Smart Agricultural Technology 2 (2022) 100034

Contents lists available at ScienceDirect

Smart Agricultural Technology

journal homepage: [Link]/locate/atech

What really impedes the scaling out of digital services for agriculture? A
Kenyan users’ perspective
John Kieti a,∗, Timothy Mwololo Waema a, Heike Baumüller b, Elijah Bitange Ndemo c,
Tonny Kerage Omwansa a
a
School of Computing and Informatics, University of Nairobi, Kenya
b
Centre for Development Research, University of Bonn, Germany
c
School of Business, University of Nairobi, Kenya

a r t i c l e i n f o a b s t r a c t

Keywords: Impediments to the scale-out of digital services for agriculture (DSAs) exist in sub-Saharan Africa (SSA). This
Digital Services is despite the promise of these information systems (IS) artifacts to unlock agricultural productivity and out-
Agriculture Sector comes. Existing literature identifies these impediments fortuitously, failing to explain the dimensionality and
Technology accessibility
patterns of association among them. To inform suitable focused effort in the digitalization of agriculture in SSA,
Value Proposition
the study set out to examine the underlying structure of impediments to DSA scaleout from a likely users’ per-
Discoverability
Policy environment spective. Employing a parallel convergent mixed-methods design, the study obtained quantitative and qualitative
evidence from likely DSA users in Kenya. Using exploratory factor analysis (EFA) on a half-split sample (n=276)
the researchers obtained a three-factor structure underlying the impediments to scale-out of DSAs. Using a sec-
ond split-half sample (n=276), the researchers undertook confirmatory factor analysis (CFA) on the three-factor
structure, yielding a good fit. The dimensions in the latent structure, christened “the three-factor structure of im-
pediments to DSA scale-out”, were technology accessibility, service discoverability, and service value proposition.
The researchers undertook thematic analysis of the qualitative evidence (n=241) and proceeded to integrate the
resulting six themes with the three-factor structure. The themes were digital skills, technology infrastructure, service
discoverability, service usability, service affordability, and the public policy environment. The researchers integrated
five of the themes into the three-factor structure. The researchers argued that the sixth theme on the public
policy environment is implicit within the three-factor structure. The study also discussed the implications of the
strong covariance between discoverability and value propositions of DSAs for practitioners. This is one of the
first research efforts to examine the structure underlying impediments to the scaling out of DSAs from a users’
perspective. It also proposes a measurement scale for such impediments. Attention to the three dimensions in
the three-factor structure can yield focussed efforts in eliminating these impediments for the increased digital
transformation of agriculture in SSA settings.

1. Introduction almost doubled from 869 US$ in the year 2000 to 1,526 US$ in the year
2019, yet the absolute numbers compare dismally with other sectors.
Agriculture is arguably the most promising sector for addressing the The sector’s highest value-added per worker attained in 2019 is tiny in
remaining poverty in Africa [1]. Studies confirm that growth in agri- comparison to the services sector values3 which remained above 6,000
culture is more poverty reducing than an equivalent amount of growth US$ since the year 2,000. The values in the industry sector4 at over
outside agriculture [2,3], more so for the poorest people in the poorest 10,000 US$ per year in the two decades period dwarf the productivity
countries [4]. Despite this potential, the sector has continued to exhibit of agriculture even further.
low productivity levels in sub-Saharan Africa (SSA) [5]. Value-added In spite of the dismal performance of agriculture, a promise exists
per worker1 for Agriculture, forestry, and fishing in SSA has steadily2 for digitalization to further unlock agricultural productivity and out-

∗
Corresponding author at: John Kieti, University of Nairobi, School of Computing and Informatics, P. O. Box 30197 - 00100 GPO, Nairobi, Kenya.
E-mail address: jkieti@[Link] (J. Kieti).
1
Value adjusted for purchasing power relative to the US Dollar in 2015.
2
[Link] - Value-added per worker in agriculture, forestry and fishing - sub-Saharan Africa.
3
[Link] - Value-added Per worker in the service sector - sub-Saharan Africa.
4
[Link] - Value-added per worker in the industry sector - sub-Saharan Africa.

[Link]
Received 14 June 2021; Received in revised form 10 January 2022; Accepted 10 January 2022
2772-3755/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license ([Link]
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

comes in SSA countries [6,7]. According to Deichmann et al. [8], the researchers argued that the focus of Kabbiri et al. was not on DSAs, and
mechanisms by which this promise can be actualized include overcom- not beyond the basic functionalities of a mobile phone network. Accord-
ing the information asymmetry challenges responsible for market in- ingly, Kabbiri et al. did not provide for scale out dynamics arising from
efficiencies in the sector. The mechanisms also include improving on- the varying characteristics of DSA providers, as observed in Ezeomah
farm productivity through information-based knowledge and extension & Duncomb [17]. Kabbiri et al. thus fell short of explaining the latent
services, as well as through the infusion of innovation in supply chain structure of impediments to the scale-out of DSAs; developed and de-
management according to the authors. The application of digital agricul- ployed as specific solutions to agriculture sector problems.
ture has also been argued to impact methods of farming and to demand The literature also falls short of explicitly addressing the perspectives
different knowledge, skills, and labor arrangements, ultimately redefin- of likely DSA users in most cases. The exceptions to this are Kabbiri et al.
ing what it means to be a farmer [9,10]. Such digitalization has been [16], Mittal et al. [18], and Gichamba et al. [19]. These studies drew di-
argued to promise hyper-transparency in the sector, shifting the roles rectly from empirical evidence among agricultural producers, laborers,
of stakeholders and broadening the boundaries of global value chains traders, commission agents, and the workers of organizations directly in-
in ways that enhance inclusivity for smallholder farmers [11]. Despite volved in agriculture. The perspective of such likely DSA users is likely
these promises, and many digital agriculture solutions being launched, divergent from the viewpoint of DSA providers, as found in Lohento &
these IS artifacts have often failed to scale out and achieve widespread Sotannde [20] and Crandall & Kieti [21]. Moreover, the co-mingling of
usage in SSA [6,8]. these perspectives as secondary evidence in literature reviews can have
Kieti et al. [12] defined a Digital Service for Agriculture (DSA) as “a a blurring effect on the specific concerns of likely DSA users for targeted
solution that uses digital equipment and devices such as mobile phones, com- intervention. For instance, the researchers argue that some impediments
puters, satellites, and sensors to solve challenges in agriculture”. The study to DSA scale-out such as high customer acquisition costs and difficulty
adopts this definition while concurring with Han et al. [13] who cau- raising investments [8,21,22] are unlikely to manifest in the likely DSA
tion against conflated use of the term digital platform. The researchers users perspectives. Such DSA users are not decision-makers in marketing
thus acknowledge a distinction between DSAs, and digital platforms for the DSA or funding it as investors. Moreover, the categories of impedi-
agriculture (DPAs) as defined in Kieti et al. [ [12] and expected to ex- ments to DSA scale-out: technology and market concerns as derived from
hibit intermediary logic ( [14,15]]. The focus of this study is not DPAs the literature may not be considered exhaustive by likely DSA users.
per se, but the DSAs which may be constituent components of a DPA Therefore, the researchers found no conceptual clarity on the underly-
[71] or may thrive as stand-alone digital artifacts. Understanding the ing structure of impediments to the scale-out of DSAs in the literature,
latent structure of impediments to the scale-out of DSAs has the poten- especially from the viewpoint of likely users.
tial to inform the focusing of interventions to unlock agriculture’s digital
transformation in SSA. This includes identifying low-hanging fruits for
3. Materials and Methods
accelerating deployment and the impacts of DSAs.
As discussed in section 2, the researchers found a paucity of literature
3.1. Research Design
examining impediments to DSA scale-out comprehensively enough to
explicate the underlying dimensions and patterns of associations among
This study was part of a larger cross-sectional research about digi-
them. This is despite the numerous publications suggesting such im-
tal platforms for agriculture. As justified in [12], the researchers chose
pediments, albeit fortuitously to their study objectives in most cases.
Kenya, the “SSA country setting with the most deployments of DSAs and
Informed by this gap, the objective of this research was to ascertain
with a heightened economic significance of agriculture” as the context of
the underlying structure among impediments to the scale-out of DSAs
inquiry. The choice of Kenya instantiates the SSA setting implied in the
from the perspectives of likely DSA users to inform focused efforts. The
research questions. This study was not guided by a predetermined theory
main research question for this study was: “What is the dimensionality and
or model. This is due to the lack of theoretical clarity on the underlying
pattern of association among impediments to DSA scale-out as perceived by
structure of impediments to the scale-out of DSAs among likely users,
likely users in an SSA setting?” (RQ1). To directly capture the sentiments
as described in sections 1 and 2. The study applied existential abduc-
of likely DSA users for consideration in the inquiry, the study sought
tion using Exploratory Factor Analysis (EFA) and Confirmatory Factor
to answer a second research question: “Which themes describe how im-
Analysis (CFA) for theory generation on the underlying structure and for
pediments to DSA scale-out are perceived by likely users in an SSA setting?”
theory testing, respectively [23,24]. Complementarily, the study applied
(RQ2).
an inductive approach using thematic analysis on qualitative evidence
In the next section, the researchers report on a survey of past rele-
for theoretical insights [25,26]. Consequently, the study design was of
vant work on impediments to DSA scale-out to demonstrate the gaps in
parallel convergent mixed methods, placing the researchers on a path
extant literature. The article then describes the materials and methods
towards heightened knowledge and validity [27]. Fig. 1. illustrates this
used in the study under section 3. Thereafter the article presents results
research design.
separately from the discussions under sections 4 and 5 respectively. The
article ends with conclusions in section 6.
3.2. Instruments and Data collection
2. Literature Review: Impediments to DSA scale-out
Similarly to Kieti et al. [12], likely DSA users can be defined as “in-
To build on existing work, the researchers surveyed the literature dividuals engaged in activities along agricultural value chains with access
resulting from the search string “(digital OR ICT) AND (agriculture OR to computing devices and the internet or basic data connectivity”. The re-
farmer) AND (impede OR limit OR obstacle OR hinder OR barrier)” on searchers conceived ten statements, five on market concerns and five
Google Scholar and ScienceDirect. The researchers filtered out publica- on technological concerns, for a survey targeting these likely users. This
tions dated earlier than 2010, prioritizing the results by recency and design drew from the literature synthesis and the researchers’ experi-
relevance to the topic and the SSA context. The authors included ad- ence in the convergence of agriculture and digital technologies. The sur-
ditional relevant literature upon applying a snowballing strategy. The vey instrument also included open-ended questions for additional qual-
literature synthesis yielded two categories of obstacles to the scale-out itative evidence. The researchers further refined the instrument with
of DSAs, namely, technology concerns and market concerns, as outlined the inputs of agriculture and digital services experts, deployed on the
in Table 1. SurveyMonkey platform, and pre-tested in a focus group setting among
Kabbiri et al. [16] undertook an inquiry into the underlying struc- farmers as described in Kieti et al. [12]. Among the resulting refinements
ture of obstacles to the adoption of mobile phones for agriculture. The was simplification for online self-administration, reducing options for

2
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

Table 1
Literature survey on impediments to DSA scale-out.

Category Impediment Sources

Technology concerns Internet, and other connectivity challenges especially in rural Ogutu et al. [7]; Karippacheril [65]; Saidu et al. [66]; Lohento
areas (uneven coverage and unreliable or costly networks) & Sotannde [20]; Kos & Kloppenburg [11]
Choice of technology deployment channels given variable Mittal et al. [18]; Karippacheril [65]; Gichamba et al. [19];
capabilities of computing devices and their affordability Aker et al. [6]; Rumanyika et al. [67]; Kansiime et al. [68]
Electricity for charging computing devices being unreliable or Mutula [45]; Jain et al. [69]; Aker et al. [6]; Saidu et al. [66];
costly with uneven coverage Ezeomah & Duncomb [17]
Low education levels and language difficulties Qiang et al. [60]; Okello et al. [55] Karippacheril [65]; Jain
et al. [69]; Gichamba et al. [19]; Aker et al. [6]; Saidu et al.
[66]; Kabbiri et al. [16];
Market concerns Low awareness among target users; and low affordability of Mittal et al. [18]; Crandall & Kieti [21]; Deichmann et al. [8];
customer acquisition costs among providers Fabregas et al. [22]; Kim et al. [59]
Sectoral fragmentation, DSAs lacking comprehensiveness, Qiang et al. [60]; Batchelor et al. [70]; Karippacheril [65];
missing one-stop-shop, missing ecosystem-wide support and Deichmann et al. [8]; Gichamba et al. [19]; Ezeomah &
collaboration and among actors Duncombe [17]; Lohento & Sotannde [20]; Kim et al. [59]
Transaction costs and costly or incompatible revenue models Qiang et al. [60]; Okello et al. [55]; Karippacheril [65];
Gichamba et al. [19]; Fabregas et al. [22]
Unconvincing usefulness and mismatch of value expectations; Mittal et al. [18]; Wyche & Steinfield [64]; Deichmann et al.
including user experience, trust, reliability, efficiency, and [8]; Aker et al. [6]; Saidu et al. [66]; Baumüller, [54]; Kabbiri
action orientedness et al. [16]; Ezeomah & Duncomb [17]; Lohento & Sotannde
[20]; Kos & Kloppenburg [11]

Fig. 1. The research design.

the Likert-like scale questions from five to three: - “I disagree”, “I nei- tractable factors using parallel analysis [30], the superior sample-based
ther agree or disagree’’, and “I agree”. Appendix A shows the instrument alternative to the more common Kaiser greater-than-1 criterion [31].
arising from these refinements. The study used the Kaiser-Meyer-Olkin (KMO) to ascertain sampling
The researchers broadcast the questionnaire as a weblink to adequacy on whether the data could be grouped into smaller sets of
over 300,000 likely DSA users, leveraging the flourishing of virtual underlying factors [28], whereby a score above 0.6 is acceptable [32].
agriculture-themed communities online. This process obtained 887 re- Complementarily, the study used Bartlett’s Test of Sphericity, with a
spondents overall, with 785 remaining valid after data cleaning, as Ki- chi-square statistic p<0.001 to indicate appropriateness to apply EFA.
eti et al. [12] described. The participant inclusion criteria for this pa- The researchers used the Principal Component Analysis (PCA) extraction
per were either responding to all ten close-ended statements (552) or method to determine the respective loadings in the EFA. They chose it
providing rich qualitative insights via the relevant open-ended ques- over Maximum Likelihood (ML) because they measured the items on an
tions (241). Respondents meeting these overlapping criteria totaled 582. ordinal scale, and the data defied the assumption of multivariate nor-
Appendix B shows a summary of the respondents. mality. The researchers rotated the resulting matrix to clarify loading
patterns in the EFA and ease of interpretation. Initially, an oblique rota-
tion method, oblimin, was used whereby the inexistence of values above
3.3. Factor Analysis and Thematic Analysis 0.32 in the inter-factor correlation matrix would imply the absence of a
strong correlation [32] hence the use of an orthogonal rotation method;
To apply EFA on the 10-item data, a minimum of 100 observations varimax. The study identified highly cross-loading items by evaluating
were required to attain the recommended threshold ratio of 10:1 [28]. the ratio of their squared loadings and eliminated them from the anal-
The sample (N=552) was five times more than adequate by this thresh- ysis if the ratio fell below 2.0 [28]. The study dropped Variables with
old. In the ideal situation, a researcher should split a sample, using one loadings less than 0.30 from the factors and considered loadings greater
half to examine and develop the structure and the other half to validate than 0.50 as practically significant [28].
the structure [29]. The researchers randomly divided the quantitative On the second split-half sample (n=276), the study employed Confir-
data into two split‐half samples (n = 276), adequate to use on the EFA matory Factor Analysis (CFA) to obtain and validate the factors associ-
by the 10:1 criterion. The researchers determined the number of ex- ated with impediments to the scale-out of DSAs as hypothesized from the

3
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

Table 2
EFA of impediments to the scaling out of DSAs.

EFA1 EFA2

KMO 0.81 0.79

Bartlett’s Test of Sphericity 𝜒2(45)=709.56 𝜒2(36)=603.83
p<0.001 p<0.001
Total Variance Explained 57% 61%
Measured Items F1 F2 F3 F1 F2 F3

T1 - Requiring high literacy levels 0.68 0.74

T2 - Requiring access to the internet 0.84 0.87
T3 - Requiring the use of smartphones rather than more affordable phones 0.81 0.83
T4 - Requiring phones to be always charged 0.59 0.30 0.54
T5 - Incurring expensive data and SMS charges 0.58 0.46 N/A N/A N/A
M1 - Awareness about them is low 0.64 0.65
M2 - There is no one-stop-shop where all of them can all be found 0.84 0.89
M3 - They do not work together or share data with each other 0.66 0.34 0.66 0.32
M4 - Their charges are too high to keep paying 0.81 0.78
M5 - The value they offer their target customers is not convincing 0.71 0.84

Extraction method = PCA; Rotation method = Oblimin

EFA. The study used the diagonally weighted least squares (DWLS) es- Table 3
timator method rather than Maximum Likelihood (ML) as it was specif- Measurements for Reliability and Construct Validity.
ically designed for ordinal data and was superior to Robust ML (MLR) HTMT ratio
in the estimation of factor loadings [33]. The researchers used the Com- Construct Items CR AVE F1 F2 F3
parative Fit Index (CFI), Tucker-Lewis index (TLI), Root Mean Squared
F1 - Technology accessibility 4 0.84 0.57 1.00
Error of Approximation (RMSEA), and Standardized Root Mean Squared F2 - Service discoverability 3 0.78 0.55 0.42 1.00
Residual (SRMSR) to assess the goodness of fit. CFI>0.90, TLI>0.90, F3 - Service value proposition 2 0.80 0.66 0.39 0.82 1.00
RMSEA<0.08, and SRMSR<.08 indicate a good fit [34,35]. For EFA and
CFA the researchers used R [36] as it was affordable, being accessible
on open-source licensing.
For the 241 open-ended responses, the researchers conducted the-
ratio of 1.6. Therefore the researchers conducted a second EFA (EFA2)
matic analysis using version 9 of [Link] [ [[37-2021] software for Ap-
without this item. This increased the total variance explained from 57%
ple MacBook computers. The researchers followed the seven-step pro-
to 61% and left no more items highly cross-loading. The three factors
cedure proposed by Friese et al. [38], which adapted the six-step pro-
extracted were conceptualized as F1 - technology accessibility, F2 - discov-
cedure in Braun & Clarke [39] for thematic analysis. The study then
erability, and F3 - value proposition and are discussed in sections 5.1,5.2,
integrated the qualitative findings as themes elaborated and aligned to
and 5.3, respectively, and together as a structure in section 5.5.
the dimensions confirmed in the factor analysis.
The researchers used the second half of the data (n=276) for CFA
4. Results to examine how well the data fit the three-factor structure emerging
from the EFA. As indicated in Table 3, the researchers ascertained the
Among the participants, the majority were male at 75%. The median internal consistency of the three subscales with the Composite Reliabil-
age range was 24-34, and 73% had attained a diploma certificate or ity (CR) measurements being 0.70 or higher, and below 0.95 [28,40].
higher education level. Reliance on agriculture for income among the This reliability and the Average Variance Extracted (AVE) being above
participants was 83%. Primary agricultural production activities such 0.50 for all three subscales meant that convergent validity was estab-
as rearing animals and growing crops were the most represented value lished [41]. For discriminant validity, the study used the Heterotrait-
chain activities at 85%. A detailed breakdown of the sample profile is Monotrait ratio (HTMT) of the correlations, which is superior to the
in appendix C. commonly used Fornell-Larker criterion [42,43]. Discriminant validity
was established under the HTMT0.85 criterion as the HTMT ratios were
4.1. The underlying structure below 0.85.
The standardized factor loadings estimated in the CFA are shown in
The quantitative ratings of the items on impediments to DSA scale- Fig. 2. With the CFA, the study obtained 𝜒2(24)=19.23; p-value=0.740.
out were not normally distributed as the study obtained the Henze- This meant a Chi-square/Degrees of freedom ratio of 0.80; suggesting
Zirkler statistics of 8.5, p-value<0.001. The researchers ascertained an acceptable fit. Further, the study obtained fit indices CFI=1.000,
three extractable factors using parallel analysis on the first split-half TLI=1.010, RMSEA=0.000, and SRMSR=0.055 which indicated this sec-
sample (n=276). The study obtained KMO 0.81 and 0.79 and Bartlett’s ond half-split of the data fit the three-factor structure well. A nine-item
test of sphericity p<0.001 as shown in Table 2, indicating the appro- measurement scale with three dimensions of impediments to the scale-
priateness of the EFA procedure. Table 2 also shows pattern matrices out of DSAs was established in line with Fig. 2.
obtained with factor loadings rotated using the oblimin method. In the With a factor loading of 0.80, “T2 - Requiring internet access” was
first EFA (EFA1), “T4 - Requiring phones to be always charged ” cross- the most impactful indicator item on technology accessibility as an im-
loaded on F1 (0.59) and F3 (0.30) but not highly as the square of the pediment to DSA scale-out. With all factors remaining constant, a unit
loadings yielded a ratio of 3.8, being higher than the 2.0 threshold in increase in this indicator increases technology accessibility as an imped-
Hair et al. [28]. Likewise, “They do not work together or share data with iment by 80%. Likewise, “M3 - Low collaboration” was the most impactful
each other” cross-loaded on F2 (0.66) and F3 (0.34) but not highly with indicator on discoverability at 0.76. The strongest covariance in the CFA
the square of the loadings yielding a ratio of 3.7. However, the item “T5 was between discoverability and value proposition at 0.83. As such, in-
- Incurring expensive data and SMS charges” cross-loaded highly on fac- tervening for a unit change in DSA discoverability impacts DSA value
tors F1 (0.58) and F3 (0.46), with the square of their loadings yielding a proposition by 83% and vice versa.

4
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

Fig. 2. CFA of 3-factor structure of impediments to DSA scale-out (n=276).

4.2. Themes from qualitative evidence tapped for expanding the uptake of DSAs. The respondents considered
young farmers not to appreciate the power of DSAs.
To answer the research question - “Which themes describe how imped-
iments to DSA scale-out are perceived by likely users in an SSA setting?”,
the researchers undertook a thematic analysis of the respondents’ qual-
itative responses. This yielded six themes, namely: (a) technology in- 4.2.2. Technology Infrastructure
frastructure, (b) digital skills, (c) service discoverability, (d) service us- The qualitative responses included difficulties in accessing technol-
ability, (e) service affordability, and (f) the policy environment. The ogy infrastructures such as the internet and basic mobile network cov-
thematic map in Fig. 3 summarises these themes and their subthemes. erage as hampering the uptake of DSAs. The respondents noted that
geographical inequality in technology access was more pronounced in
remote and rural areas where most agricultural production occurs. The
4.2.1. Digital Skills
qualitative responses linked the technological infrastructure limitations
The qualitative evidence indicated that general literacy per se was
to electricity access and mobile network coverage:
not crucial as an impediment to the scale-out of DSAs. The respondents
deemed literacy levels high, with educated people venturing into agri- “DigiFarm may sometimes not operate in remote areas due to network in-
culture after failing to secure professional jobs and de-emphasizing chal- accessibility” ∼ male respondent aged 18-24 year involved in agricultural
lenges of general literacy. Nonetheless, they considered farmers to lack production.
operational digital skills in using smartphones even when they could
afford such devices. On the one hand, the respondents indicated that large-scale farm-
ers can access advanced smartphones, including those with enhanced
“... most of the players in agriculture are rural farmers who most of the
battery life. On the other hand, they argued that small-scale farmers
time lack the digital literacy to use smartphones” ∼ Female respondent
could only afford feature phones for essential communication and mo-
aged 25-34 providing specialised advisory services to farmers.
bile money services. The respondents desired universal access to DSAs
The respondents considered the unavailability of trainers to facil- through low technology channels such as Short Message Service (SMS),
itate digital skills enhancement among farmers an impediment. They Unstructured Supplementary Service Data (USSD), and Interactive Voice
indicated younger people as adaptable to digital technology and more Response(IVR), not needing smartphones. Moreover, they suggested
strategic to involve than older people in the scale-out of DSAs. However, that deployment channels other than mobile phones diversify access to
the enthusiasm for digital services among young people was deemed un- DSA benefits.

5
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

Fig. 3. Thematic map as generated from qualitative evidence (n=241).

4.2.3. Service Discoverability expressed concerns that DSAs offered broad-based solutions with little
The qualitative responses indicated low discoverability of DSAs as a demonstrable value at a user-specific level. The respondents noted that
limitation to their scale-out. They mainly attributed the low discover- nuanced preferences and niches remained unaddressed by DSAs, in ex-
ability of DSAs to ineffectiveness among the awareness creation chan- amples of unmet expectations of solution specificity. Using DSAs was
nels used and the absence of collaboration among players in the digi- indicated as time-consuming, yet the work of farmers was deemed too
tal agriculture ecosystem. The respondents observed that farmers were laborious to afford distractions of using DSAs. Moreover, the respon-
likely to own and use smartphones for communication via social me- dents expressed unease about user safety in otherwise useful DSAs. The
dia platforms yet remain blind-sided on the opportunity to use DSAs. As fears included concerns about protecting user privacy, being hacker-
such, the farmers were unaware of DSAs accessible through the same proof, and detecting or eliminating abuse by fellow users:
technological infrastructure they were familiar with:
“ scammers are very many and have distorted the true platforms from the
“Lack of exposure is one of the biggest reasons for not using digital plat- fake ones, this has led to mistrust”. ∼ male respondent aged 35-44 years
forms. I use my smartphone a lot but mostly WhatsApp. How do I know involved in agricultural production
there are other apps that can teach me better agribusiness?” ∼ female
respondent aged 55-64 years involved in agricultural production 4.2.5. Service Affordability
Expensive charges for data connectivity to access DSAs, although
The respondents went further to suggest an array of possibilities to
not payable to DSAs, were deemed to undermine the scaling out of
market DSAs effectively. These included advertisements on websites and
DSAs. Likewise, the respondents considered the cost of smartphones
social media platforms, including the use of social media influencers.
to undermine the affordability of DSAs and their scale-out. Beyond
They also suggested offline channels such as newspapers, journals, ra-
the cost of data and devices, the respondents suggested free access to
dio, billboards, school education curricula, and TVs. In-person demon-
DSAs as websites, mobile applications, and other deployment channels.
stration of value was deemed effective for the persuasion of farmers to
They expected Government agencies, in particular, to provide essential
use DSAs. The inexistence of one-stop-shop platform for accessing DSAs
information-based DSAs for free. While the respondents indicated reluc-
was deemed to not only undermine their discoverability. It was depicted
tance among specialists to provide services for free, they called for fair
to undermine the efficient use of time by agricultural actors as well. Low
pricing of DSAs, including avoiding undue profiteering.
collaboration at the peer level among farmers was observed, character-
ized by the prioritization of individual-level effort over the collective “There should be more customer-based / oriented solutions at a fair price,
effort in the usage of DSAs. This, and weak collaboration among other not (only) after making fame and profit” ∼ male respondent aged 25-34
stakeholders in agriculture were depicted to hinder the growth and up- years engaged in agricultural production
take of DSAs.
4.2.6. Public Policy
4.2.4. Service Usability Respondents faulted the Government for not adequately supporting
The scaling out of DSAs was depicted as constrained by their low the digitalization of agriculture. They particularly faulted the national
usability. For this, the respondents expressed unmet user expectations, and county governments for not leveraging digital services to enhance
user experience friction, and user safety concerns. They faulted DSAs their input subsidization programs and not adequately supporting the
for not being dependable in fulfilling their promises. The respondents use of drones in agriculture. The respondent also faulted governments

6
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

for not supporting the marketing and selling of agricultural produce on- mental access barriers characterizing the digital divide [48]. Hargittai
line and not enhancing markets with transparent and timely informa- [49] posited these mental access barriers as more manifest among older
tion on commodity prices. The respondents faulted government-backed people than younger people. The researchers contend that mental access
research institutions for not adequately disseminating the information barriers undermine acquiring digital skills and the requisite self-efficacy
they hold. The respondents went further to call for Government invest- to use technology, comprising technology accessibility impediments to
ment to promote transparency of commodity prices and support farmers the scale-out of DSAs. This study holds technology accessibility as a major
in marketing their commodities using DSAs. The respondents also sug- dimension of impediments to DSA scale-out encompassing the technol-
gested that Governments should finance DSAs: ogy infrastructure and digital skills themes generated in the qualitative
evidence.
“They (DSAs) should get government funding so that they get their ser-
A DSA provider may seek to address technological infrastructure ac-
vices down to the lowest farmers.” ∼ male respondent aged 45-54 in-
cess concerns among its likely users through “low-tech” deployments.
volved in agricultural production.
These include basic phones with USSD and SMS technologies where ac-
In addition, the respondents called upon Governments to facilitate cess to the internet and smartphones is limited. The researchers contend
training, sensitization, and technology access among agriculture stake- that these technological infrastructure concerns are of a broader scope
holders to eliminate impediments to the scale-out of DSAs relating to national infrastructure provision and regulation, beyond the
purview of the DSA provider. Likewise, while a DSA may provide tu-
5. Discussion torials and technical support functions to supplement operational skills
among its users, digital skills access as an impediment to DSA scale-
For the first research question, the study ascertained three dimen- out is best addressed as a macro-level concern. DSA providers lack the
sions and their patterns of association by performing factor analy- resources and the influence to increase digital skills at a large scale,
sis on two half-split samples and presented the results in section 4.1. an intervention in the purview of policymakers and other actors in the
For the second question, the researchers generated the six themes in larger digital agriculture ecosystem. The researchers position technology
section 4.2 from the qualitative evidence. accessibility as an impediment to DSA scale-out in the first and second
This paper’s contribution to the literature on DSA scale-out is levels of the digital divide and contend that it requires interventions at
twofold. Firstly, the study integrates the qualitatively generated a broad level beyond the efforts of individual DSAs.
themes with the structure arising from the quantitative findings. In
subsections 5.1,5.2, and 5.3, the researchers show that five of the themes 5.2. Service Discoverability
align with the dimensions of the three-factor structure while highlight-
ing their relation to the existing literature on the digital divide and busi- This dimension was manifested highly in three items in the CFA
ness model design. In sub-section 5.4, the researchers argue how the namely:- “M3 - Low collaboration”, “M2 - no one-stop-shop”, and “M1 -
sixth theme is implicit in the three-factor structure. The second theo- Low awareness”. The quantitative evidence on this dimension directly
retical contribution in subsection 5.5 is to propose a three-factor struc- aligns with the corresponding theme in the qualitative evidence. Low
ture explaining the dimensionality and patterns of association among awareness about DSAs was consistent with the difficulties among DSA
impediments to DSA scale-out, including a measurement scale for such providers in creating awareness of their services [21,22]. The weak
impediments. collaboration observed among actors in the digital agriculture ecosys-
tem was consistent with the fragmentation [8] and lack of ecosystem-
5.1. Technology Accessibility wide collaboration support among actors [20]. The findings indicate
an extended collaboration concern to cover not only DSA providers
Three items loaded highly in the CFA to technology accessibility, con- complementing the work of network operators and device providers
noting physical access barriers to the scale-out of DSAs. The items are [50] among the high-level stakeholders. Collaboration among end-user
“T2 - Requiring internet access”, “T3 - Requiring smartphone use”, and “T4 actors called for as an intervention area for increased scale out of DSAs
- Requiring phones always charged”. The essence of these indicators fea- is noteworthy. It is indicative of the awareness creation potential of
tures prominently in the technology infrastructure theme in the qualita- end-user collaboration and the potential for data-level collaboration as
tive evidence. The researchers contend that these three items consti- with agricultural data cooperatives and coalitions in the clamor for data
tute the technological infrastructure aspect of technology accessibility as sovereignty [51]. The absence of a one-stop-shop contributing to DSA
an impediment to DSA scale-out. The researchers also argue that the scaleout impediments as evidenced in this study is consistent with the
prominence of technology infrastructure aspects in the evidence depicts clamor for comprehensiveness among DSAs and DPAs [17,19]. The re-
the significance of the first-level digital divide, which comprises mate- searchers further argue that service discoverability as an impediment to
rial access barriers [44,45]. The respondents’ sentiments of rural and the scale-out of DSAs relates to the channels and partnerships components
remote areas suffering unreliable network connectivity are consistent of the DSA business model [52]. Low awareness relates to the channel
with the identification by Ernest III et al. [46] of geography as one of effectiveness, while the collaboration or its absence between the DSA
the demographic dimensions of the digital divide. provider and other digital agriculture ecosystem actors relates to part-
The researchers relate the high factor loading in the CFA of the fourth nerships. The DSA provider may forge new or join existing partnerships
item, “T1 - Requiring high literacy levels” on technology accessibility, that contribute to its enhanced visibility and discoverability.
to the significance of the second-level digital divide [47] in impeding The researchers contend that large corporations can leverage their
DSA scale-out. At this level of the digital divide, skills access barri- marketing budgets and distribution networks to overcome service dis-
ers impede the use of technology even if material access barriers no coverability challenges of the DSAs they acquire, develop or advance in
longer exist. However, the researchers observe that general literacy per partnership with digital entrepreneurs. Moreover, these large organiza-
se is de-emphasized as an impediment to DSA scale-out in the qualita- tions with a global or nationwide presence, including mobile network
tive evidence under the digital skills theme. Instead, the respondents operators, device manufacturers, software platforms, and agricultural
emphasized operational digital skills among targeted users as missing inputs firms, are considered major drivers of digital agriculture [53].
enablers of DSA scale-out. The researchers contend that literacy is a Further, the researchers argue that an Aggregator Platform for Digital
proxy indicator of the required digital skills as it unlocks the under- Services in Agriculture (AP4DSA), as conceptualized in Kieti et al. [12],
standing of DSAs and how to apply them to ones’ tasks. Furthermore, can address service discoverability concerns. An AP4DSA can be a col-
the researchers link the observed high enthusiasm among young people laboration platform and a one-stop-shop where all DSAs may be located,
and negative attitudes among older people regarding technology to the accessed, reviewed, and rated. Such a digital platform can be a market-

7
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

place for DSAs. It can foster openness, collaboration, and competition 5.4. Public policy as an implicit theme
among digital agriculture ecosystem actors in ways that enhance DSA
benefits to populations at the base of the pyramid [50]. The collabora- This theme generated from the qualitative evidence was consistent
tion can include actor intermediation and automated sharing of data, with extant literature on the important role played by the policy and
subject to appropriate data protection provisions. Such collaboration regulatory environment in e-business diffusion [62]. Moreover, the im-
can constitute alliances of specialized DSAs, and mobile network op- portance of this role has been argued to be more pronounced in de-
erators brought together by entrepreneurial support programs, as with veloping than developed countries [57]. The findings tally with Birner
the case of Mercy Corps’ AgriFin accelerator5 giving rise to the DigiFarm et al. [53], who argue for public efforts to leverage private-sector ef-
platform in Kenya. Moreover, the thriving of an AP4DSA can lead to forts in digital agriculture for a vibrant industry with benefits distributed
enhanced awareness creation about its constituent actors. This is more fairly among actors in developing countries. This theme was not explic-
so if the AP4DSA design minimizes search costs related to discovering itly manifest in the three dimensions resulting from the factor analysis.
and using constituent DSAs [12]. The researchers add a caveat that an However, the researchers argue that public policy concerns are implicit
AP4DSA can independently suffer the impediments of “M1 - Low aware- across the three dimensions. For instance, technology accessibility as an
ness” and “M3 - Low collaboration” if its implementation does not con- impediment to the scaling out of DSAs can be minimized through in-
sciously address them. For instance, an AP4DSA not collaborating with creased public policy interventions to improve the telecommunications
other actors in a digital agriculture ecosystem undermines any leverage infrastructure for universal access and to reduce the digital divide [58].
it may have to foster such collaboration among its constituent DSAs. This is especially in rural areas where most primary agricultural pro-
duction occurs, as suggested in Aker et al. [6] and concurring with Kim
et al. [59]. Technological infrastructure access interventions by Govern-
5.3. Service Value Proposition
ments could also be through rural ICT community centers as suggested
by the respondents. Promoting the large-scale acquisition of digital skills
The high factor loading of “M4 - charges too high” on Service Value
through public formal education systems can also enhance Technology
Proposition as an impediment to DSA scaleout can be linked to the Ser-
accessibility. Public policy investment in an AP4DSA through the fund-
vice affordability theme generated in the qualitative evidence. Service
ing, coordination or public endorsement of such a platform can also min-
affordability is a manifestation of the income demographic dimension
imize Service discoverability as an impediment to DSA scale-out. In con-
of the digital divide [46]. As such, likely users may not always afford
currence with Kim et al. [59], governments can also invest in enhanced
fees charged by the DSA despite compelling value. This study’s finding
awareness creation and distribution frameworks for existing beneficial
is consistent with Gichamba et al. [19], who found that in low resource
DSAs. Governments can also promote collaboration among actors to ad-
settings, agricultural actors preferred low-cost DSAs that they could af-
vance initiatives that produce common good benefits such as open data
ford even during seasons of financial hardship. Furthermore, the ability
and data cooperatives. Subsidizing DSA costs to reduce service afford-
of DSAs to charge their users, more so for information-based DSAs, is
ability challenges for essential DSAs can minimize impediments related
undermined in economic theory by information as a commodity being
to their Value proposition. Fabregas et al. [22] uphold the case for Gov-
non-rival [22]. The researchers further argue that service affordability
ernment subsidization of DSAs as social investments given the expected
as an aspect of service value proposition relates to the cost structure and
difficulties for providers charging for information-based DSAs. From a
revenue streams components of a DSA’s business model [52]. Moreover,
DSA provider perspective, Lohento & Sotannde [20] argue that funding
affordability relates to whether the DSA can collect revenue from deliv-
limitations hinder DSA success. The authors contend that public policy
ering the service enough to cover their costs and earn a surplus.
can inform the funding of DSAs by the Government as suggested by the
Similarly, service value proposition as manifested in the high factor
respondents. This could be in financial and talent development support
loadings for the item “M5 - value not convincing” can be linked to the
to DSA providers to enhance service usefulness hence their value propo-
Service usability theme in the qualitative evidence. The service usability
sitions. Regulatory and enforcement mechanisms for data security and
theme is indicative of the prospects of the user to derive value from using
the elimination of user safety-related barriers can enhance the value
the DSA. Such value has to be compelling enough to exceed the costs and
propositions of DSAs. Moreover, this study concurs with Qiang et al.
inconveniences of accessing it. According to the qualitative evidence,
[60] on the need for regulatory forbearance in digital agriculture. This
impediments to DSA scale-out in the form of impaired service usefulness
can include granting space for DSAs providers to experiment with new
such as the lack of solution specificity, user-experience friction, and lack
services, be innovative, and be rewarded for taking risks. Furthermore,
of user safety make a DSA’s value unconvincing. This finding concurs
policy and regulatory interventions are also needed to manage potential
with past studies where usefulness of DSAs was not always convincing
threats to digital agriculture, such as market imperfections [58,59] and
[6,17,54]. It is also consistent with the characterization of the digital
unhealthy dominance by agricultural input providers [53].
divide to comprise usage-related barriers [44,48].
Consequently, Service value proposition as an impediment to DSA
5.5. The proposed three-factor Structure
scale-out is an evaluation of benefits versus the costs incurred by users
beyond overcoming the initial technology accessibility and service dis-
In answering the first research question of this study, the researchers
coverability impediments. It relates directly to the value proposition com-
submit their “the three-factor structure of impediments to DSA scale-out” il-
ponent in a business model by the taxonomy in Osterwalder & Pigneur
lustrated in Fig. 2 for evaluation and use by researchers and practition-
[52]. The researchers contend that a cyclic relationship exists between
ers alike. The researchers explained the three dimensions of technology
the usefulness and affordability of DSAs as the value derived by users
accessibility, service discoverability, and service value proposition by inte-
can lead to increased income generation and economic empowerment,
grating the qualitatively generated themes in the preceding subsections.
hence increased affordability. Such increased service affordability yields
The study established the unidimensionality of variables in the three-
increased demand and value derivation [55], hence usefulness. The re-
factor structure after dropping the item “T5 - Incurring expensive data
searchers contend that such a virtuous cycle leading to increased inten-
and SMS charges”. A probable reason for this item’s highly cross-loading
sity of DSA access and beneficial use can bridge the third-level digital
on technology accessibility and service value proposition is that it combined
divide on technology impact as well as address usage access barriers in
cost aspects of physical infrastructure and individual DSA pricing. It de-
the second-level of the digital divide [47,56].
picts the inability among users to distinguish internet data and SMS costs
charged by network providers from the extra costs charged by DSAs on a
5
[Link] one-off or recurring basis. The authors contend that while all these costs
made-for-farmers/. remain implicit in both technology accessibility and service value proposi-

8
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

tion, fees charged by a DSA provider explicitly relate to their service’s provide a reference point for further research analyzing impediments to
affordability, hence uniquely impacting its value proposition. DSA scale-out to unlock the promise of DSAs in SSA settings. The au-
The authors also propose adaptation and use of the three-factor, thors recommend the incorporation of the qualitative findings in fur-
nine-item measurement scale for impediments to DSA scale-out. This is ther adapting or enhancing the measurement scale. Further research
building on the reliability and construct validity ascertained in Table 3. may also ascertain whether the qualitatively generated theme of the
Guidelines by Hair et al. [28] suggest three or more manifest items per public policy can stand on its own as a fourth independent dimension.
factor for maximized construct reliability and adequate coverage of the This study’s findings can be analytically replicated in other human de-
construct’s theoretical domain. Hair et al. [28] also encourage the use velopment sectors such as education and health. The researchers also
of the least number of items for parsimony. On the one hand, the two recommend further research into the prospects of AP4DSAs to facilitate
items in the structure covering the value proposition result in adequate organizing and consolidating DSA benefits in SSA settings at the digital
composite reliability at 0.80. This compares favorably with the gener- agriculture ecosystem level.
ally recommended composite reliability above 0.70 and not above 0.95
in Hair et al. [28]. Conversely, it can be argued based on the quali-
tative evidence that the item “M5 - value not convincing” may not ad- 6. Conclusion
equately cover aspects of service value proposition such as user safety,
user-experience friction, dependability, and service relevance. The re- This study is one of the first research efforts examining the under-
searchers contend that their consolidation of such concerns into “M5 lying structure of impediments to the scaling out of DSAs from a users’
- Value not convincing” avoids creating sub-factors, possible with inad- perspective. It uses parallel convergent mixed methods, including fac-
vertent focus on the service usability aspect of “value proposition”, as tor analysis and thematic analysis, combining inductive and deductive
cautioned against in Hair et al. approaches, to pursue conceptual clarity. Quantitatively, the study as-
The strong and positive covariance between service discoverability certained a three-factor structure underlying impediments to DSA scale-
and service value proposition is noteworthy. A probable reason is that out. The three dimensions of the structure are “technology accessibility”,
increased discoverability can result in expanded usage, network effects, “service discoverability”, and “service value proposition”. Qualitatively,
economies of scale, and user feedback which can, in turn, enhance the the study identified six themes on impediments to DSA scaleout. These
service’s value offerings. Improved value propositions of DSAs can also themes include “technology infrastructure” and “digital skills” which
gain social validation such as media coverage, and the virality likely correspond to the technology accessibility dimension in the three-factor
with peer-to-peer referrals among gratified users, leading to increased structure. Other themes are “service usability” and “service affordabil-
discoverability. Moreover, these strongly co-varying dimensions are typ- ity” which correspond to the “service value proposition” dimension. The
ically in the DSA provider’s purview since they constitute aspects of the fifth theme named “service discoverability” aligned directly with the
provider’s business model. As discussed in section 5.1, this is not neces- corresponding dimension in the three-factor structure. The researchers
sarily the case with technology accessibility. argued that the sixth theme of “public policy” is implicit within the
three-factor structure. The researchers also proposed a scale for mea-
5.6. Implications for policy and practice suring impediments to DSA scaleout based on the ascertained structure.
With the increased conceptual clarity through the three-factor struc-
Policymakers can use the findings of this study to prioritize and des- ture, interventions in policy and practice can be suitably crafted for fo-
ignate intervention areas by addressing environmental aspects impeding cus and effectiveness as suggested in sections 5.4 and 5.6. This is more so
the scale-out of DSAs in SSA countries. As elaborated in section 5.4, pol- to reinvigorate efforts in Kenya and other SSA countries, unlocking the
icy and regulatory interventions should address impediments implicit promise of digital transformation for agriculture, a sector whose growth
in the three dimensions of the three-factor structure. Increasing digital offers the promise to eliminate poverty in low and middle-income coun-
inclusivity by eliminating access barriers is implied as a policy interven- tries.
tion for technology accessibility, especially in rural areas where most
agricultural production is situated. Subsidizing or funding DSAs activ-
Funding
ities is also implied to eliminate service value proposition-related bottle-
necks. So is policy and regulatory effort to catalyze the digitalization
The authors disclosed receipt of the following financial support for
of agriculture through DSA uptake and to protect the industry from ex-
the research and authorship of this article: This work was supported by
ploitative actors. Service discoverability impediments call for policy in-
the Programme of Accompanying Research for Agricultural Innovation
terventions in awareness and collaboration enhancing initiatives such
(PARI) funded by the German Federal Ministry for Economic Coopera-
as an AP4DSA.
tion and Development (BMZ).
The authors recommend that digital entrepreneurs and intrapreneurs
in SSA facing agriculture sector-related business model innovation and
DSA scale-out challenges exploit the cyclic relationship between afford- Declaration of Competing Interest
ability and usefulness under service value proposition. The authors recom-
mend leveraging marketing budgets and distribution networks among The authors declare that they have no known competing financial
managers of large organizations with a nationwide physical presence for interests or personal relationships that could have appeared to influence
enhanced discoverability of DSAs. Further, the authors recommend col- the work reported in this paper.
laboration among digital agriculture ecosystem actors towards a thriv-
ing AP4DSA as described in section 5.2.
Acknowledgmentss
5.7. Limitations and directions for future research
The authors acknowledge the kind and generous participant mobi-
Using three options on the survey’s Likert-like scale rather than five lization support from Beatrice Chepkorir and Kibet Koech as the ad-
or seven possibly limited the findings in this study. However, the trade- ministrators of Digital Farmers Kenya, and Dairy Farmers Kenya Face-
off for this limitation paid off in more timely and cost-effective acqui- book groups respectively. The researchers also acknowledge the gener-
sition of reasonably large samples of the qualitative and quantitative ous writing and proofreading support from Janet Mutiso, Dr. Oliver Wa-
evidence than if the researchers designed the study differently. The pro- songa, and Peter Ndungo and the diligent administrative support from
posed measurement scale for impediments to the scale-out of DSAs may Lucille Odongo.

9
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

Appendix A. The instrument∗ Appendix C. Sample Profile

∗ Thiswas part of a longer questionnaire in the larger research. Table C1.

(Q9) Do you agree or disagree that the following can be preventing
digital services for agriculture from being successful enough to attain Table C1
Demographic distribution of the respondents in the study.
the scale of M-PESA, Uber, Ali baba etc? (Technology Concerns)
Category Measure Frequency Percentage
I Neither Agree
I Disagree Nor Disagree I Agree Gender (n=575) Female 146 25%
Male 429 75%
T1 - Requiring high literacy levels among users □ □ □ Age range (n=574) 18-24 114 20%
T2 - Requiring access to the internet □ □ □ 25-34 209 36%
T3 - Requiring use of smart phones rather than □ □ □ 35-44 137 24%
more affordable phones 45-54 83 14%
T4 - Requiring phones to be always charged □ □ □ 55-64 26 5%
T5 - Incurring expensive data and SMS charges □ □ □ 65+ 5 1%
Highest education Did not complete any formal 4 1%
Please type the additional views you have on technological concerns level attained education
with respect to success of digital services in the space below. (n=578) Primary School 11 2%
Secondary School 61 11%
Short Course Certificate After 78 13%
Secondary School
Diploma Certificate 151 26%
(Q10) Do you agree or disagree that the following can be preventing
University Degree 221 38%
digital services for agriculture from being successful enough to attain Master’s degree 48 8%
the scale of M-PESA, Uber, Ali baba etc (Market concerns) Doctor of Philosophy (PhD) 4 1%
Income reliance on Fully reliant 280 49%
I Neither Agree agricultural Partial reliant 193 34%
I Disagree Nor Disagree I Agree activities (n=567) Not reliant 94 17%
Involvement in Farm inputs provision 67 12%
M1 - Awareness about them is low □ □ □
Agricultural value Production - crops and animals 492 85%
M2 - There is no one-stop-shop where all □ □ □
chain activities Post harvest logistics 46 8%
of them can all be found
(n=582) Marketing and brokerage services 69 12%
M3- They do not work together or share □ □ □
Specialized services 84 14%
data with each other
Finance and insurance 21 4%
M4 - Their charges are too high to keep □ □ □
paying
M5 - The value they offer their target □ □ □
customers is not convincing

Please type the additional views you may have on this in the space
below.

Appendix B. Respondents Summary

Table B1.

Table B1
Summary of Respondents and their Affiliations.

Social Media Group / Category Platform Membership Size in Feb’ 2019 Total Participants Valid for larger researcha Valid for this studyb

Africa Farmers Club Facebook 126,364 14 13 8

Dairy Farmers Forum Facebook 64,129 5 5 2
Dairy Farmers Kenya Facebook 299,000 233 204 147
Digital Farmers Kenya Facebook 307,498 530 505 384
Farming Gurus Facebook 120,000 4 4 2
Kenya Farmers Market Facebook 51,000 2 2 1
Kienyeji Poultry Farming Facebook 220,000 8 7 7
Poultry farming group Ghana Facebook 23,000 43 0 0
Direct invites by DSA Providers N/A N/A 2 2 2
Whatsapp and telegram groups Mixed N/A 46 43 29
Total N/A 887 785 582
a
In the scope of the larger research - beyond examining impediments to DSA scale-out.
b
Participants who either had no missing values on numerically measured statements or gave additional views on impediments to DSA scale
out.

10
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

References [28] J.F. Hair, W.C. Black, B.J. Babin, R.E. Anderson, Multivariate Data Analysis,
Pearson Education, Harlow, 2014 (new int. ed.) [Link]
[1] AGRAAfrica Agriculture Status Report: The Business of Smallholder Agricul- p/multivariate-data-analysis-pearson-new-international-edition/P100000057627 .
ture in Sub-Saharan Africa (Issue 5), Alliance for a Green Revolution in [29] J.C. Anderson, D.W. Gerbing, Structural equation modeling in practice: A review and
Africa (AGRA), Nairobi, Kenya, 2017 Issue No. 5 [Link] recommended two-step approach, Psychological Bulletin 103 (3) (1988) 411–423,
uploads/2017/09/[Link] . doi:10.1037/0033-2909.103.3.411.
[2] Christiaensen, L., & Martin, W. (2018). Agriculture, structural transforma- [30] J.L. Horn, A rationale and test for the number of factors in factor analysis, Psychome-
tion and poverty reduction: Eight new insights. [Link] trika 30 (2) (1965) 179–185, doi:10.1007/bf02289447.
[Link].2018.05.027. [31] J.C. Hayton, D.G. Allen, V. Scarpello, Factor retention decisions in exploratory fac-
[3] M. Ivanic, W. Martin, Sectoral productivity growth and poverty reduction: National tor analysis: A tutorial on parallel analysis, Organizational research methods 7 (2)
and global impacts, World Development 109 (2018) 429–439. (2004) 191–205, doi:10.1177/1094428104263675.
[4] E. Ligon, E. Sadoulet, Estimating the relative benefits of agricultural growth [32] B.G. Tabachnick, L.S. Fidell, J.B. Ullman, in: Using multivariate statistics, 5,
on the distribution of expenditures, World Development 109 (2018) 417–428, Pearson, Boston, MA, 2007, pp. 481–498. [Link]
doi:10.1016/[Link].2016.12.007. assets/preface/0/1/3/4/[Link].
[5] A. Shimeles, A. Verdier-Chouchane, A. Boly, Building a resilient and [33] C.H. Li, Confirmatory factor analysis with ordinal data: Comparing robust maximum
sustainable agriculture in sub-Saharan Africa, Springer Nature (2018), likelihood and diagonally weighted least squares, Behavior research methods 48 (3)
doi:10.1007/978-3-319-76222-7. (2016) 936–949, doi:10.3758/s13428-015-0619-7.
[6] J.C. Aker, I. Ghosh, J. Burrell, The promise (and pitfalls) of ICT for agriculture ini- [34] D. Hooper, J. Coughlan, M.R. Mullen, Structural equation modelling: guide-
tiatives, Agricultural Economics 47 (S1) (2016) 35–48, doi:10.1111/agec.12301. lines for determining model fit, Electron J Bus Res Methods 6 (2008) 53–60,
[7] S.O. Ogutu, J.J. Okello, D.J. Otieno, Impact of information and communica- doi:10.21427/D7CF7R.
tion technology-based market information services on smallholder farm input use [35] L.T. Hu, P.M. Bentler, Cutoff Criteria for Fit Indexes in Covariance Structure Analysis:
and productivity: The case of Kenya, World Development 64 (2014) 311–321, Conventional Criteria Versus New Alternatives, Structural Equation Modeling 6 (1)
doi:10.1016/[Link].2014.06.011. (1999) 1–55, doi:10.1080/10705519909540118.
[8] U. Deichmann, A. Goyal, D. Mishra, Will digital technologies transform agriculture [36] R Core TeamR: A language and environment for statistical computing, R Foundation
in developing countries?, The World Bank, 2016, doi:10.1596/1813-9450-7669. for Statistical Computing, Vienna, Austria, 2020 [Link]
[9] L. Klerkx, E. Jakku, P. Labarthe, A review of social science on digital agri- [37] [Link]. Version 9 (2020-2021). [Link] Scientific Software Development GmbH.
culture, smart farming and agriculture 4.0: New contributions and a future re- Qualitative Data Analysis.
search agenda, NJAS-Wageningen Journal of Life Sciences 90 (2019) 100315, [38] Friese, S., Soratto, J., & Pires, D. (2018). Carrying out a computer-aided the-
doi:10.1016/[Link].2019.100315. matic content analysis with [Link]. [Link]
[10] M. Trendov, S. Varas, M. Zeng, Digital technologies in agriculture and rural areas: 324720405_Carrying_out_a_computer-aided_thematic_content_analysis_with_ATLASti.
status report, Digital technologies in agriculture and rural areas: status report, 2019 [39] V. Braun, V. Clarke, Using thematic analysis in psychology, Qualitative research in
[Link] psychology 3 (2) (2006) 77–101, doi:10.1191/1478088706qp063oa.
[11] D. Kos, S. Kloppenburg, Digital technologies, hyper-transparency and smallholder [40] D. McNeish, Thanks coefficient alpha, we’ll take it from here, Psychological Methods
farmer inclusion in global value chains, Current Opinion in Environmental Sustain- 23 (3) (2018) 412, doi:10.1037/met0000144.
ability 41 (2019) 56–63, doi:10.1016/[Link].2019.10.011. [41] C. Fornell, D.F. Larcker, Evaluating structural equation models with unobservable
[12] J. Kieti, T.M. Waema, E.B. Ndemo, T.K. Omwansa, H. Baumüller, Sources variables and measurement error, Journal of marketing research 18 (1) (1981) 39–
of value creation in aggregator platforms for digital services in agriculture - 50, doi:10.2307/3151312.
insights from likely users in Kenya, Digital Business 1 (2) (2021) 100007, [42] M.R. Ab Hamid, W. Sami, M.M Sidek, Discriminant validity assessment:
doi:10.1016/[Link].2021.100007. Use of Fornell & Larcker criterion versus HTMT criterion, Journal of
[13] X. Han, V. Martinez, A. Neely, Service in the Platform Context: A Review of the State Physics: Conference Series In, 890, IOP Publishing, 2017 [Link]
of the Art and Future Research, in: Collaborative Value Co-creation in the Platform doi:10.1088/1742-6596/890/1/012163.
Economy, Springer, Singapore, 2018, pp. 1–27, doi:10.1007/978-981-10-8956-5_1. [43] J. Henseler, C.M. Ringle, M. Sarstedt, A new criterion for assessing discriminant
[14] A. Gawer, Bridging differing perspectives on technological platforms: To- validity in variance-based structural equation modeling, Journal of the Academy of
ward an integrative framework, Research policy 43 (7) (2014) 1239–1249, Marketing Science 43 (1) (2015) 115–135, doi:10.1007/s11747-014-0403-8.
doi:10.1016/[Link].2014.03.006. [44] C. Fuchs, E. Horak, Africa and the digital divide, Telematics and informatics 25 (2)
[15] P. Langley, A. Leyshon, Platform capitalism: the intermediation and capitalisa- (2008) 99–116, doi:10.1016/[Link].2006.06.004.
tion of digital economic circulation, Finance and society 3 (1) (2017) 11–31, [45] S.M. Mutula, Digital divide and economic development: case study of sub-Saharan
doi:10.2218/finsoc.v3i1.1936. Africa, The Electronic Library (2008), doi:10.1108/02640470810893738.
[16] R. Kabbiri, M. Dora, V. Kumar, G. Elepu, X. Gellynck, Mobile phone adoption in agri- [46] J. Ernest III, E.J. Wilson, E.J. Wilson III, The information revolution
food sector: Are farmers in Sub-Saharan Africa connected? Technological Forecast- and developing countries, MIT press, 2004 [Link]
ing and Social Change 131 (2018) 253–261, doi:10.1016/[Link].2017.12.010. books/information-revolution-and-developing-countries.
[17] B. Ezeomah, R. Duncombe, The role of digital platforms in disrupting agricultural [47] A. Scheerder, A. van Deursen, J. van Dijk, Determinants of Internet skills, uses and
value chains in developing countries, in: International Conference on Social Impli- outcomes. A systematic review of the second-and third-level digital divide, Telem-
cations of Computers in Developing Countries, Springer, Cham, 2019, pp. 231–247, atics and informatics 34 (8) (2017) 1607–1624, doi:10.1016/[Link].2017.07.007.
doi:10.1007/978-3-030-18400-1_19. [48] J. Van Dijk, K. Hacker, The digital divide as a complex and dynamic phenomenon,
[18] S. Mittal, S. Gandhi, G. Tripathi, Socio-economic impact of mobile phones on Indian The information society 19 (4) (2003) 315–326, doi:10.1080/01972240309487.
agriculture (No. 246), Working paper, 2010 [Link] [49] Hargittai, E. (2001). Second-level digital divide: Mapping differences in people’s
[19] A. Gichamba, P. Waiganjo, D. Orwa, R. Wario, B. Ngari, Prototyping mAgriculture online skills. arXiv preprint cs/0109068. [Link]
applications among smallholder farmers, in: 2016 IST-Africa Week Conference, IEEE, [50] T.G. Karippacheril, F. Nikayin, M.D. Reuver, H. Bouwman, Serving the poor:
2016, pp. 1–7, doi:10.1109/istafrica.2016.7530687. Multisided mobile service platforms, openness, competition, collaboration and
[20] Lohento, K. & Sotannde, M. (2019). Digital Start-ups Servicing the Agro-food Indus- the struggle for leadership, Telecommunications Policy 37 (1) (2013) 24–34
try in West Africa: Analysis of Business Models and Key Success Drivers. Retrieved [Link]
from [Link] [51] A. Fraser, Land grab/data grab: Precision agriculture and its new
[21] A. Crandall, J. Kieti, Startup business models and challenges for east african ma- horizons, The Journal of Peasant Studies 46 (5) (2019) 893–912,
griculture innovations, in: 2013 IST-Africa Conference & Exhibition, IEEE, 2013, doi:10.1080/03066150.2017.1415887.
pp. 1–10. [Link] [52] A. Osterwalder, Y. Pigneur, Business model generation: a handbook for
[22] R. Fabregas, M. Kremer, F. Schilbach, Realizing the potential of digital develop- visionaries, game changers, and challengers, John Wiley & Sons, 2010,
ment: The case of agricultural advice, Science 366 (6471) (2019), doi:10.1126/sci- doi:10.1111/j.1540-5885.2012.00977_2.x.
ence.aay3038. [53] R. Birner, T. Daum, C. Pray, Who drives the digital revolution in agriculture? A
[23] B.D. Haig, Exploratory factor analysis, theory generation, and scien- review of supply-side trends, players and challenges, Applied Economic Perspectives
tific method, Multivariate Behavioral Research 40 (3) (2005) 303–329, and Policy (2021), doi:10.1002/aepp.13145.
doi:10.1207/s15327906mbr4003_2. [54] H. Baumüller, The little we know: an exploratory literature review on the utility
[24] G. Schurz, Patterns of abduction, Synthese 164 (2) (2008) 201–234, of mobile phone-enabled services for smallholder farmers, Journal of International
doi:10.1007/s11229-007-9223-4. Development 30 (1) (2018) 134–154, doi:10.1002/jid.3314.
[25] N.M. Heydarian, Developing theory with the grounded-theory approach and the- [55] J.J. Okello, O.K. Kirui, Z.M. Gitonga, G.W. Njiraini, J.M. Nzuma, Determi-
matic analysis, APS Observer 29 (4) (2016) - [Link] nants of awareness and use ICT-based market information services in devel-
observer/developing-theory-with-the-grounded-theory-approach-and-thematic-analysis oping country agriculture: The case of smallholder farmers in Kenya, Quar-
. terly Journal of International Agriculture 53 (2014) 263–282 (892-2016-65235),
[26] M.E. Kiger, L. Varpio, Thematic analysis of qualitative data: AMEE Guide No. 131, doi:10.22004/[Link].195738.
Medical teacher 42 (8) (2020) 846–854, doi:10.1080/0142159X.2020.1755030. [56] A.J. Van Deursen, E.J. Helsper, The third-level digital divide: Who benefits most
[27] J. Schoonenboom, R.B. Johnson, How to Construct a Mixed Methods Research from being online? Communication and information technologies annual In, Emer-
Design Wie man ein Mixed Methods-Forschungs-Design konstruiert, KZfSS Köl- ald Group Publishing Limited, 2015, doi:10.1108/s2050-206020150000010002.
ner Zeitschrift für Soziologie und Sozialpsychologie 69 (2) (2017) 107–131, [57] K. Zhu, K.L. Kraemer, J. Dedrick, Information technology payoff in e-business envi-
doi:10.1007/s11577-017-0454-1. ronments: An international perspective on value creation of e-business in the finan-
cial services industry, Journal of management information systems 21 (1) (2004)
17–54, doi:10.1080/07421222.2004.11045797.

11
J. Kieti, T.M. Waema, H. Baumüller et al. Smart Agricultural Technology 2 (2022) 100034

[58] G. Myovella, M. Karacuka, J. Haucap, Digitalization and economic growth: A com- [67] J. Rumanyika, M. Tedre, M. Apiola, S.S. Oyelere, N.R. Mramba, Contem-
parative analysis of Sub-Saharan Africa and OECD economies, Telecommunications porary Challenges in Street Trader-Customer Interaction Through Mobile De-
Policy 44 (2) (2020) 101856 [Link] vices in Dodoma, Tanzania, in: International Conference on Social Implica-
[59] J. Kim, P. Shah, J.C. Gaskell, A. Prasann, Scaling up disruptive agricultural tech- tions of Computers in Developing Countries, Springer, Cham, 2019, pp. 76–88,
nologies in Africa, World Bank Publications, 2020. doi:10.1007/978-3-030-18400-1_7.
[60] Qiang, C. Z., Kuek, S. C., Dymond, A., & Esselaar, S. (2012). Mobile applications for [68] M.K. Kansiime, A. Alawy, C. Allen, M. Subharwal, A. Jadhav, M. Parr, Ef-
agriculture and rural development. [Link] fectiveness of mobile agri-advisory service extension model: evidence from Di-
[62] J.L. Gibbs, K.L. Kraemer, A cross-country investigation of the determinants of scope rect2Farm program in India, World Development Perspectives 13 (2019) 25–33,
of e-commerce use: an institutional approach, Electronic markets 14 (2) (2004) 124– doi:10.1016/[Link].2019.02.007.
137, doi:10.1080/10196780410001675077. [69] L. Jain, H. Kumar, R.K. Singla, Assessing mobile technology usage for knowledge
[64] S. Wyche, C. Steinfield, Why don’t farmers use cell phones to access market prices? dissemination among farmers in Punjab, Information Technology for Development
Technology affordances and barriers to market information services adoption in 21 (4) (2015) 668–676, doi:10.1080/02681102.2013.874325.
rural Kenya, Information Technology for Development 22 (2) (2016) 320–333, [70] Batchelor, S., Scott, N., Valverde, A., Manfre, C., & Edwards, D. (2014).
doi:10.1080/02681102.2015.1048184. Is there a role for Mobiles to support Sustainable Agriculture in Africa?.
[65] Karippacheril, T. G. (2015). Serving the Poor: Designing a Mobile Ser- [Link]
vice Platform for Smallholder Farmer Inclusion in Global Value Chains. [71] Michael Blaschke, Kazem Haki, Stephan Aier, Robert Winter, Taxonomy of
[Link] Digital Platforms: A Platform Architecture Perspective, 14th International
[66] A. Saidu, A.M. Clarkson, S.H. Adamu, M. Mohammed, I. Jibo, Application of Conference on Wirtschaftsinformatik (2019) [Link]
ICT in agriculture: opportunities and challenges in developing countries, Interna- [Link]?article=1220&context=wi2019.
tional Journal of Computer Science and Mathematical Theory 3 (1) (2017) 8–18
[Link]
20of%[Link].

12