Roche Placement Drive — Preparation Roadmap
ROCHE SERVICES & SOLUTIONS (INDIA)
Campus Recruitment Drive — Preparation Roadmap
Prepared for: 3rd-Year [Link] (AI & Data Science) Candidate
Role Track: Data / Business Analytics — Non-SDE Process
How to read this document
This is a probability-based preparation model, not a leaked syllabus. No previous-year
questions, official weightage, or confirmed cut-offs exist for this drive. Every prediction below
is built from three hard facts you provided — the round structure, the handbook contents,
and Roche's analytics/business-support role profile — and is labeled with a confidence level
so you can allocate your limited prep time rationally instead of evenly.
Page 1
Roche Placement Drive — Preparation Roadmap
Contents
• Section 1 — Assessment Prediction Model
• Section 2 — Round 1 (iMocha) Preparation Strategy
• Section 3 — Case Study Round Strategy
• Section 4 — Group Discussion Strategy
• Section 5 — Interview Strategy
• Section 6 — Selection Maximization Plan
Page 2
Roche Placement Drive — Preparation Roadmap
Section 1 — Assessment Prediction Model
Roche is not running a software-engineering pipeline here. A one-hour written test, a one-hour case
study, and a one-hour GD on Day 1, followed by two interviews on Day 2, is the signature of a
business/analytics support hiring loop — this format is common across GCC (Global Capability Centre)
analytics and reporting roles in pharma and FMCG shared-services hubs. The reasoning below treats
your handbook as the strongest available signal, since the placement team would not distribute it
unless it mapped closely to what gets tested.
1.1 Why the handbook is the single best predictor
Three things point to the handbook being curriculum, not background reading: (1) it was handed out
explicitly and tied to this drive, not posted generically; (2) every category in it — Excel, KPIs, statistics,
frameworks — maps one-to-one onto the listed role responsibilities (data analysis, trend identification,
statistical methods, business recommendations); (3) the categories form a coherent analyst skill stack,
not a random grab-bag, suggesting someone at Roche or the placement cell built it deliberately from
the actual JD.
1.2 Topic-probability model for Round 1 (iMocha)
iMocha is a skills-assessment platform that typically mixes MCQ knowledge checks, scenario-based
reasoning, and sometimes a short Excel-simulation or SQL-output-prediction module. Based on the
handbook weighting and role description, here is the estimated topic distribution for the 1-hour written
test:
Topic Cluster Est. Weight Confidence Likely Format
Excel formulas & logic 25–30% High MCQ + “what does this
(VLOOKUP/XLOOKUP, INDEX- formula return” / output-
MATCH, SUMIFS/COUNTIFS, prediction
IFERROR, Pivot logic)
Business KPIs (definitions, 15–20% High MCQ + 1-line calculation
formulas, interpretation) problems
Statistics fundamentals 15–20% High MCQ + small numeric
(mean/median/mode, std dev, computation
correlation, percentiles, z-score,
outliers)
Quantitative & logical aptitude 15–20% Medium-High Standard aptitude MCQs —
(generic placement-test staples) ratios, %, data interpretation
tables
Analytics frameworks (Pareto, 5–10% Medium Scenario MCQ — “which
funnel, cohort, RFM, RCA, framework fits this situation”
scenario)
SQL basics (SELECT, WHERE, 5–10% Medium MCQ or output-prediction,
GROUP BY, JOIN, aggregate unlikely to be live coding
functions)
Python basics (syntax, pandas 0–5% Low-Medium If present at all, conceptual
fundamentals, simple logic) MCQ — not coding under time
pressure
Page 3
Roche Placement Drive — Preparation Roadmap
Read this as a planning tool, not a guarantee: Excel + KPIs + Statistics together account for roughly
55–65% of the paper on this model, which is why Section 2 below front-loads those three areas heavily
over SQL/Python.
1.3 Expected question styles across the test
• Definition-recall MCQs — “Which function returns an approximate match?”
• Output-prediction MCQs — given a small data snippet and a formula, choose the result.
• Applied-business MCQs — given a scenario (e.g. “customers stopped renewing”), pick the correct
KPI or framework.
• Data-interpretation sets — a small table/chart followed by 2–4 questions (standard aptitude-test
format).
• Possible short free-text or scenario-justification item — less likely in a strict 1-hour MCQ format,
but worth a one-line mental template just in case.
1.4 What this round is actually screening for
Round 1 is a filter, not a differentiator. Roche almost certainly is not trying to find the single best analyst
here — it's eliminating candidates who lack baseline Excel/stats/business fluency before spending two
full days of interviewer time on Day 2. Your goal in this round is accuracy and speed on the basics, not
depth. Don't over-invest study time chasing edge-case Excel functions or advanced statistics here —
that effort pays off more in Round 2 and the interviews.
Page 4
Roche Placement Drive — Preparation Roadmap
Section 2 — Round 1 (iMocha) Preparation Strategy
2.1 Handbook chapters ranked by priority
Rank Chapter Why it ranks here
1 Excel formulas Largest estimated weight; also the most “trainable” in the time
you have — high return per hour studied.
2 Business KPIs Directly reused in Round 2 (case study) and interviews —
studying this once pays off three times.
3 Statistics fundamentals High test weight and also the basis for the “outlier/cohort/RFM”
frameworks below it.
4 Analytics frameworks Lower direct MCQ weight, but this is your Round 2 toolkit — don't
skip it even though Round 1 weight is modest.
5 Pivot Tables / Power Likely tested conceptually (what each does, when to use which)
Query rather than hands-on.
6 SQL/Python Lowest confidence weight in Round 1; treat as a light pass now,
deeper pass before interviews.
2.2 Must-know Excel formulas (highest yield)
Function What to know cold
VLOOKUP vs XLOOKUP XLOOKUP searches in any direction and defaults to exact match — know
this contrast, it's a classic MCQ trap.
INDEX + MATCH Why it's used over VLOOKUP (works left-lookup, doesn't break when
columns are inserted).
SUMIF / SUMIFS Difference between single vs multiple criteria; argument order (criteria
range before sum range in SUMIFS).
COUNTIF / COUNTIFS Same logic as above, applied to counting not summing.
IF / nested IF / IFS IFS as a cleaner alternative to deeply nested IFs.
IFERROR Wraps a formula to catch #N/A, #DIV/0!, etc. — know typical use with
VLOOKUP.
Pivot Tables Rows vs Columns vs Values vs Filters — and when a pivot is the right
tool vs a formula.
Power Query Conceptual only: it's for cleaning/transforming/combining data before
analysis, not formula syntax.
2.3 Must-know KPIs — formulas and one-line meaning
KPI Formula What it tells you
Gross Margin (Revenue − COGS) / Revenue Profitability before operating costs
Net Margin Net Profit / Revenue Profitability after all costs
Page 5
Roche Placement Drive — Preparation Roadmap
KPI Formula What it tells you
CAC Total acquisition spend / new Cost to win one customer
customers acquired
LTV Avg. revenue per customer × avg. Total value a customer brings over time
customer lifespan
Retention Rate (Customers at end − new % of existing customers kept
customers) / customers at start
Churn Rate 1 − Retention Rate % of customers lost
Conversion Rate Conversions / total visitors or leads Funnel efficiency
ARPU Total revenue / total active users Average revenue per user
AOV Total revenue / number of orders Average order value
NPS % Promoters − % Detractors Customer loyalty/advocacy
LTV:CAC is a classic applied-MCQ pairing — know that a ratio below roughly 1:1 signals an
unsustainable acquisition cost relative to value, while higher ratios (often cited around 3:1+) are
considered healthy.
2.4 Must-know statistics
• Mean vs median vs mode — and specifically when median is preferred (skewed data, presence of
outliers).
• Variance vs standard deviation — std dev is the square root of variance, same units as the data.
• Correlation — value range (–1 to +1), and that correlation ≠ causation (near-guaranteed concept
check).
• Quartiles/percentiles — Q1/Q2/Q3, IQR, and IQR's role in outlier detection.
• Z-score — how far a value sits from the mean in standard-deviation units; used to flag outliers.
• Outlier detection — IQR method (below Q1−1.5×IQR or above Q3+1.5×IQR) and z-score
thresholds (commonly |z|>3).
2.5 Likely MCQ patterns to drill (self-test style)
1. “Which formula returns the position of a value, not the value itself?” (MATCH)
2. “A customer base loses 8% of accounts monthly. What is the churn rate / retention rate?”
3. “Which measure of central tendency is least affected by an extreme outlier?” (Median)
4. “Given a small table, which Excel function would correctly sum sales only for Region = ‘West’ and
Month = ‘Jan’?” (SUMIFS)
5. “Which framework best diagnoses *why* a metric dropped, rather than just measuring it?” (Root
Cause Analysis)
6. “A dataset's Q1 is 20, Q3 is 60. Which value is an outlier by the IQR rule?” (Apply the 1.5×IQR
fence.)
Page 6
Roche Placement Drive — Preparation Roadmap
Section 3 — Case Study Round Strategy
This round is where the handbook's “analytics frameworks” section stops being theory and becomes
your operating toolkit. Assume the case will hand you a business symptom (a KPI moved in the wrong
direction) and reward you for diagnosing it structurally rather than guessing at causes.
3.1 The universal response structure (use this every time)
7. Clarify the metric — restate the problem in one sentence and confirm the timeframe/segment.
8. Frame the hypothesis space — list 3–5 plausible cause categories before picking one (don't jump
to a single cause).
9. Select the right framework — match the symptom to a tool (see 3.3 below).
10. Quantify where possible — even rough estimates (“if X drove half the drop, that's roughly ₹Y”)
beat pure narrative.
11. Recommend with a tradeoff — every recommendation should name what you'd watch to
confirm it's working, and one risk/cost of the action.
3.2 Sample case studies to practice against
Case A — Retention drop
“Monthly retention for a subscription service fell from 92% to 81% over two quarters. How
would you investigate and what would you recommend?” — Practice: cohort analysis to
isolate which sign-up cohort is churning, RFM to segment by value, and a clear hypothesis
list (pricing change, competitor entry, onboarding friction, product issue) before
recommending.
Case B — Funnel conversion dip
“Website-to-purchase conversion dropped 30% after a redesign launch. Diagnose and
recommend next steps.” — Practice: funnel analysis to find the exact stage that leaks, then
root-cause that stage specifically rather than the whole funnel.
Case C — Cost spike
“Customer acquisition cost doubled in one quarter while new customer volume stayed flat.
What's happening?” — Practice: separate spend-side causes (channel mix, bidding costs)
from demand-side causes (market saturation, weaker creative/targeting), and use CAC:LTV
to frame whether this is still a sustainable spend.
Case D — Concentration risk (Pareto)
“80% of revenue comes from 20% of clients. Is this a problem, and what would you do about
it?” — Practice: this is a direct Pareto-analysis prompt — discuss both the upside (efficient
focus) and the risk (concentration/dependency), then propose a balanced action (diversify
vs. deepen).
3.3 Framework selection cheat-sheet
If the symptom is… Reach for…
A metric dropped and you need to know why Root Cause Analysis (5 Whys / fishbone-style
Page 7
Roche Placement Drive — Preparation Roadmap
If the symptom is… Reach for…
hypothesis tree)
Customers fall off at a specific step Funnel Analysis
Behaviour differs by signup/launch group over Cohort Analysis
time
You need to prioritize which customers/products RFM Analysis (Recency, Frequency, Monetary)
matter most
A small share of inputs explains most of the Pareto Analysis (80/20)
outcome
You're deciding between future options under Scenario Analysis (best/base/worst case)
uncertainty
3.4 Common mistakes that sink case answers
• Jumping straight to one cause without listing alternatives first — looks like guessing, not analysis.
• Pure narrative with zero numbers — even rough math signals analytical rigor.
• Naming a framework without applying it (“I'd do a cohort analysis” and then never using it).
• Giving a recommendation with no success metric attached — always state what you'd measure
afterward.
Page 8
Roche Placement Drive — Preparation Roadmap
Section 4 — Group Discussion Strategy
A 1-hour GD slot for an analytics-track drive most likely runs multiple small groups rather than one
continuous hour per person. Topics for analytics/business-support hiring tend to cluster around three
buckets: data-and-technology-in-business themes, general current-affairs/ethics themes, and abstract
reasoning prompts. Prepare across all three rather than betting on one.
4.1 Likely topic pool
Category Example topics Why likely
Data & technology in AI/automation's impact on jobs in analytics Directly on-theme for an
business Data privacy vs. personalization analytics hire — tests
Should companies prioritize data-driven whether you can discuss
decisions over intuition? the field you're entering.
Healthcare/pharma- Access to healthcare vs. cost of innovation Roche is a healthcare
adjacent (Roche Role of technology in modern healthcare company; a healthcare-
context) delivery adjacent prompt is a
reasonable bet even if not
certain.
General Remote work vs. office work productivity Standard GD staples used
business/current affairs Is social media good or bad for society? across most campus drives
regardless of role.
Abstract/reasoning “Is failure necessary for success?” style Used when evaluators want
prompts prompts to see structured thinking
independent of subject
knowledge.
4.2 How to structure what you say
• Open with a one-sentence framing of the topic, not an opinion — shows you're structuring, not
reacting.
• Use a simple structure aloud: “There are two sides here — on one hand… on the other… and I'd
lean toward… because…”
• Bring in one data point or real-world reference if you have one — it stands out against pure
opinion.
• Build on others by name (“Building on what [name] said…”) rather than only pushing your own
point — this is what “team discussion” evaluators actually score.
• Close strong if you get the chance to summarize — a one-line synthesis at the end is
remembered more than mid-discussion points.
4.3 Common mistakes
• Talking the most instead of contributing the clearest point — evaluators score structure and
listening, not airtime.
• Interrupting repeatedly to force entry — wait for a natural pause or explicitly request one (“Can I
add a point here?”).
• Staying silent for long stretches — aim for 3–4 distinct, well-placed contributions over the session.
Page 9
Roche Placement Drive — Preparation Roadmap
• Taking an extreme, one-sided stance — analytics roles value balanced, evidence-aware
reasoning over strong opinions.
Page 10
Roche Placement Drive — Preparation Roadmap
Section 5 — Interview Strategy
Two rounds on Day 2 most likely split as one more technical/analytical conversation and one more
HR/fit/communication conversation — though in analytics-hiring loops both interviewers often probe
both areas. Prepare every category below for either room.
5.1 Likely technical/analytics questions
• Walk me through how you'd analyze [a dataset/metric] from scratch.
• Difference between VLOOKUP and INDEX-MATCH — and why you'd choose one.
• How would you detect and handle outliers in a dataset?
• Explain a KPI you've worked with and what it told you.
• What's the difference between correlation and causation? Give an example.
• How would you explain a technical finding to a non-technical stakeholder?
5.2 Likely SQL questions (concept-level, not deep coding)
• Difference between WHERE and HAVING.
• What does GROUP BY do, and when do you pair it with an aggregate function?
• Difference between INNER JOIN and LEFT JOIN, with a simple example.
• Write a simple query to find the top N rows by a metric (e.g., top 5 customers by revenue).
5.3 Likely Python questions (concept-level)
• What is pandas used for, and what's the difference between a Series and a DataFrame?
• How would you read a CSV and check for missing values in pandas?
• Difference between a list and a dictionary, and when you'd use each.
5.4 Project questions — be ready to defend, not just describe
• Walk me through one project end-to-end: problem, data, method, result.
• What was the hardest analytical decision in that project, and why did you make it that way?
• If you had one more week on this project, what would you improve?
• What would you have done differently if the data had been messier/smaller/biased?
5.5 Standard fit/HR questions to have ready
• Why analytics, and why Roche specifically?
• Tell me about a time you had to explain something technical to someone non-technical.
• Tell me about a time your analysis or recommendation was wrong — what did you learn?
• How do you prioritize when you have multiple deadlines?
Page 11
Roche Placement Drive — Preparation Roadmap
Section 6 — Selection Maximization Plan
Clearing each round and getting selected are different optimization problems. Most candidates plateau
at “competent” on Excel/KPIs/stats because that's what the handbook visibly asks for. The
differentiation budget below is about where to spend extra effort that the average candidate won't, since
that's what actually separates a pass from a final offer.
6.1 Highest-ROI actions (do these first)
Action Time cost Why it's high ROI
Build one fully worked mini case 2–3 hrs Gives you a concrete story for Round 2 patterns
(pick any real or sample dataset, and a strong project answer in interviews —
run it through KPI → framework → reused 3 times for 1 unit of effort.
recommendation end-to-end)
Drill Excel formula output-prediction 2–3 hrs Directly targets the single largest Round 1
questions until near-instant recall weight bucket.
Memorize KPI formulas + 1-line 1 hr Reused across Round 1, Round 2, and
“what it tells you” for all 11 listed interviews — the single most cross-leveraged
KPIs chapter in the handbook.
Practice saying the Section 3.1 1 hr Fluency under time pressure is what separates a
case-response structure out loud good case answer from a great one; content
3–4 times knowledge alone isn't enough.
Prepare 2 GD opening lines and 2 20 min Removes the hardest part of a GD — finding
“building on that…” transition your entry point — before you're in the room.
phrases in advance
6.2 What most candidates will under-prepare (your edge)
• Connecting KPIs to frameworks: most candidates memorize KPI formulas in isolation. Practice
pairing each KPI with the framework that diagnoses it (e.g., Churn Rate ↔ Cohort Analysis; CAC
spike ↔ Root Cause Analysis) — this pairing is exactly what a case study or interview will reward.
• Quantified storytelling: most candidates describe projects narratively. Add one number to every
project answer (“this reduced X by roughly Y%”) — it signals analytical instinct even if the number
is an estimate, clearly labeled as such.
• Stakeholder framing: practice explaining one technical concept (e.g., outlier detection) in two
ways — once to a technical interviewer, once as if to a non-technical business stakeholder. This
directly matches the JD's “stakeholder communication” requirement and is rarely rehearsed by
candidates.
• Roche/healthcare context: a one-paragraph point of view on why data analytics matters in a
healthcare/pharma services context will make your “why this company” answer stand out from
generic “great learning opportunity” answers.
6.3 48-hour pre-drive schedule (suggested)
Block Focus
Day −2, Morning Excel formulas — drill output-prediction questions until fast and automatic.
Page 12
Roche Placement Drive — Preparation Roadmap
Block Focus
Day −2, Afternoon KPIs — memorize formulas + meaning; pair each with a likely framework
(6.2).
Day −2, Evening Statistics fundamentals — outlier detection, correlation vs causation, z-
score.
Day −1, Morning Build the one worked mini-case end-to-end (Section 6.1).
Day −1, Afternoon Practice the Section 3.1 structure out loud against Cases A–D.
Day −1, Evening Light SQL/Python concept review + prepare GD openers and 2–3 project
stories.
Day of Light review only — KPI sheet + framework cheat-sheet skim, then rest.
6.4 Honest calibration
This plan is built on strong but indirect evidence (the handbook + role description), not confirmed
Roche-specific past papers. Treat the probability weights in Section 1 as a prioritization aid, not a
certainty. The highest-confidence, lowest-regret move is still mastering Excel + KPIs + Statistics deeply,
since those three show up — in some form — in literally every round of this process, written test
through final interview.
Page 13