Interview guides · TD

TD Data Scientist Interview Questions & Guide 2026

Data scientist interview prep for TD: original practice on experiment design, SQL and metric reasoning, with worked solutions.

The 60-second brief

At TD, the Data Scientist role sits at the intersection of complex financial modeling, massive-scale data engineering, and strategic business decision-making. You will not just be building models in a vacuum; you will be acting as a bridge between raw, high-velocity financial data and actionable insights that drive the banking experience for millions of customers. Whether you are working on fraud detection, personalized banking services, or risk assessment, your work directly influences the stability and innovation of TD’s digital footprint.

If the team owns experimentation, expect depth past a two-sample test: minimum detectable effect and its roughly inverse-square-root dependence on sample size (holding power, significance level and variance fixed), variance reduction from pre-period covariates, interference between units, and when a sequential design is the right call.

Focus areas

Pick a randomisation unit that respects interference
Decompose a metric move by segment and mix
Size an experiment before anyone launches it

Overview

The role is increasingly focused on the "Full Stack" lifecycle—meaning you are expected to own the journey from data ingestion and feature engineering to model deployment and monitoring. You will collaborate with cross-functional teams, including software engineers, product managers, and business stakeholders, to ensure that machine learning solutions are not only accurate but also scalable and production-ready. This is an environment where precision, ethics, and performance are paramount.

TD values candidates who can articulate the "why" behind their technical choices. Be prepared to explain how your models solve specific business problems rather than just demonstrating technical complexity.

The TD interview process

TD candidates report 5 rounds · ≈ 4-6 weeks. The stages below are what candidates describe, not a published process.

Recruiter Screen

A screening call is a matching exercise run by someone who will not evaluate your statistics. They are checking that the work described on your resume is work you personally did, and that its scope matches the level the role is written for. Logistics get settled in the same half hour so nobody spends an interviewer's afternoon on a mismatch. The answer that fails is the one narrated in the plural. If every sentence is 'we built' and 'the team decided', there is nothing specific to write down about you. Name the piece that was yours, the decision you made inside it, and what changed after.

What to demonstrate

Whether the ownership implied by your resume survives one round of follow-up about who actually did which part
Whether your described scope (data size, stakeholders, what shipped) matches the seniority the role is written at
Whether timeline, location and compensation expectations make the rest of the loop worth scheduling

How to prepare

Rewrite your top three resume bullets in the first person singular, each with the decision you made and what moved afterwards, then say them out loud once so the 'we' does not return under pressure
Attach one number to each project: the baseline, the change, and the window it was measured over. Where impact was never measured, say that plainly rather than inventing a figure
Settle your compensation range before the call and give it as a range with a reason behind it, such as current total comp or a competing timeline, instead of deflecting the question twice

Technical Deep-Dives

A handful of shapes account for most of what gets asked in this format: a ranking or deduplication inside groups, a running or rolling total, a period-over-period comparison, and a cohort tracked forward over time. Recognising the shape quickly is most of the speed here; deriving it from scratch while a clock runs is where the time goes. Know that a window function keeps every row while a GROUP BY collapses them, and know which one the question needs. If the exercise is in Python instead of SQL, the same shapes arrive as groupby with transform, shift and merge, and the same grain mistakes are available.

What to demonstrate

Whether you reach the right construct without a detour, such as ROW_NUMBER over a partition to deduplicate instead of a self-join against a MAX subquery
Whether you know what your window frame actually is, since adding ORDER BY inside OVER changes the default frame and silently changes a running total
Whether the thing runs. A near-miss that throws an error scores below a plainer query that returns the right rows.

How to prepare

Write each of the four shapes once from memory against a small schema and keep the working version somewhere you will reread it: dedupe with ROW_NUMBER, a running total, a month-over-month change with LAG, and a retention table
Compute one running total twice on data with tied timestamps, once on the default frame and once with ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW, and look at where the two disagree
If Python is on the table, rebuild the dedupe and the running total with groupby and cumsum, then assert the two implementations return identical rows

Live Coding Assessment

Before anything else, this round is a reading test. You are given a small schema and a question phrased in business language, and most of the difficulty sits in the gap between them. Who counts as an active user, does a refunded order still count as an order, is that date column an event time or a load time. Weak answers start typing immediately and compute something precise about the wrong population. Strong ones pin the definition in one sentence, name the column that encodes it, then write the query. On a timed assessment with nobody to tell, write the definition in a comment anyway.

What to demonstrate

Whether an ambiguous term becomes a specific column and filter before any computation happens
Whether you read the schema for keys and cardinality rather than only for column names
Whether the result answers the question at the grain it was asked at, per user or per session or per day

How to prepare

Take three metrics you already use and write down the exact filter and exact grain behind each, then practise stating one of them in a single sentence out loud
On a schema you have never seen, spend the first minute writing what one row of each table means and which key it is unique on, then predict which joins can duplicate rows
Rehearse a version where the definition changes halfway through, and edit the query you have instead of starting over

System Design Whiteboard

Rounds outside the standard loop often open with something deliberately under-specified: a loose business problem, an open question about a product area, a dataset described in one sentence. The common failure is surveying, listing six plausible approaches and committing to none of them. The thing that separates a strong answer is scoping out loud. State what you are treating as the goal, name the metric you would move, say what you are choosing not to do and why, then take one path through to an actual answer. An interviewer can follow you down a narrow path. Nobody can grade a menu.

What to demonstrate

Whether you turn an ambiguous prompt into a stated question with a measurable outcome before doing any work
The judgement visible in what you cut, and whether you say why you cut it rather than silently dropping it
Whether you land on a concrete recommendation with its caveat attached, rather than an unranked set of options

How to prepare

Take three vague prompts, such as 'is this feature working', 'why did retention drop', and 'should we expand into a new segment'. For each, write one sentence of goal, one primary metric with its window, and two things you are explicitly not doing.
Practise giving the recommendation first and the reasoning second, in five minutes. Loosely defined rounds are usually time-boxed, and an answer that arrives last often does not arrive.
Keep a running assumption list as you talk, on paper or in the shared doc, so the interviewer can challenge one assumption instead of your whole answer.

Behavioral Interviews

Most of the weight in this round sits on the disagreement questions. Data work routinely produces an answer someone senior did not want, and the interviewer is trying to learn what you do in that hour. Both failure modes are common: folding as soon as a director pushes back, and treating the pushback as ignorance to be corrected with a better chart. A strong answer usually contains a specific thing the other person knew that you did not, and describes how you found out whether it changed the conclusion.

What to demonstrate

Whether you can state the other side's argument accurately before you explain why you disagreed
What you treated as evidence during the disagreement, such as a rerun under their assumption or a holdout check, rather than persuasion technique
Whether you distinguish being overruled from being wrong, and can give an example of each

How to prepare

Write out one disagreement where you turned out to be wrong, and say what in the data misled you. Candidates prepare the story where they were right, and the follow-up asks for the other one.
For your main disagreement story, be ready to say what result would have made you drop your position. If no such result exists, you were not arguing from the data.
Practise stating the opposing position out loud in one sentence the stakeholder would accept, then continue the story.

Where candidates lose points

Comparing cohort retention curves of different maturities, or building the curve from users who are still present

A cohort four weeks old has no week-8 value, so an average taken across cohorts silently drops young cohorts from the later columns and keeps them in the earlier ones. The curve then bends upward at the tail, and the reading that 'retention is improving over time' is an artefact of which cohorts survived to be measured. The same error appears in the denominator when retention is computed over users active in the current period rather than over the full original cohort, which conditions on survival and guarantees a flattering number. The fix is a triangle: fix the cohort at signup, bound every window on both sides, and only compare cells where every cohort has had the full elapsed time, publishing the rest as blank rather than as a partial average.

Watching an experiment daily and stopping when it crosses significance

A fixed-sample test controls type I error at one pre-declared look. Checking repeatedly and stopping at the first p < 0.05 inflates the false positive rate to roughly 0.15 to 0.20 for ten looks, and it rises further with more frequent checks, because the p-value takes a random walk that will eventually dip below the threshold under the null. The usual defences are a fixed horizon declared before launch, group-sequential boundaries such as O'Brien-Fleming that spend alpha across a planned number of looks, or always-valid confidence sequences that are correct under continuous monitoring. Compounding it, the effect size reported conditional on having crossed the threshold is biased away from zero, and the bias is larger the lower the power was, so an underpowered test that 'won' typically overstates the lift it found.

Generalising beyond the population the sample actually supports

State the frame the sample was drawn from and where it diverges from the population you want to talk about: time window, platform, geography, opt-in. If a group is excluded from the frame, either weight to known margins or narrow the claim rather than quietly extending it.

Reporting a p-value with no effect size or interval

Give the estimated difference with a confidence interval in the units the business cares about, then say whether that whole interval is worth acting on. A p-value only addresses whether you can rule out exactly zero; it says nothing about magnitude.

Questions & practice

How do you manage model drift once a model is in production?

Medium · machine learning and modelling

Approach

Set a baseline first, so any model has something honest to beat.
Say how the offline result would be validated online before it is trusted.
Check what information would not exist at prediction time, and exclude it.

Follow-up

What would you monitor after launch to know the model is still valid?
Where could label leakage enter this setup?

How do you ensure model interpretability when working with complex black-box algorithms?

Medium · machine learning and modelling

Approach

Check what information would not exist at prediction time, and exclude it.
Set a baseline first, so any model has something honest to beat.
Say how the offline result would be validated online before it is trusted.

Follow-up

How would you choose the decision threshold, and who owns that choice?
What would you monitor after launch to know the model is still valid?

Describe the architecture you would use to deploy a machine learning model to a cloud environment.

Medium · machine learning and modelling

Approach

Set a baseline first, so any model has something honest to beat.
Pick an evaluation metric that matches the cost of each error type, not a default.
Frame the prediction: the label, the moment of prediction, and the action it triggers.

Follow-up

How would you choose the decision threshold, and who owns that choice?
What would you monitor after launch to know the model is still valid?

Cluster bootstrap for a per-session rate randomised on users

Hard · cluster bootstrap · randomisation unit · variance

An experiment randomised on user_id reports a per-session conversion rate, so sessions inside a user are correlated. Input: one row per session with user_id, variant in {control, treatment} and converted in {0,1}. Write a cluster bootstrap from scratch: resample users with replacement within each arm, keep every session of a drawn user, recompute each arm's ratio of converted sessions to sessions, and take the difference. Return the point estimate, a 95 percent percentile interval from at least 2,000 resamples, the naive session-level interval that ignores clustering, and the ratio of their widths.

Approach

Name the estimand precisely: it is a ratio of sums, sum(converted) over sum(sessions) within an arm, not the mean of per-user rates. Those differ whenever session counts vary across users, and the ratio is what the reported metric is.
Resample the cluster, not the row. Draw n_users user ids with replacement inside each arm and take every session belonging to each draw, including duplicate draws of the same user. Keeping the user count fixed per arm rather than the session count is what preserves the sampling design.
Precompute per-user (converted_sum, session_count) once, so each resample is two vector lookups and a division rather than a repeated filter over the session frame. That turns 2,000 resamples from minutes into under a second.
Take the 2.5th and 97.5th percentiles of the 2,000 differences for the interval, and report the point estimate from the full data rather than from the bootstrap mean, since the bootstrap mean carries the resampling bias.
Compute the naive interval from the session-level binomial standard error and compare widths. The expected inflation is roughly sqrt(1 + (m-1)*rho), with m the mean sessions per user and rho the intraclass correlation of converted within users, so a computed ratio far from that value points at a bug in one of the two intervals.

Worked solution (35 min)

per_user = df.groupby(['variant','user_id'])['converted'].agg(['sum','size']); split into two arrays per arm.
point = (t_sum.sum() / t_n.sum()) - (c_sum.sum() / c_n.sum()).
For b in range(B): idx = rng.integers(0, len(t_sum), len(t_sum)); ratio_t = t_sum[idx].sum() / t_n[idx].sum(); same for control; store the difference. Vectorise by drawing a (B, n) index matrix if memory allows.
ci = np.percentile(diffs, [2.5, 97.5]); naive_se = sqrt(p_t*(1-p_t)/n_sessions_t + p_c*(1-p_c)/n_sessions_c); naive_ci = point +/- 1.96*naive_se.
width_ratio = (ci[1]-ci[0]) / (naive_ci[1]-naive_ci[0]).

Expected result: A point estimate identical to the full-data ratio difference, a percentile interval containing that point estimate, and a width ratio greater than 1 whose value approximates sqrt(1 + (m-1)*rho) for the data's mean cluster size m and intraclass correlation rho. A ratio of approximately 1.0 means sessions were resampled instead of users.

Verify it yourself

Construct data where every user has exactly one session: the cluster and naive intervals then agree to within Monte Carlo error, since m = 1 makes the design effect 1.
Construct data where converted is constant within each user: rho is 1, and the width ratio should approach sqrt(m).
The bootstrap distribution's mean sits close to the point estimate; a large offset means the resampled arms have drifted in size.
Doubling B narrows the Monte Carlo noise in the percentile endpoints without shifting the interval's centre.

Follow-up

Users average 3.4 sessions and the intraclass correlation is 0.12. What width ratio do you predict before running it, and does your bootstrap land there?
Give the delta-method standard error for this ratio and say when you would prefer it to the bootstrap.
Half the users in the treatment arm have exactly one session. What does that do to the cluster bootstrap's coverage, and how would you check it?

Rebuild sessions from raw events with a thirty-minute gap

Medium · window-functions · sessionisation · gaps-and-islands

From fct_event (event_id, visitor_id, occurred_at_utc) alone, rebuild sessions: a new session begins when the gap from that visitor's previous event exceeds 30 minutes, and every session is force-closed at UTC midnight so none spans two calendar dates. Return one row per session with visitor_id, session_start, session_end, event_count and session_date. Do not read fct_session; the point is to reproduce it. Assume duplicate occurred_at_utc values exist for the same visitor.

Approach

Get the previous timestamp per visitor with LAG(occurred_at_utc) OVER (PARTITION BY visitor_id ORDER BY occurred_at_utc, event_id). The event_id tiebreaker is required, not stylistic: with duplicate timestamps an unstable ordering makes the boundary flags non-deterministic between runs.
Set a boundary flag when prev IS NULL, or occurred_at_utc - prev > interval '30 minutes', or occurred_at_utc::date <> prev::date. The third disjunct is the midnight rule, expressed as a date change rather than a clock comparison so it holds across any gap length.
Number the islands with SUM(flag::int) OVER (PARTITION BY visitor_id ORDER BY occurred_at_utc, event_id ROWS UNBOUNDED PRECEDING). Because event_id is unique the ordering is total, so no two rows are peers and RANGE UNBOUNDED PRECEDING would compute exactly the same numbers here. Write ROWS anyway: it is the half of the guard that survives someone later simplifying the ORDER BY back to occurred_at_utc alone, at which point the default RANGE frame gives every row sharing a timestamp one shared running total.
GROUP BY visitor_id and the island number, then MIN(occurred_at_utc) AS session_start, MAX(...) AS session_end, COUNT(*) AS event_count, session_start::date AS session_date.
Reconcile against fct_session on one sample day. The counts should agree except for server-emitted events carrying no client session, so a systematic difference beyond those is a bug in the gap rule or in the ordering.

Worked solution (30 min)

Pick one high-volume visitor and dump their ordered event timestamps for a day to trace by hand.
Add LAG with the tiebreaker and eyeball the computed gaps.
Add the three-part boundary flag and confirm the first event of each date is flagged.
Add the running SUM with an explicit ROWS frame, then group and aggregate.
Compare total event_count against the raw input count, and compare session counts to fct_session for the sample day.

Expected result: One row per (visitor_id, island). SUM(event_count) equals the input row count exactly, with no events lost or duplicated. No session_end falls on a different date from its session_start.

Verify it yourself

Total events in equals total events out.
MAX(session_end::date - session_start::date) is 0 across all rows.
For a visitor ordered 09:00, 09:10, 09:55, 23:58, 00:02, the result is four sessions, not three.
No two sessions for the same visitor have overlapping [session_start, session_end] ranges.
Swapping ROWS for RANGE while event_id stays in the ORDER BY changes nothing; dropping event_id from both ORDER BY clauses does change the boundaries for a visitor with duplicate timestamps, which locates the guard in the tiebreaker rather than in the frame keyword.

Follow-up

Why 30 minutes? What does a 5-minute rule do to sessions-per-visitor and to any per-session conversion rate?
The same person uses phone then laptop. Two visitor_ids, two sessions. What breaks if you sessionise on user_id instead?
The midnight rule splits an overnight session. Which metrics does that bias, and in which direction?

Find reactivation gaps in account paid-period history

Hard · gaps-and-islands · window-functions · subscriptions

fct_subscription_period holds account_id, subscription_id, period_start_utc, period_end_utc, period_status and change_reason. A mid-period plan or seat change closes one row and opens another, so a single continuous paid tenure is often many rows, and an account may hold two overlapping subscriptions. Collapse rows with period_status in ('active','past_due') into continuous tenures per account, treating gaps of three days or less as continuous. Return account_id, tenure_start, tenure_end, and for every tenure after the first, the gap in days that preceded it.

Approach

Filter to paid rows only: period_status IN ('active','past_due'). Trialing periods are not tenure, and including them turns every trial that never converted into a one-period tenure followed by a fake churn.
Order by period_start_utc and take a running maximum of all prior ends: MAX(period_end_utc) OVER (PARTITION BY account_id ORDER BY period_start_utc, period_end_utc, subscription_id ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING). Those three columns are the only stable ordering this schema exposes, so check first that they are unique within an account; if rows tie on all three, the island numbering is order-dependent between runs and you need a real row key before the result is reproducible.
LAG on its own is wrong here because with overlapping or nested periods the immediately preceding row by start date is not the one that ends latest, so the running maximum is the part that cannot be shortcut.
Flag a new island when prior_max_end IS NULL OR period_start_utc > prior_max_end + interval '3 days', then number islands with a running SUM of the flag over the same ordering and an explicit ROWS frame.
Group to (account_id, island) taking MIN(period_start_utc) and MAX(period_end_utc), then LAG(tenure_end) OVER (PARTITION BY account_id ORDER BY tenure_start) to compute the preceding gap in days for every tenure after the first.
Sanity-check with change_reason, which is the only lineage this schema carries: list its distinct values first, then confirm that rows recording a plan or seat change sit inside a tenure rather than opening one, and that every tenure after the first opens on a row whose reason records a restart rather than an ordinary renewal. Do not reconcile against a churn timestamp on dim_account, which this schema does not define; and where such a column does exist, a cancellation timestamp records when the request was made and routinely sits weeks before the period it ends.

Follow-up

Why three days of grace? What do 0 and 30 days each do to the count of accounts classed as reactivated?
An account runs two concurrent subscriptions for different teams. One tenure or two, and what does the revenue reader expect?
How would you turn these tenures into a monthly gross logo churn series without double-counting an account that churned and returned in the same month?

Explain the trade-offs between gradient boosting machines and deep learning models for time-series forecasting.

Medium · product sense and metrics

Approach

State what result would change your recommendation, so the answer is falsifiable.
Decompose the metric into the rates that drive it, and say which one you would check first.
Restate the decision this analysis has to support, and who acts on the answer.

Follow-up

How would you detect that the metric is being gamed rather than genuinely improving?
What would you do if the primary metric and the guardrail moved in opposite directions?

How would you design a real-time scoring pipeline for credit card transactions?

Medium

Approach

Say what you would check first and why it is the highest-information step.
State your assumptions explicitly before working the problem.
Clarify what is being asked and what a complete answer would contain.

Follow-up

How would you know your answer was wrong?
What assumption would you test first?

What considerations do you make when choosing between batch processing and streaming for a specific data science use case?

Medium

Approach

State your assumptions explicitly before working the problem.
Work from the decision backwards to the evidence you would need.
Clarify what is being asked and what a complete answer would contain.

Follow-up

How would you know your answer was wrong?
What assumption would you test first?

Define success for halving the free trial length

Hard · monetisation · cohort horizon · trials

A proposal cuts the free trial from 30 days to 14. Trial-to-paid conversion is defined on fct_subscription_period: numerator, subscription_id whose first row with is_first_paid_period = TRUE has period_status IN ('active','past_due') and period_start_utc no later than 14 days after trial end; denominator, subscription_id whose first row has period_status = 'trialing' and period_start_utc in the cohort week. dim_account (account_id, account_created_at_utc, signup_surface) is also available and joins to fct_subscription_period on account_id. Explain why conversion rate alone cannot decide this, define the metric that can, and state the horizon and lag your readout needs. Deliverable: the decision metric and the readout schedule.

Approach

Show the incomparability numerically before arguing about it: at a fixed calendar readout date, a larger share of the 14-day arm's cohorts have completed trial end plus 14 days plus settlement, so the short arm leads mechanically and the lead shrinks as both arms mature.
Fix the censoring first, and do not mistake a rescaling for a fix. The given conversion rate already has trial starts as its denominator, so 'paid accounts per 1,000 trial starts' is that same ratio multiplied by 1,000 and answers nothing new. What the censoring needs is a fixed cohort age: read both arms at account_created_at_utc plus 60 days, which clears the long arm's full 30 + 14 + 3 = 47-day path to a settled first payment, so neither arm is censored at readout.
Then move the denominator upstream, which is a separate fix and the only one that catches a higher share of a smaller population: paid accounts per 1,000 dim_account rows created in the cohort week. Precondition — this diverges from the conversion rate only when trial length is visible before the trial starts, on pricing pages, in ads or in the signup flow. If the arms are assigned at trial start and nothing upstream differs, trial starts per 1,000 new accounts is equal across arms by construction and the two metrics are the same comparison rescaled. Report that ratio per arm and say which regime you are in rather than assuming one.
Add the leg neither denominator can see: a shorter trial converts users who had less time to reach activation, so read month-3 gross revenue churn and week-4 retention of the converted cohort, and read activation at trial day 7 to see whether the converted population changed composition.
Audit 'past_due' before comparing, since the definition counts it as converted: report the rate with and without it per arm, because a gain that sits entirely inside past_due is a billing artefact rather than a conversion effect.
Be explicit that the twelve-month consequence is not observable inside a planning cycle: pre-register that the decision is made on the 60-day metric, that the twelve-month cohort read will be published as a check, and what action follows if the two disagree.

Worked solution (35 min)

Tabulate, at a fixed calendar readout date, the share of each arm's cohorts that have completed trial end plus 14 days plus 3 days of settlement, and show the gap.
Define the decision metric as paid accounts per 1,000 dim_account rows created in the cohort week, evaluated at account_created_at_utc plus 60 days for both arms, and write the 30 + 14 + 3 = 47 arithmetic that justifies 60 as the common clock.
Report trial starts per 1,000 new accounts per arm beside it, so the reader can see whether the upstream denominator is doing any work in this test or whether it is arithmetically pinned to the conversion rate.
Compute the earliest honest readout date from the enrolment window: 60 days after the last account creation in the window plus a 3-day settlement lag, with no partial-cohort comparison permitted before it.
Add the quality leg: month-3 gross revenue churn, week-4 retention of converted accounts, and day-7 activation per arm.
Report the conversion rate with and without past_due per arm, and state which version the decision uses.

Expected result: A decision metric stated as paid accounts per 1,000 new accounts at a fixed 60-day cohort age measured from account creation, the trial-start rate published per arm so it is explicit whether that denominator change carries any information here, a readout date computed from the enrolment window, and at least two quality metrics capable of overturning a conversion-rate win.

Verify it yourself

Both arms must have identical elapsed time at readout: compute min and max of (readout_date minus account_created_at_utc) per arm and confirm the two ranges match.
Trial starts per 1,000 new accounts must be reported per arm. If that ratio is equal across arms, the conversion rate and the per-new-account metric are the same comparison rescaled and cannot diverge — so never present them as two independent confirmations. Only a gap in that ratio gives the per-new-account metric information the conversion rate does not carry.
The past_due share of the numerator should be similar across arms; a gap there is a payment-failure effect and does not belong in a conversion claim.

Follow-up

What happens to a user who would have converted on day 20, and how would you detect that population in the data?
The short arm has lower day-7 activation but higher conversion. Reconcile those two facts into one story.
What randomisation unit do you use here, and what goes wrong with the obvious alternatives?

Separate a definition edit from a behaviour change

Easy · definition-change · metric-versioning · cohorts

Seven-day activation rate is documented as core actions on at least 2 distinct UTC dates inside the first seven days after account_created_at_utc, and every cohort published through last week was computed on that definition. The dashboard now reads 38% for the signup cohort that started Monday, against 31% for the cohort before it. Two things landed in that week: an onboarding change, and a commit to the metric's SQL that you have not read yet and that came with no documentation update. You have dim_user, fct_event and the metric's version history. Deliver two numbers: how many of the 7 points are behaviour and how many are the definition. Then say which published cohorts are no longer comparable.

Approach

Diff the metric SQL across the commit boundary and enumerate every changed predicate individually: the distinct-day threshold, the is_internal exclusion, the is_core_action source, the window bounds, and the identity key. A single commit routinely changes more than one, and each has its own sign.
Dual-run both definitions over the same 26 cohort weeks. The definition effect is the level shift between the two series computed on identical data, which isolates it from anything that happened in the world.
Read the behaviour effect only from the old-definition series, comparing the new cohort to its own trailing cohorts. The new-definition series cannot answer the behaviour question because it has no pre-period.
Explain the mechanism of the changed predicate rather than only its size. Lowering a distinct-day threshold from 2 to 1 admits every user who acted once in a single session, which is a large and low-intent population, so a several-point jump is the expected magnitude and not a surprise.
State the restatement plan: pick one definition, backfill the full history on it, and annotate the break date. A series carrying two definitions is not a time series.

Follow-up

The onboarding team wants the 1-day threshold kept because it is easier to move. What is your argument, and what would change your mind?
How would you make a definition change visible to every consumer of this metric without relying on people reading a changelog?

Suggested practice checklist

For someone who can already write the query and train the model but stalls when asked what to measure or whether a change is worth making. Metric definition and case structure come first; the technical work is kept as maintenance rather than the centre of the week.

Day 1 — Metric anatomy

For three products you use daily, write one primary metric, two input metrics that plausibly move it, and one guardrail that would catch a cheap way of moving the primary at the cost of the product.
For one of them, specify the metric precisely enough that two analysts would return the same number: numerator, denominator, unit of observation, time window, and how returning and deleted accounts are treated.
Pick a ratio metric and write what happens to it when the denominator shrinks for reasons unrelated to the numerator, with a concrete example of that happening.

Deliverable: A one-page metric tree for one product, with the primary metric written as an unambiguous spec.

Day 2 — Diagnosing a drop without guessing

Take the prompt "weekly active users fell 8 percent week over week" and write the segmentation plan before proposing any cause: platform, region, tenure cohort, acquisition channel, and whether the movement sits in the numerator or in a changed denominator.
List the instrumentation failures that manufacture fake drops (a client release that stopped firing an event, a bot filter change, a shifted date boundary or timezone) and write the query that rules out each one.
Rehearse stating the boring explanations first, seasonality and day-of-week composition, before reaching for a product cause.

Deliverable: A drop-diagnosis checklist short enough to recite from memory in under a minute.

Day 3 — Should we build it

Take a feature idea and write it as a bet: what you believe is true, what would have to be true for it to pay off, the metric that would confirm it, and the effect size that would justify the engineering cost.
Size the opportunity top-down and bottom-up, then reconcile the two numbers in writing instead of quoting whichever is friendlier.
Write the counter-metric that would make you kill the feature even if it wins on the primary metric.

Deliverable: A one-page product memo ending in a decision rather than a list of considerations.

Day 4 — The places aggregate numbers lie

Construct a Simpson's paradox numerically: two segments where the treatment wins within each segment yet loses overall, and identify the shift in segment weights that causes it.
Take a heavy right-tailed quantity such as revenue per user and write why the mean is the wrong summary, which percentile you would report instead, and what a moving mean with a stable median tells you.
Write your definition of a session for the product from day one, then name two real behaviours it misclassifies.

Deliverable: One page holding a worked Simpson's paradox table and a session definition with its two known failure cases.

Day 5 — Technical maintenance, aimed at metrics

Solve four timed SQL prompts that all end in a ratio metric, so the question of grain stays live in every answer.
Compute a 95 percent confidence interval for a proportion on a small sample, and state why the normal approximation is unreliable when either np or n(1 minus p) falls below roughly 10, along with which interval you would use instead.
Take one metric from your day-one tree, write the query that computes it correctly, then write the query that computes it wrong in the most plausible way and explain how you would notice.

Deliverable: Four solved prompts plus a matched correct and plausible-wrong query for one metric.

Day 6 — Turning engineering work into data science stories

Write three project stories as situation, decision, trade-off, outcome, each carrying one number and one thing you got wrong.
For the story you will lead with, prepare an answer to "what would you do differently" that names a decision you made, not a constraint you were handed.
Practise the sentence that reframes a systems project as a question project: the question the work answered, ahead of the pipeline it shipped.

Deliverable: Three written stories with the lead story delivered aloud and timed under four minutes.

Day 7 — Mock case and gap list

Run a 40-minute mock case with someone playing a product manager who pushes back on your metric choice, and record it.
Listen back and mark every moment you proposed a solution before the success metric existed.
Rewrite those moments as the question you should have asked, and rehearse the first 90 seconds of the case until scoping comes before solving.

Deliverable: A recorded case plus a rewritten opening 90 seconds.

Being wrong in public

A number you shipped turned out to be wrong, and someone had already acted on it. That is one of the most useful stories a data person can carry. What is being scored is how fast you noticed, who you told first, and what you changed in the process so the same class of error could not repeat quietly.

How do you handle disagreement within a technical team regarding the choice of a model or tool?

Medium · behavioural and stakeholder questions

Approach

Close with what you would do differently, concretely.
Quantify the outcome, including what you would not claim credit for.
Name the disagreement or constraint, and how you resolved it with evidence.

Follow-up

How did you know the outcome was caused by your change?
What did you decide not to do, and why?

How do you handle imbalanced datasets in fraud detection scenarios?

Medium · behavioural and stakeholder questions

Approach

State the situation in two sentences and spend the rest on your reasoning.
Pick a story where you drove the decision, not one where you observed it.
Close with what you would do differently, concretely.

Follow-up

What did you decide not to do, and why?
How did you know the outcome was caused by your change?

Turn an ambiguous onboarding question into a measurable metric

Easy · scoping · metric definition · stakeholder

Two days before a planning review, a director asks whether onboarding is working. You have dim_user (account_created_at_utc, signup_surface, is_internal), fct_event (is_core_action, flow_id, flow_instance_id, event_name, occurred_at_utc, received_at_utc) and fct_session. No further meeting with the director is possible before you start work. Deliver three clarifying questions you would send in writing, the metric you will compute in the meantime with its numerator, denominator, window and exclusions, and one sentence naming the question you are deliberately not answering.

Approach

Recognise what is being probed: whether you convert a goal into a computable predicate without stalling for requirements or guessing in silence. Listing clarifying questions is the generic answer; shipping a defensible default alongside them is the strong one, because the review is in two days and it will happen with or without you.
Infer the decision behind the request. A question about whether onboarding works, arriving before a planning cycle, usually means whether to staff it next quarter. That points at a rate with visible headroom over several cohorts, not at a descriptive dashboard.
Write the three questions so that each one changes the SQL. Which population, all signups or only self-serve from dim_user.signup_surface. What counts as working, reaching a core action or completing the onboarding flow_id. Against what bar, last quarter's cohorts or a stated target.
Propose the default explicitly: seven-day activation on weekly signup cohorts. Numerator, users with is_core_action = TRUE events on at least two distinct UTC dates inside [account_created_at_utc, account_created_at_utc + 7 days). Denominator, the signup cohort with is_internal = FALSE. Publish with an eight-day lag, and state that the two-distinct-days threshold is a frozen choice rather than a discovery.
Name the exclusion in the same breath as the number. The series shows whether users activate; it does not establish that onboarding caused the level, which needs a staged rollout or an experiment.

Follow-up

The director replies that they meant the onboarding flow specifically, not activation. What changes in the query and in the caveats?
Your cohort metric needs an eight-day lag and the review is in two days. What do you present, and how do you label it?
Two of your three questions come back unanswered. Which one do you refuse to proceed without?