adjoe Data Scientist Interview Questions & Guide 2026

The 60-second brief

As a Data Scientist at adjoe, you sit at the intersection of high-scale mobile advertising, user behavior analysis, and predictive modeling. adjoe operates a massive, complex ecosystem where the ability to derive actionable insights from billions of events is not just a competitive advantage—it is the core of the business. You will be tasked with optimizing ad delivery, refining recommendation engines, and ensuring that the platform remains both profitable and user-centric.

Nearly every loop contains a round whose deliverable is a recommendation to someone non-technical. Practise stating a conclusion, the confidence attached to it, and the cost of being wrong in each direction, because that triple is the artifact being graded.

Focus areas

• Turn a vague request into a measurable question
• Size an experiment before anyone launches it
• Pick a randomisation unit that respects interference

Overview

The role is highly product-biased, meaning your work directly influences how users interact with mobile applications and how advertisers reach their audiences. You will work closely with engineering and product teams to translate ambiguous business challenges into rigorous analytical frameworks. Whether you are diagnosing a sudden drop in a key metric or designing an experimentation roadmap, your impact is immediate and visible across the company's entire product portfolio.

The adjoe interview process

adjoe candidates report 4 rounds · ≈ 3-5 weeks. The stages below are what candidates describe, not a published process.

Recruiter Screen

Most candidates lose this call inside the first two minutes, during the walkthrough of their own background. The account runs chronologically, sits at the level of tools and titles, and never arrives at a decision anyone could have disagreed with. Anchor on a problem instead of a timeline: what the team could not answer, what you did about it, what happened next. Ninety seconds is enough, and stopping on time leaves room for the half of the call that belongs to you. What you ask about how work gets prioritised signals your level more reliably than the walkthrough does.

What to demonstrate

• Whether your background summary has a shape (problem, decision, consequence) or is a chronological list of tools and employers
• Whether you can account for gaps, short stints and the reason you are looking, unprompted and without hedging
• The substance of the questions you ask back, which an experienced screener reads as a level signal

How to prepare

• Time your opening walkthrough against a clock. If it runs past two minutes, compress the earliest role into a single clause and spend the recovered time on the most recent one
• Write one honest sentence for every gap or short stint visible on your resume and offer it before being asked about it
• Prepare questions about how work arrives and gets prioritised: who writes the request, how often priorities change, and what happens to an analysis after it is delivered

Technical 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

Team Evaluation

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

Leadership Interviews

Because the format is not fixed, prepare the reasoning rather than the ritual. Nearly every version of this round draws on the same underlying material: a design you can defend, a metric you can define exactly, an analysis whose assumptions you can state out loud. Only the wrapper changes, whether that is a take-home, a live case, a deep dive on past work, or a rough estimate on a whiteboard. Answers rehearsed to fit one shape stall the moment the shape differs. Practise naming the assumption behind a number, then saying how much the conclusion moves if that assumption is wrong.

What to demonstrate

• Whether your justification for a method survives the question 'why not the simpler thing', including when the simpler thing would have worked
• Precision under pressure: what exactly counts as an active user, a conversion or a success, over what window, with what exclusions
• Whether you carry an argument through to a recommendation instead of stopping at a list of tradeoffs

How to prepare

• For each project you plan to mention, write the metric definition in one sentence: numerator, denominator, time window, exclusions. Say it out loud once, because vagueness shows up in speech before it shows up on paper.
• Rehearse the same project at three lengths: two minutes, ten minutes, and a deep dive on one technical decision. Cutting live is harder than it sounds.
• For your headline result, write down what would have had to be true for it to be wrong, and how you ruled that out.

Where candidates lose points

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.

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.

SQL that silently fans out on a one-to-many join

State the grain of each table and the grain you want in the result before writing the join. Pre-aggregate the many side to the join key, or use EXISTS or a window function, and verify with a row count against COUNT(DISTINCT id) rather than trusting that the numbers look plausible.

Averaging per-user rates to produce a population rate

Decide which quantity you want: the mean of per-user ratios and the ratio of summed numerator to summed denominator are different estimands, and heavy users dominate one but not the other. For a ratio metric, aggregate numerator and denominator separately and use the delta method for its variance.

Questions & practice

Audit a one-day event extract for structural defects

Easy · data quality · late arrival · pandas

You receive a one-day extract of fct_event as a DataFrame with event_id, occurred_at_utc, received_at_utc, visitor_id, user_id, account_id, event_name, is_bot_flagged and surface. Write a function returning one row per data-quality rule with the rule name, the failing row count and the failing share of the extract. Cover at minimum: duplicate event_id, received_at_utc earlier than occurred_at_utc, occurred_at_utc later than the extract's maximum received_at_utc, account_id present while user_id is NULL, and rows whose occurred_at date differs from their received_at date. Do not drop rows; report only.

Approach

1. Compute the extract's own reference clock first: max(received_at_utc). Wall-clock now() is wrong here because the extract may be replayed days later, which would turn every row into a future-dated failure.
2. Express each rule as a boolean Series over the same index so the checks compose, then aggregate with .sum() and divide by len(df). Building a list of (name, mask) pairs keeps the rule set extensible and keeps one code path for counting.
3. For the duplicate rule, decide and state the convention: df.duplicated('event_id', keep=False).sum() counts every member of a duplicated group, df.duplicated('event_id').sum() counts only the surplus copies. Either is defensible; an unstated choice is not. The rest of this item assumes keep=False.
4. Treat received_at < occurred_at as clock skew, not corruption: occurred_at is client-supplied. Separate it from the date-mismatch rule, which is the one that actually breaks a daily metric keyed on occurred_at.
5. Know which rules imply which before you read the counts. A row whose occurred_at exceeds max(received_at_utc) has its own received_at no later than that maximum, so it is necessarily a clock-skew row as well: the future-dated mask is a subset of the skew mask, always. Neither is a subset of the date-mismatch mask, because skew of a few minutes inside one UTC date mismatches nothing.
6. Return a tidy DataFrame sorted by failing_share descending, and add a boolean column saying whether the rule should block publication, so the output is a decision rather than a list of numbers.

Worked solution (20 min)

1. Parse both timestamp columns with utc=True and assert the dtype, since a silently-object column makes every comparison string-wise and wrong.
2. Set ref = df['received_at_utc'].max() and build the five masks against it.
3. Assemble results as pd.DataFrame(rows) with columns rule, failing_rows, failing_share, blocks_publication.
4. Keep the masks addressable (a dict of name to Series) rather than only their sums, so the overlap between rules can be asserted rather than assumed.
5. Verify the function is pure: assert the input frame's shape is unchanged after the call.

Expected result: A DataFrame with one row per rule, failing_share equal to failing_rows divided by len(df) for every row, and the input frame returned unmodified. The rules overlap rather than partition the extract, and one containment is structural: every future-dated row is also a clock-skew row. The counts therefore must not be summed or presented as a total.

Verify it yourself

• Plant violations deliberately and assert an exact expected vector rather than assuming one hit per rule. On a 10-row frame carrying two rows that share an event_id, one same-date skew row, one row whose occurred_at sits an hour past max(received_at_utc), one row with account_id set and user_id NULL, and one row whose occurred_at falls on the previous UTC date: duplicate_event_id is 2 under the keep=False convention, clock_skew is 2 (the planted skew row plus the future-dated row, which cannot avoid tripping it), future_dated is 1, orphan_account is 1, date_mismatch is 1.
• Assert the containment directly rather than by counting: (future_dated_mask & ~clock_skew_mask).sum() is 0. A non-zero value means the reference clock was not max(received_at_utc).
• Feed a clean frame and confirm all counts are 0 and no rule errors on empty masks.
• Confirm len(input) is identical before and after, since the deliverable is a report and not a cleaned frame.

Follow-up

• The date-mismatch count is 2.1 percent on this extract. What late-arrival rule would you write for a daily metric, and how many days would you hold the number open?
• Duplicate event_id values appear only on the 'core_action_completed' event. What upstream cause would you check before deduplicating?
• Which of these rules should fire an alert at the pipeline, and which should only appear in a weekly review?

Rebuild per-visitor ordering without groupby convenience methods

Easy · pandas · vectorisation · window logic

You have a DataFrame of 2 million fct_event rows with visitor_id, occurred_at_utc and event_id, unsorted and containing duplicate timestamps within a visitor. Produce three new columns: event_rank, the 1-based position of the event within its visitor ordered by occurred_at_utc; seconds_since_prev, the gap to that visitor's previous event, NULL for the first; and is_first_for_visitor. You may use sort_values, shift, cumsum, numpy and boolean masking. You may not use groupby.transform, groupby.apply, groupby.cumcount, groupby.rank or merge_asof. Break timestamp ties on event_id.

Approach

1. Sort once by ['visitor_id', 'occurred_at_utc', 'event_id'] and reset the index. The whole exercise reduces to row arithmetic on a sorted frame, and the tiebreak on event_id is what makes the result reproducible across runs.
2. Mark visitor boundaries with is_first = df['visitor_id'].ne(df['visitor_id'].shift()). This is the single fact every other column derives from.
3. Compute seconds_since_prev as the diff of the timestamp column, then overwrite it with NaT/NaN wherever is_first is True. The shift crosses the boundary between visitors and will otherwise hand the first row of each visitor the last event of the previous one.
4. Build event_rank from a running counter that resets at boundaries: take a global cumulative position (np.arange(len(df))) and subtract, per row, the global position at which that visitor started. Get the start position by forward-filling the positions where is_first is True, which is a cumsum-free reset and is O(n).
5. Verify against the forbidden method once, as a test rather than as the implementation, and confirm the two agree on every row.

Follow-up

• The frame does not fit in memory. How does your approach change if you can only process one visitor-partitioned chunk at a time?
• occurred_at_utc is client-supplied and sometimes runs backwards within a visitor. Does your seconds_since_prev go negative, and should it?
• How would you extend this to reset the counter at every change of surface as well as visitor?

Simulate the false positive cost of repeated peeking

Medium · simulation · peeking · type i error

Quantify the cost of peeking. Simulate a two-arm experiment with no true effect: each arm accumulates Bernoulli conversions at a base rate of 0.10 up to 40,000 units per arm. Run a two-sided two-proportion z-test at alpha 0.05 at ten equally spaced interim points, and record whether the test ever crossed. Report the false positive rate over at least 10,000 replications, alongside the rate for a single look at the final sample only. Use a fixed seed and report a Monte Carlo standard error on both figures.

Approach

1. Generate each replication as two cumulative sums of Bernoulli draws, then read the interim points off the cumulative arrays. Regenerating data at each look would make the looks independent, which destroys exactly the dependence the exercise is about: later looks share data with earlier ones.
2. Use the pooled-variance two-proportion z: p_pool = (x1+x2)/(n1+n2), z = (p1-p2) / sqrt(p_pool*(1-p_pool)*(1/n1 + 1/n2)), reject when |z| > 1.96. State that the normal approximation is fine here because the smallest look has roughly 400 expected conversions per arm.
3. Vectorise across replications rather than looping: draw a (reps, n) array of uniforms, threshold at 0.10, cumsum along axis 1 and slice the ten look indices. A per-replication loop at 10,000 by 40,000 is unnecessarily slow.
4. Record the any-cross indicator per replication, take the mean, and compute the Monte Carlo standard error as sqrt(p*(1-p)/reps) so the reported figure comes with its own precision.
5. Report the single-look rate in the same run as a control. If it does not land near 0.05, the bug is in the test statistic and not in the peeking argument.

Follow-up

• Re-run with 40 looks instead of 10. Why does the curve flatten rather than continue rising linearly?
• Among the replications that crossed, what is the mean observed lift, and why is it not zero?
• What does an O'Brien-Fleming boundary or an always-valid confidence sequence change about this simulation, and what does each cost in power?

Write a SQL query to calculate the rolling 7-day average of active users using window functions.

Medium · sql and data manipulation

Approach

1. Compute rates by summing numerator and denominator separately, never by averaging rates.
2. State the window function and its partition and ordering out loud before writing it.
3. Handle the rows that do not match: a LEFT JOIN with a NULL check is usually the question.

Follow-up

• How does the query change if the join becomes one-to-many?
• What breaks if events arrive late or out of order?

How would you identify "power users" in a large, sparse dataset using SQL?

Medium · sql and data manipulation

Approach

1. Check whether any join is one-to-many before aggregating, or the sums inflate.
2. Compute rates by summing numerator and denominator separately, never by averaging rates.
3. Handle the rows that do not match: a LEFT JOIN with a NULL check is usually the question.

Follow-up

• How would you verify this result without re-running the same query?
• How does the query change if the join becomes one-to-many?

Given a table of ad impressions and clicks, how do you calculate the conversion rate per user segment?

Medium · sql and data manipulation

Approach

1. Check whether any join is one-to-many before aggregating, or the sums inflate.
2. Compute rates by summing numerator and denominator separately, never by averaging rates.
3. Handle the rows that do not match: a LEFT JOIN with a NULL check is usually the question.

Follow-up

• How does the query change if the join becomes one-to-many?
• How would you verify this result without re-running the same query?

Read an experiment from first exposure, not assignment

Hard · experiments · window-functions · srm

fct_experiment_exposure holds experiment_id, unit_type, unit_id, variant, user_id, assigned_at_utc, first_exposed_at_utc, is_in_analysis_population and planned_end_utc. fct_event holds user_id, occurred_at_utc, is_core_action, and carries events up to a known data cut, :data_cut_utc. For one experiment randomised on unit_type = 'user', return per variant: exposed units, units with at least one core action in the seven days after that unit's own first exposure, the rate, and the variant share of exposed units. Only units whose seven-day window has fully elapsed as of the data cut belong in the readout. Units appearing under more than one variant are excluded from both arms and counted separately.

Approach

1. Run the contamination pass as an aggregate, not a window: SELECT unit_id FROM fct_experiment_exposure WHERE experiment_id = :exp GROUP BY unit_id HAVING COUNT(DISTINCT variant) > 1, then anti-join it away. PostgreSQL rejects COUNT(DISTINCT variant) OVER (PARTITION BY unit_id) outright, since DISTINCT is not implemented for window functions; if you want the test inline, MIN(variant) OVER (PARTITION BY unit_id) <> MAX(variant) OVER (PARTITION BY unit_id) is the equivalent that does run.
2. Do not resolve contamination by keeping the earliest variant. A unit that saw both arms carries treatment from both, so assigning it to either one biases that arm.
3. Define the population as is_in_analysis_population = TRUE AND unit_type = 'user' AND first_exposed_at_utc < planned_end_utc AND first_exposed_at_utc + interval '7 days' <= :data_cut_utc. The horizon filter is what makes the readout reproducible next week instead of drifting with every re-run; the data-cut filter is the one that actually buys seven days of follow-up, since a unit exposed an hour before the horizon otherwise contributes an hour of observation to a seven-day rate.
4. Measure the outcome on a per-unit relative window: LEFT JOIN fct_event on user_id with is_core_action = TRUE and occurred_at_utc in [first_exposed_at_utc, first_exposed_at_utc + interval '7 days'). LEFT JOIN so units with no outcome stay in the denominator at zero rather than being deleted by an inner join.
5. Check the sample ratio before reading the effect: variant share of exposed units against the intended split, tested as a binomial. Run it on the truncated population as well as on the full exposed set, because if one arm exposes later on average the data-cut filter removes more of that arm and can manufacture a ratio mismatch the randomisation did not have. A mismatch on the full set means the exposure data is not a valid randomisation and invalidates the readout rather than being a footnote under it.
6. Report the per-variant rate, the absolute difference, and the fact that the variance unit is unit_id. That is straightforward here only because the grain is already one row per user; a per-session outcome under user randomisation would need a delta-method or bootstrap standard error instead.

Worked solution (35 min)

1. Count units per variant and units with more than one variant; report the contaminated count before anything else.
2. Build the exposed population CTE with both the horizon filter and the seven-days-before-data-cut filter, assert one row per unit_id, and record per variant how many units the second filter removed.
3. LEFT JOIN the bounded outcome events and flag conversion with a boolean aggregate.
4. Aggregate per variant to exposed, converted, rate and share, and run the binomial check on the share, both before and after the truncation.
5. Re-run with a fixed calendar window from experiment_start_utc and record the difference as the size of the dilution you avoided.

Expected result: One row per variant with exposed, converted, rate and share, plus a scalar count of excluded contaminated units. Every retained unit was first exposed at least seven days before the data cut, so each one has a full seven days of observation and the rates are directly comparable.

Verify it yourself

• Exposed counts across variants sum to the population count minus the contaminated units.
• MAX(first_exposed_at_utc) among retained units is at least seven days before :data_cut_utc.
• Variant shares sum to 1.
• No unit_id contributes to two variant rows.
• The truncation drops a similar share of each arm; if it does not, note the imbalance alongside the SRM result rather than reading the effect.
• Swapping the relative window for a calendar one changes the rates; if it does not, exposure timing is balanced and you can say so.

Follow-up

• Some units were assigned days before they were exposed. What does analysing the assigned set instead do to the estimated effect, and in which direction?
• The split is 51/49 on 400,000 exposed units. Do you read the result?
• The treatment arm exposes on average two days later than control. What does that do to a fixed calendar outcome window, and which arm does it favour?

A key engagement metric has suddenly dropped by 10%. How do you systematically diagnose the root cause?

Medium · product sense and metrics

Approach

1. Restate the decision this analysis has to support, and who acts on the answer.
2. Fix the population and the time window before naming any metric.
3. Name one primary metric, then the guardrail that stops it being gamed.

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 metric to measure the success of a new ad placement?

Medium · product sense and metrics

Approach

1. State what result would change your recommendation, so the answer is falsifiable.
2. Decompose the metric into the rates that drive it, and say which one you would check first.
3. 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 do you balance short-term revenue gains with long-term user retention?

Medium · product sense and metrics

Approach

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

Follow-up

• Which segment would you cut first, and what would that rule out?
• What would you do if the primary metric and the guardrail moved in opposite directions?

How do you define statistical significance for a test with a small sample size?

Medium · a/b testing and experimentation

Approach

1. Name the guardrails that would stop a launch even on a positive primary result.
2. Name the randomisation unit first; it decides the variance and what the test can detect.
3. Say whether units interfere with each other, and switch design if they do.

Follow-up

• What would you do if you could not randomise at all?
• How would you handle interference between treated and control units?

How would you explain the concept of regularization to a non-technical person?

Medium · a/b testing and experimentation

Approach

1. Name the randomisation unit first; it decides the variance and what the test can detect.
2. State the primary metric and the minimum effect worth shipping, then size the test.
3. Say whether units interfere with each other, and switch design if they do.

Follow-up

• What would you do if you could not randomise at all?
• How would you handle interference between treated and control units?

A surrogate for twelve-month value inside a three-week test

Hard · surrogate metrics · experimentation · long-term value

A monetisation change — an earlier and harder paywall — will be tested for three weeks. The decision quantity is twelve-month cumulative net revenue per acquired account, which cannot be observed in three weeks. Available: fct_event, fct_session, fct_subscription_period (mrr_cents_constant_fx, period_status, change_reason, is_first_paid_period), dim_account, and fct_experiment_exposure (unit_type, unit_id, variant, first_exposed_at_utc, is_in_analysis_population). Construct a surrogate index readable at three weeks, state the assumption that makes it valid, name the mechanism that breaks it, and give the pre-registered rule for refusing to decide. Deliverable: the index, the assumption, and the refusal rule.

Approach

1. State the surrogacy condition before building anything: a surrogate is valid only if the treatment's entire effect on twelve-month revenue runs through it. Then say where it fails here, because a paywall moves short-run revenue directly and long-run revenue through churn and through who becomes a payer at all.
2. Fit the index on history rather than on intuition: regress twelve-month cumulative net revenue per account on features observable by day 21 — first-paid flag, activation days, week-3 core-action days, seats billed — using cohorts old enough to have a twelve-month outcome, and hold out a later cohort.
3. Report the out-of-sample fit as the headline rather than the point estimate: the held-out R-squared and the calibration of predicted against actual deciles are what license any use of the index, and a decile plot catches the mis-calibration an R-squared hides.
4. Name the bias direction explicitly: the relationship was estimated under the old paywall, so under a harder paywall the marginal payer is a different person and the index over-predicts their value, biasing the treatment arm optimistic.
5. Surround the index with the direct three-week readouts it cannot contain — revenue per exposed unit, cancel-within-first-period rate, and the free-side signup and activation counts — and pre-register a numeric refusal rule, because the failure mode of a surrogate is being used confidently in exactly the case it was not fit for.

Worked solution (40 min)

1. Assemble the training set: accounts with first_paid_at_utc at least twelve months old, their realised twelve-month cumulative net revenue, and the day-21 features.
2. Fit on pre-change cohorts, evaluate on a held-out later cohort, and report both out-of-sample R-squared and predicted-versus-actual decile calibration.
3. Write the surrogacy assumption as a falsifiable sentence, then write the specific mechanism in this treatment that violates it.
4. Specify the direct arm readouts on fct_experiment_exposure with is_in_analysis_population = TRUE: revenue per exposed unit, cancel-within-first-period rate, and free-side signup and activation counts.
5. Write the refusal rule with numbers attached, covering both the fit threshold and the free-side divergence case.

Expected result: A fitted surrogate with a reported out-of-sample R-squared and calibration, a written surrogacy assumption plus the mechanism that violates it, the resulting bias direction, and a numeric refusal rule agreed before launch.

Verify it yourself

• Standard errors are clustered on fct_experiment_exposure.unit_id, the identifier of the randomisation unit. unit_type is the kind label for that unit — a single value such as 'user' across the whole table — so clustering on it collapses the design to one or two clusters and carries no information about who was randomised together. A per-event or per-session standard error on a user-randomised test is the opposite error: too small, and it will manufacture significance.
• The analysis population is the exposed set with is_in_analysis_population = TRUE, and a sample-ratio check against the intended split passes before any outcome is read.
• The index applied to the control arm should reproduce the historical relationship. If control is already mis-calibrated, the treatment estimate is not interpretable at all.

Follow-up

• What evidence would make you trust a surrogate for this particular change rather than a different one?
• The index reads plus 8 percent while three-week cancel-within-period is up. Which do you act on, and what do you tell the decision-maker?
• How would you size this test, on what unit, and what does that do to the three-week horizon?

Size every candidate cause of a trial-to-paid decline

Hard · cohort-maturity · experiment-readout · mix-shift · contribution-sizing

Trial-to-paid conversion on weekly trial-start cohorts from fct_subscription_period reads 3.1 points below the trailing eight-week mean for the three most recent cohorts. Three things happened in that window: a pricing experiment reached 50% of new trials, a payment processor migration added settlement delay, and paid_search spend tripled. Using fct_subscription_period, fct_experiment_exposure and dim_user, rank the causes by their contribution in points of the headline, state the remainder, and give the decision you would take on Monday.

Approach

1. Kill the immature cohorts first, because everything downstream is computed on them. The metric is lagged by the trial length plus a 14-day conversion window plus a settlement allowance, and a processor migration lengthens exactly that last term; recompute each cohort at a fixed cohort age rather than as of today, and confirm the newest cohort's value is still climbing day over day.
2. Hold the experiment analysis to the exposed population. Join fct_experiment_exposure on unit_id with is_in_analysis_population = TRUE rather than reading an assignment log, then check the variant split for a sample-ratio mismatch before believing any effect at all. Contribution to the headline is the variant effect multiplied by the exposed share, which is not the same number as the variant effect.
3. Decompose the cohort mix by dim_user.first_touch_channel using the same weight-times-rate arithmetic as any other mix question, so the paid_search increase is sized as a weight change at a measured conversion rate rather than asserted from the spend figure.
4. Convert all three to points of the headline, sum them, and print the residual against the historical week-to-week standard deviation of the metric. If the residual is inside that band, say so and stop looking; if it is outside, name what you would investigate next rather than leaving it implied.
5. Land the decision. Only one of the three is actionable on Monday, so state whether the experiment has accrued enough exposed units to stop at the pre-declared horizon, and state separately what the settlement-lag correction does to the published series and its lag rule.

Follow-up

• How do you choose the fixed cohort age, and what do you lose by choosing it too long?
• If the pricing variant is genuinely 1.2 points worse, does that settle whether to stop it? What else is on the other side of that decision?
• The trailing eight-week mean spans the processor migration. What is the right baseline instead?

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.

Bring numbers, including the ones you are unsure of

An answer without a quantity is hard to interrogate, so interviewers keep probing until they find one. Come with the baseline, the change, the window it was measured over, and how confident you were. If the effect never got measured, say so and say what you would have measured. Fabricated precision is worse than an honest gap.

Tell me about a time you navigated an ambiguous problem without a clear roadmap.

Medium · behavioural and stakeholder questions

Approach

1. Close with what you would do differently, concretely.
2. Name the disagreement or constraint, and how you resolved it with evidence.
3. Pick a story where you drove the decision, not one where you observed it.

Follow-up

• How did you know the outcome was caused by your change?
• What would you do differently if you ran that project again?

Explain a wide interval to a non-technical executive

Medium · communication · uncertainty · pricing

A pricing change is under consideration. Your best estimate of its effect on trial-to-paid conversion is a 1.8pp drop, with a 95% interval from a 4.6pp drop to a 1.0pp rise, read from a geo holdout rather than a randomised test. An executive preparing a board slide asks you for 'the number'. You have ninety seconds and one slide, and the words confidence interval, p-value and significance are not usable with this audience. Deliver the slide headline, the single supporting line, and what you say aloud.

Approach

1. Recognise what is being probed: whether you can carry uncertainty into a decision instead of either hiding it or hiding behind it. The generic answer promises to explain the interval in plain English; the strong one replaces the question 'what is the number' with 'across this range, where does the decision change'.
2. Find the threshold before you draft anything. Ask what the pricing case assumes, then compute the conversion drop at which the higher price stops adding revenue: price uplift on the conversions kept against the revenue lost from conversions forgone. That single figure is what makes the range legible.
3. Restate the estimate and both bounds in the unit the audience already reasons in. Convert percentage points into monthly first-paid conversions at current trial volume, then into mrr_cents_constant_fx, so the slide reads as money per month rather than as statistics.
4. Place the range against the break-even and say which part of it sits on each side. If most of the range clears the threshold, that is a recommendation to proceed with a monitoring plan; if the range straddles it, that is a recommendation to narrow the range first.
5. Name what would narrow it and what that costs in weeks, then give one recommendation with an explicit condition for revisiting it. Uncertainty stated without a next step is read as indecision and the midpoint gets used anyway.

Follow-up

• The executive says to give the midpoint and they will manage the risk. What do you do?
• How does the slide change if the interval were a 4.6pp to 0.2pp drop, with no positive outcomes in range?
• Why is a geo holdout the credible read here rather than the attributed channel numbers you already have?

Choose between three teams' requests with one analyst-week

Medium · prioritisation · stakeholder · judgement

You have one analyst-week. Three requests land the same morning. A growth team wants a paid-channel readout before a Friday spend decision. A billing team wants gross monthly revenue churn rebuilt, because the current figure recognises cancellation at canceled_at_utc rather than period_end_utc and is therefore wrong. A product team wants a dashboard for a feature launching in six weeks. All three sponsors are peers of your manager. Deliver your ranking, the explicit rule that produced it, and the message you send to the two teams you defer.

Approach

1. Recognise what is being probed: whether you prioritise on decision value and reversibility or on who asked most recently and most loudly. The generic answer sorts by importance; the strong one states a rule, applies it, and accepts the ranking it produces even where that is uncomfortable.
2. Score each request on three statable things: the decision it unblocks and the date that decision is made, the cost of being wrong in the meantime, and whether the work is one-off or compounding. A wrong published churn figure compounds, because it is quoted downstream and enters forecasts; the channel readout has a fixed date that cannot move; the dashboard has six weeks of slack.
3. Notice the tension between value and urgency rather than resolving it by feel. The churn defect is the most valuable item and the least urgent one, which is exactly the shape of work that never gets done.
4. Break the churn item in two. A one-hour severity check, sizing the gap between the two recognition points in MRR, is cheap enough to do before ranking anything and may promote the item outright. Do that first, then rank.
5. Make the deferrals concrete. Each deferred team gets a date, a reason expressed as another team's decision deadline rather than as relative importance, and the smallest thing you can hand them immediately.

Follow-up

• The dashboard team escalates to your manager. What do you say in that conversation?
• Your severity check shows churn is overstated by 15%. Does the ranking change, and does anybody need to be told today regardless of the ranking?
• A fourth request arrives Wednesday with a Thursday deadline. What comes off the list, and who do you tell first?