Interview guides · Flock Safety

Flock Safety Data Scientist Interview Questions & Guide 2026

A two-week prep plan for Flock Safety data scientist candidates: write your answer before reading the solution, revisit misses three days later.

The 60-second brief

As a Data Scientist at Flock Safety, you are at the intersection of computer vision, public safety, and large-scale data infrastructure. You will work on problems that directly impact the mission of eliminating crime and helping communities thrive. This role is not just about building models; it is about extracting actionable insights from a massive, real-world sensor network to solve complex, high-stakes problems.

Learn the economics of the product category before the loop. Marketplaces, subscription products and ad-supported products turn on different core quantities (match rate and liquidity, retention and churn, fill rate and yield) and fail in different characteristic ways.

Focus areas

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

Overview

You will contribute to a product ecosystem that processes significant volumes of data daily. Whether you are optimizing algorithmic performance, analyzing patterns in vehicle telemetry, or designing experiments to improve system reliability, your work will directly influence the security and efficiency of the Flock Safety platform. This role requires a balance of rigorous technical execution and the ability to translate complex data findings into strategic business decisions.

Candidates should prepare to articulate how their technical work contributes to the broader mission of public safety, as the company highly values alignment with its core values.

The Flock Safety interview process

Flock Safety candidates report 3 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

Hiring Manager Interview

Underneath the questions about your past work sits a resourcing question. Given four things worth doing and one of you, which gets done and what happens to the rest? Managers ask because that is the daily texture of the job, and because the answer shows whether you rank work by effort or by what it changes. The weak version sorts by personal interest or by whoever asked most insistently. The strong version ties each candidate piece of work to a decision somebody downstream is waiting on, and then names the one you would drop and who you would tell.

What to demonstrate

Whether you rank work by the decision it unblocks or by how interesting the method is
How you describe a request you declined, and whether you can say who you said it to
Whether your sense of how long something takes survives one follow-up question about the messy part
How you decide something is good enough to hand over unfinished

How to prepare

Write out your current queue and, next to each item, the decision that stays stalled until it lands. Anything with no waiting decision becomes your example of work you would cut
Rehearse turning down a plausible stakeholder request out loud, including the smaller alternative you offered instead
Have one case where you shipped a rough answer early and one where you refused to, with the reason that separated them

Multi-Stage Round

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

Slicing a flat experiment until a segment reaches significance

Testing one metric across twenty segments at a nominal 5% level produces a significant result about two thirds of the time when nothing is happening anywhere, and the segment that surfaces is by construction the one with the most favourable noise. The reported effect in that slice is then badly overstated, because selection on significance conditions the estimate on being large. What makes it dangerous rather than merely wrong is that a post-hoc segment always has a plausible story attached, so it survives the meeting. The controls are declaring the small number of segments of interest before launch, correcting across the ones tested, and treating anything discovered afterwards as a hypothesis that needs its own adequately-powered test rather than a finding.

Treating last-touch attribution as the causal value of a channel

The attribution label on dim_user is the output of a rule that assigns full credit to whichever touch happened to be recorded last inside a lookback window, and that rule systematically rewards channels that sit close to the conversion, especially branded search and retargeting, which largely intercept demand that already existed. Reallocating spend on those labels moves budget toward the channels that are best at being last, which is why attributed return on ad spend often improves while total signups do not. Nothing in the touchpoint data can settle this, because the counterfactual of not running the channel was never observed. The credible reads are a geo holdout or a scheduled pause, sized in advance on the total-signups metric rather than on the attributed one, and the honest framing in the meantime is that the label describes correlation with conversion and not incremental contribution.

Dropping rows with missing values without naming the mechanism

Say whether the values are missing at random, missing by a known process, or missing in a way that depends on the outcome, and handle them accordingly. Deleting incomplete rows silently redefines the population whenever missingness correlates with what you are measuring.

Stopping an experiment the moment it crosses significance

Fix the sample size or duration before launch, or use a method built for continuous monitoring such as a sequential test, always-valid confidence intervals, or group-sequential boundaries. Repeatedly checking a fixed-horizon p-value against 0.05 pushes the real false-positive rate well above 5 percent.

Questions & practice

Can you explain the bias-variance tradeoff in machine learning and how you manage it in production models?

Medium · machine learning and modelling

Approach

Check what information would not exist at prediction time, and exclude it.
Frame the prediction: the label, the moment of prediction, and the action it triggers.
Pick an evaluation metric that matches the cost of each error type, not a default.

Follow-up

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

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.

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?

Sessionise an event stream with gap and midnight rules

Hard · sessionisation · vectorisation · event streams

Sessionise a raw event stream. Input: a DataFrame with visitor_id, user_id (often NULL), occurred_at_utc and event_name, unsorted, up to 5 million rows. A session breaks when the gap from that visitor's previous event exceeds 30 minutes, and is force-closed at UTC midnight so no session spans two calendar dates. A gap of exactly 30 minutes does not break. Emit one row per session with session_id, visitor_id, the user_id as of the last event in the session, started_at_utc, ended_at_utc, session_date, duration_seconds and event_count. Vectorise; do not loop per visitor.

Approach

Sort by ['visitor_id', 'occurred_at_utc', 'event_id'] once, then express the whole problem as one boolean vector: a row starts a new session when the visitor changed, or the gap exceeds 30 minutes, or the UTC date differs from the previous row's UTC date. Cumsum that vector and you have the session key.
Get the comparison direction right on the gap: the rule is strictly greater than 1800 seconds, so an event at exactly 1800 seconds continues the session. Write it as gap > pd.Timedelta(minutes=30), and make the tie a test case rather than an assumption.
Derive the midnight break from the date change, not from inserting synthetic boundary rows. A date change implies a break even when the gap is two seconds, which is precisely the force-close rule and is why the two conditions are ORed rather than one subsuming the other.
Aggregate with a single groupby on the session key: min and max of occurred_at_utc, size for event_count, and last for user_id, which is correct because the frame is already sorted so 'last' is the final event in the session. That is the identity-as-of-session-end rule.
Compute duration_seconds as (max - min).dt.total_seconds(), which makes a single-event session 0 seconds. Say so explicitly, because a downstream mean session duration is sensitive to whether single-event sessions are 0 or excluded.

Worked solution (35 min)

df = df.sort_values(['visitor_id','occurred_at_utc','event_id']).reset_index(drop=True).
new_visitor = df.visitor_id.ne(df.visitor_id.shift()); gap = df.occurred_at_utc.diff(); new_day = df.occurred_at_utc.dt.date.ne(df.occurred_at_utc.dt.date.shift()).
is_start = new_visitor | (gap > Timedelta(minutes=30)) | new_day; df['session_key'] = is_start.cumsum().
g = df.groupby('session_key'); out = g.agg(visitor_id=('visitor_id','first'), user_id=('user_id','last'), started_at_utc=('occurred_at_utc','min'), ended_at_utc=('occurred_at_utc','max'), event_count=('event_id','size')).
out['session_date'] = out.started_at_utc.dt.date; out['duration_seconds'] = (out.ended_at_utc - out.started_at_utc).dt.total_seconds(); assign session_id from the sorted index.

Expected result: One row per session where session_count equals is_start.sum(), event_count sums exactly to len(input), every session's started_at and ended_at share one UTC date, and duration_seconds is at least 0 and strictly less than 86400.

Verify it yourself

out['event_count'].sum() equals len(df); a mismatch means rows were lost to a join or a dropna.
started_at_utc.dt.date equals ended_at_utc.dt.date on every row.
For a planted visitor with events at 12:00:00 and 12:30:00 exactly, one session is emitted; move the second to 12:30:01 and two are emitted.
For a planted visitor with events at 23:59:50 and 00:00:10, two sessions are emitted on consecutive dates despite a 20-second gap.

Follow-up

Sessions are used as the denominator of a conversion rate. How does moving the inactivity gap from 30 to 45 minutes move that rate, and in which direction?
A visitor's clock is 40 minutes ahead, so their events arrive with future occurred_at values. What does your sessioniser do, and what would you rather it did?
The same person signs up mid-session on mobile and continues on desktop. How many sessions and how many users does your output show, and is that the right answer?

How would you debug a piece of SQL code that is returning unexpected results?

Medium · sql and data manipulation

Approach

Compute rates by summing numerator and denominator separately, never by averaging rates.
Say which table is the grain you start from, and join outward from it.
Handle the rows that do not match: a LEFT JOIN with a NULL check is usually the question.

Follow-up

What breaks if events arrive late or out of order?
How would you verify this result without re-running the same query?

Walk me through a complex SQL query you wrote to solve a specific data problem.

Medium · sql and data manipulation

Approach

Check whether any join is one-to-many before aggregating, or the sums inflate.
State the window function and its partition and ordering out loud before writing it.
Say which table is the grain you start from, and join outward from it.

Follow-up

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

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

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.
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.
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.
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.
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.
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)

Count units per variant and units with more than one variant; report the contaminated count before anything else.
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.
LEFT JOIN the bounded outcome events and flag conversion with a boolean aggregate.
Aggregate per variant to exposed, converted, rate and share, and run the binomial check on the share, both before and after the truncation.
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?

How do you prioritize projects when you have multiple competing requests from different departments?

Medium · product sense and metrics

Approach

Decompose the metric into the rates that drive it, and say which one you would check first.
Fix the population and the time window before naming any metric.
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?

Given a hypothetical scenario involving traffic pattern analysis, what metrics would you track?

Medium · product sense and metrics

Approach

State what result would change your recommendation, so the answer is falsifiable.
Fix the population and the time window before naming any metric.
Decompose the metric into the rates that drive it, and say which one you would check first.

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 an experiment to test the effectiveness of a new feature?

Medium · a/b testing and experimentation

Approach

Say whether units interfere with each other, and switch design if they do.
State the primary metric and the minimum effect worth shipping, then size the test.
Decide the analysis before seeing data, including how long it runs and when you look.

Follow-up

What would you do if you could not randomise at all?
What would you conclude if the result is positive but the test is underpowered?

If we noticed a sudden drop in detection accuracy, how would you go about diagnosing the issue?

Medium

Approach

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

Follow-up

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

Diagnose a 50.5 to 49.5 split before reading the result

Easy · sample ratio mismatch · exposure logging · chi-square

An experiment randomised on visitor_id ran for 11 days. fct_experiment_exposure holds 104,320 control rows and 102,180 treatment rows for the experiment_id, all with is_in_analysis_population = TRUE, against a configured 50/50 split. The treatment arm shows a 3.1% relative lift in visit-to-signup conversion at p = 0.004, and the team wants to ship. Decide whether that lift can be reported, show the arithmetic behind your decision, and name the three checks you would run first on the exposure logging.

Approach

Test the split instead of eyeballing it. Chi-square goodness of fit against 103,250 expected per arm gives 2 x 1070^2 / 103,250 = 22.2 on 1 degree of freedom, p about 2.5e-6. The conventional alarm threshold is p < 0.001, so this is a sample ratio mismatch.
Treat the mismatch as invalidating, not as a caveat. The arms are no longer exchangeable, so the 3.1% lift has no causal reading and rebalancing on covariates afterwards does not restore it.
Walk the logging path in the order units flow through it: assignment service, then the exposure event fired on the treated surface, then the filters that build the analysis population. Each stage can drop one arm preferentially.
Name the three suspects this stack actually has. Exposure emitted by client code the treatment arm reaches later or not at all when its bundle errors; an extra redirect or render in one arm that loses slow connections; and a post-assignment filter (is_bot_flagged, is_internal, consent_state) that correlates with variant.
Check whether the imbalance is uniform by splitting counts on session_date, surface, app_version and device_type. A mismatch confined to one client build or to day one points at a deployment fault that can be fixed and re-run, not at a design fault.

Worked solution (15 min)

N = 206,500, expected 103,250 per arm, observed deviation 1,070, which is 0.518% of the total pushed into one arm.
Chi-square = 2 x 1070^2 / 103,250 = 22.18 on 1 df, p about 2.5e-6.
Compare with the p < 0.001 threshold and stop: the readout is void.
Re-pull the same counts grouped by session_date and app_version to localise where the rows were lost.
Deliver one line: the split is off by 1,070 visitors, a roughly one-in-400,000 event under correct randomisation, so the experiment is being fixed and re-run rather than read.

Expected result: Chi-square(1) = 22.2, p about 2.5e-6, far past the p < 0.001 mismatch threshold. The correct deliverable is an exposure-logging investigation and a re-run, not a caveated 3.1% lift.

Verify it yourself

Cross-check the statistic as a z on one arm count: under a fair split that count has SD sqrt(N/4) = sqrt(51,625) = 227 visitors, so the 1,070 deviation is 4.71 SD, and 4.71^2 = 22.2 reproduces the chi-square.
Chi-square scales linearly with N at a fixed ratio, so the same 50.518% split at N = 2,000 returns 22.18 / 103.25 = 0.21, p = 0.64. That reconciles the arithmetic and nothing more; it is not evidence that the alarm is independent of N. What rules out the large N is that the p-value is uniform under correct randomisation at every sample size, so p = 2.5e-6 is a one-in-400,000 draw whether N is 2,000 or 2,000,000. Large N buys the power to resolve a 0.5% skew; it does not manufacture one.
The mismatch p-value should be orders of magnitude smaller than the lift's p-value; if they are comparable, recheck the arithmetic.
Arm counts must sum to the total exposure rows for that experiment_id. If they do not, the analysis-population filter is already dropping rows before you looked.

Follow-up

The imbalance is entirely in one app_version on android. Can you salvage the other surfaces, and under what argument?
How would you monitor this across every running experiment without drowning in false alarms from a daily chi-square at alpha 0.05?
The missing rows turn out to be disproportionately visitors with consent_state = 'unknown'. What does that do to the estimate even after the logging bug is fixed?

Decide whether a one-day core-action drop is real

Easy · seasonality · day-of-week · late-arriving-data

A daily dashboard counts distinct fct_event.user_id with is_core_action = TRUE, filtered on occurred_at_utc, and is read at 09:00 UTC. This morning it shows yesterday down 22% against the day before. fct_event is partitioned on received_at_utc. You have fct_event, fct_session and dim_user with thirteen months of history. Deliver a one-paragraph verdict, escalate or do not escalate, with the evidence that settles it, before anyone proposes a product hypothesis.

Approach

Identify which two weekdays the comparison actually spans, then pull the same weekday-pair transition for the last 52 weeks and place the observed 22% inside that distribution. A day-over-day comparison in a product with a weekday pattern is a comparison of two different populations, so the reference class is the same transition historically, not the prior day.
Measure partition completeness rather than assuming it. For each of the last 30 days compute the share of that day's occurred_at_utc rows that had landed by 09:00 UTC the following morning, split by surface; mobile clients buffer events offline, so the freshest partition is systematically short and the shortfall is not uniform across surfaces.
Recompute the same series keyed on received_at_utc. If the drop survives on both keys it is not a lateness artefact; if it exists only on occurred_at_utc it is the partition filling in.
Check the two exclusion flags before segmenting anything: a change in is_bot_flagged coverage or a batch of is_internal accounts entering or leaving moves a distinct-user count with no user behaviour behind it.
Only if the movement survives all of the above, begin the segment decomposition. Say explicitly in the verdict which of these four checks the movement passed, so the next reader does not repeat them.

Follow-up

What publication lag would you set for this dashboard, and how would you derive the number rather than pick it?
If you switch the metric to received_at_utc, what does that break for anyone comparing to historical figures?
How would you detect the same problem automatically, so a human does not have to notice it each morning?

Suggested practice checklist

For a candidate whose interviews will centre on A/B testing, metric movement and causal claims. Design comes before arithmetic, arithmetic before analysis, and the week ends by rehearsing the readout rather than the derivation.

Day 1 — Design one test end to end on paper

Take a single feature change and write the full design: randomization unit, the exact point of exposure, the primary metric with its grain, guardrails, allocation, planned duration, and the decision rule committed before any data exists.
Write why the randomization unit must sit at or above the level where treatment can spill over, and give one case where user-level randomization is still contaminated (shared accounts or devices, or two participants in the same marketplace).
State in advance what you will do if the primary metric is flat while a secondary metric is significant.

Deliverable: A one-page test design with a decision rule written before launch.

Day 2 — Power arithmetic until it is automatic

Compute required sample size per arm for a binary metric with the normal approximation, n is approximately 2 times (z for alpha/2 plus z for power) squared times p(1 minus p) divided by delta squared, for baselines of 2, 10 and 40 percent at a 5 percent relative lift, and note that for a fixed relative lift the requirement falls as the baseline rises because delta grows proportionally with p.
Redo the calculation for a continuous metric using variance in place of p(1 minus p), and show why a heavy-tailed quantity such as revenue per user needs either far more traffic or a capped version with a stated cap.
Convert one of the results into weeks given a weekly eligible traffic figure, then list the two honest ways to shorten it (accept a larger detectable effect, or reduce variance) and write why quietly lowering the power target is a decision to miss more real wins, not a speedup.

Deliverable: A small script or sheet that maps baseline, minimum detectable effect, alpha and power to sample size and weeks, cross-checked against a published calculator.

Day 3 — Variance and the unit-of-analysis problem

Take a ratio metric whose denominator is not the randomization unit (clicks per session, randomized by user) and compute the standard error twice, once naively at session level and once by the delta method or a user-level bootstrap, then record how much the naive version understates it.
Implement CUPED on simulated data: choose a pre-period covariate X measured before assignment, estimate theta as Cov(Y, X) divided by Var(X), and analyse Y minus theta times (X minus its mean) in place of Y. Confirm the variance of the adjusted outcome equals the raw variance multiplied by one minus the squared correlation between Y and X, so a correlation of 0.45 removes about 20 percent of the variance and not 80.
Now run that simulation a few hundred times and confirm the adjusted effect estimate is unbiased for the same effect rather than numerically identical to the raw one. Within any single run the two differ, sometimes by a large fraction of the true effect, because the two arms' pre-period covariate means never coincide exactly in a finite sample; they agree in expectation, which is the property that matters and the one to state out loud.

Deliverable: A notebook showing the adjusted estimator with a measurably smaller variance than the raw one, plus a repeated-simulation table showing the two estimators agreeing on average while differing run by run.

Day 4 — Validity threats you can actually test for

Run a sample ratio mismatch check as a chi-square goodness-of-fit test against the intended allocation, and write the three causes you would chase first (assignment logged before exposure, an arm-specific redirect or load failure, bot filtering applied asymmetrically).
Simulate peeking: generate A/A data, test daily at alpha 0.05 across 14 looks, record the inflated false positive rate, then apply an alpha-spending boundary or commit to a fixed horizon and confirm the rate returns to nominal.
Write how you would separate a novelty effect from a durable lift using the treatment effect plotted against days since first exposure, and what shape would change your recommendation.

Deliverable: One table showing the peeking false positive rate before and after correction, plus a written SRM triage list.

Day 5 — When randomization is not available

Write the identifying assumption for difference-in-differences (parallel trends in the absence of treatment), then plot pre-period trends for two candidate control groups and justify rejecting one of them.
Design a switchback test for a change where user-level randomization would leak across participants, choosing a time-block length against the carryover you expect and saying how you would detect carryover in the data.
List what an interrupted time series or a synthetic control buys you and the one thing neither can rule out: an unobserved shock that coincides with the launch.

Deliverable: A one-page memo recommending a single quasi-experimental design and naming its weakest assumption explicitly.

Day 6 — The readout query

Write the assignment-to-exposure join that returns exactly one row per unit per experiment, and handle units appearing in both arms by excluding and counting them rather than silently keeping one.
Compute the per-arm metric, its variance and the relative lift with a confidence interval in SQL, then reproduce the identical numbers in a notebook as a cross-check.
Add a segment breakdown and write the sentence that keeps it from being p-hacking: segments declared in advance, everything else reported as exploratory and corrected for multiplicity.

Deliverable: A single query that outputs the full readout table, matched to a notebook recomputation.

Day 7 — Present it to someone who will not read the appendix

Give a 10-minute readout of a real or simulated experiment in the order decision, number, uncertainty, caveat.
Have your listener ask "can we ship it" in the case where the primary is flat and a guardrail moved, and answer with a recommendation rather than a request for more data.
Rewrite your opening line so the recommendation lands before any methodology.

Deliverable: A one-page readout whose first line is the recommendation.

Rigour against a deadline

Nearly every data role forces a trade between the analysis you want and the one that fits the decision window. Prepare a case where you deliberately shipped something less rigorous, named the weakness to the person relying on it, and said what would change your answer. The naming is the part interviewers listen for.

Tell me about a time you had to explain a complex technical concept to a non-technical stakeholder.

Medium · behavioural and stakeholder questions

Approach

Quantify the outcome, including what you would not claim credit for.
Close with what you would do differently, concretely.
State the situation in two sentences and spend the rest on your reasoning.

Follow-up

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

Describe a time you had to optimize a model or query for performance.

Medium · behavioural and stakeholder questions

Approach

Pick a story where you drove the decision, not one where you observed it.
Name the disagreement or constraint, and how you resolved it with evidence.
Quantify the outcome, including what you would not claim credit for.

Follow-up

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

Walk through an analysis you got wrong and what changed

Easy · postmortem · data quality · self-assessment

Describe an analysis of yours that turned out to be wrong after somebody had already acted on it. You have four minutes. The account must name the defect mechanically, the join, the filter, the window or the identity key, rather than describing it as a communication problem. It must also say who did what because of the wrong number, how the error surfaced, how long it stood, and what control you put in place so that class of error cannot reach a decision again. Do not pick an error nobody acted on.

Approach

Recognise what is being probed: whether you can be specific about your own failure without minimising it or performing contrition. The discriminator is whether the defect has a mechanism the listener could reproduce in their own warehouse.
Choose the case by blast radius rather than by comfort. An error nobody acted on tests nothing, and picking one signals that you are managing the interview instead of answering it.
Structure the account in six beats: the number, the decision it drove, the defect, the detection, the correction, the control. Keep the defect to one reproducible sentence, for example an inner join to fct_subscription_period that dropped accounts with no subscription row and so computed retention over payers only.
State the direction of the bias, not only its existence. A filter or join that removes rows usually moves a metric predictably, and knowing which way shows you diagnosed the mechanism rather than patched the symptom.
Be exact about detection and elapsed time. 'A colleague noticed' and 'the row-count assertion failed before publication' are different answers about the same organisation, and the second one is the one your control is supposed to produce next time.
End on the control, its cost, whether it has fired since, and one thing it does not cover.

Follow-up

What did the control cost, and has it fired since? If it never has, how do you know it works?
How long did the wrong number stand before anyone questioned it, and what does that say about the review path it went through?
What is the equivalent mistake you are most likely to make in this role, given the tables you would be working in?