Interview guides · Calfus

Calfus Data Scientist Interview Questions & Guide 2026

Causal inference drills for Calfus data scientist prep: diff-in-diff, matching, instruments, and the limits of observed data.

The 60-second brief

As a Data Scientist at Calfus, you serve as a critical bridge between raw data and actionable business intelligence. This role is pivotal to the organization’s ability to solve complex, high-stakes problems for clients, requiring you to translate ambiguous business requirements into robust mathematical models and scalable data solutions. You will work within cross-functional teams, collaborating closely with engineers and product managers to ensure that your analytical insights directly influence product strategy and operational efficiency.

SQL is seldom the hardest round and is often the one that eliminates people. The working bar is usually window functions, correct deduplication, and joins that do not silently fan out rows, rather than obscure syntax.

Focus areas

Turn a vague request into a measurable question
Separate novelty effects from durable behaviour change
Decompose a metric move by segment and mix

Overview

The work at Calfus is characterized by its diversity and intellectual rigor. You will be expected to handle end-to-end project lifecycles—from data extraction and feature engineering to model deployment and performance monitoring. Because Calfus operates in a fast-paced environment, your ability to communicate technical findings to non-technical stakeholders is just one of the many ways you will drive tangible impact. You are not just building models; you are crafting the narrative that guides business decisions.

Be prepared to discuss the business impact of your past projects. At Calfus, technical accuracy is only half the battle; the other half is demonstrating how your work improves the bottom line or enhances user experience.

Where to focus your preparation

Treat the sections below as preparation areas and confirm the format with your recruiter.

Preparation focus

No round sequence has been reported for this company, so work the categories below and confirm the format with your recruiter.

What to demonstrate

Breadth across SQL, experimentation and product reasoning
Ability to state assumptions before choosing a method

How to prepare

Drill the practice exercises below and time yourself
Prepare three quantified stories about decisions you drove

Where candidates lose points

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.

Counting on an identity key that changes underneath the metric

visitor_id is per browser and per device, and it resets on cookie clearance, private browsing and platform privacy changes, so the distinct-visitor count drifts upward for reasons unrelated to reach. Any rate with visitors in the denominator therefore decays over time even when behaviour is constant, and any rate with visitors in the numerator inflates. The stitching at signup makes it worse in both directions: a user who signed up on mobile and returns on desktop is two visitors and one user, while a shared device is one visitor and several users. Decide which key each metric is counted on, write it into the definition, and when comparing a period before and after a platform privacy change, expect a level shift in every visitor-keyed metric and do not attribute it to the product.

Over-explaining the method and under-explaining the implication

Lead with the answer and what you would do about it, then give the approach when asked. Roughly one sentence of method per three of implication is the right ratio for a stakeholder-facing answer; the interviewer already knows what a regression is.

Comparing periods without accounting for seasonality or day-of-week

Compare whole weeks against whole weeks and check whether the same swing appeared in prior cycles or prior years before attributing it to anything you changed. Weekday and weekend populations often differ enough that a Tuesday-to-Saturday comparison is meaningless.

Questions & practice

Explain the trade-offs between precision and recall in the context of your last model.

Medium · machine learning and modelling

Approach

Check what information would not exist at prediction time, and exclude it.
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?
Where could label leakage enter this setup?

Walk me through the most challenging ML project you have worked on.

Medium · machine learning and modelling

Approach

Pick an evaluation metric that matches the cost of each error type, not a default.
Check what information would not exist at prediction time, and exclude it.
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?
Where could label leakage enter this setup?

How did you handle data leakage in your previous project?

Medium · machine learning and modelling

Approach

Say how the offline result would be validated online before it is trusted.
Pick an evaluation metric that matches the cost of each error type, not a default.
Check what information would not exist at prediction time, and exclude it.

Follow-up

How would you choose the decision threshold, and who owns that choice?
Where could label leakage enter this setup?

How would you optimize a search algorithm for a large, unsorted dataset?

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.
Pick an evaluation metric that matches the cost of each error type, not a default.

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?

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

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.
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.
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.
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.
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.

Worked solution (30 min)

rng = np.random.default_rng(seed); for memory, batch the replications in chunks and accumulate the any-cross count across chunks.
Per chunk: draw (chunk, 40000) uniforms per arm, x = (u < 0.10).cumsum(axis=1), slice columns at indices 3999, 7999, ..., 39999.
Compute the ten z statistics vectorised over the chunk, take crossed = (np.abs(z) > 1.96).any(axis=1).
Aggregate: peek_rate = total_crossed / reps; single_rate = mean of |z_final| > 1.96; mc_se = sqrt(p*(1-p)/reps) for each.

Expected result: The single-look rate lands at 0.05 within about 0.005 at 10,000 replications. The ten-look rate lands near 0.19, which with 10,000 replications has a Monte Carlo standard error of about 0.004, so anything in roughly 0.18 to 0.20 is consistent and anything near 0.40 indicates independent redraws.

Verify it yourself

The single-look control recovers 0.05 within two Monte Carlo standard errors; if it does not, the z statistic is wrong and the headline number is meaningless.
The crossing rate is monotone in the number of looks: 1 look near 0.05, 2 near 0.08, 5 near 0.14, 10 near 0.19.
Halving the base rate to 0.05 leaves the peeking rate essentially unchanged, since the result is about the test's random walk and not about the conversion level.
Re-running with the same seed reproduces the figure exactly.

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?

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?

Weekly visit-to-signup conversion split by acquisition channel

Easy · funnel · conversion · aggregation

From fct_session (session_id, visitor_id, started_at_utc, referrer_channel, is_bot_flagged, consent_state) and fct_event (visitor_id, occurred_at_utc, event_name), compute visit-to-signup conversion for one ISO week, split by channel. session_id is the unique key of fct_session. Denominator: distinct visitor_id with a session starting in the week, is_bot_flagged = FALSE and consent_state <> 'denied'. Numerator: those visitors with a 'signup_completed' event in the same week. Label each visitor with the referrer_channel of their first session in the window. Return channel, visitors, signups and rate, plus one all-channel total row.

Approach

Build a visitor spine that is one row per visitor: filter sessions to the week, drop is_bot_flagged and consent_state = 'denied', then take the first session per visitor with ROW_NUMBER() OVER (PARTITION BY visitor_id ORDER BY started_at_utc, session_id) = 1 to carry the channel label. Collapsing to one row here is what makes the channel buckets mutually exclusive and the totals additive.
session_id is the unique key, so that ordering is total and the label is reproducible. If the table carried no unique key you would have to write an explicit tie rule instead, because two sessions on different channels at the identical timestamp would otherwise label the visitor differently between runs.
Attach the outcome as a semi-join (EXISTS on a signup_completed event for that visitor inside the same week) rather than a join to the event table, so a visitor who fires the event twice does not count twice and inflate the numerator past the denominator.
Aggregate with COUNT() as visitors and COUNT() FILTER (WHERE signed_up) as signups, and compute the rate as signups::numeric / NULLIF(visitors, 0) so an empty channel returns NULL rather than a division error.
Produce the total with GROUP BY GROUPING SETS ((channel), ()), which re-sums numerator and denominator for the total row. Averaging the channel rates gives a different and wrong number whenever channel volumes differ, which they always do.
Verify the spine before trusting the output: COUNT(*) must equal COUNT(DISTINCT visitor_id), and the per-channel visitor counts must sum to the total row.

Worked solution (20 min)

Write the filtered session CTE and check its row count against an unfiltered count, so you know how much volume the bot and consent filters removed.
Add the ROW_NUMBER first-session pick and assert one row per visitor.
Add the EXISTS outcome flag and aggregate with FILTER.
Add GROUPING SETS for the total and format the rate to four decimal places.
Spot-check one channel by hand: pull its visitor list, count signups directly, compare.

Expected result: One row per referrer_channel plus a total row, with visitors, signups and a decimal rate. Channels with zero signups appear with rate 0, not missing. The total rate lies strictly between the minimum and maximum channel rate unless one channel holds all the volume.

Verify it yourself

Per-channel visitor counts sum exactly to the total row's visitor count.
COUNT(*) equals COUNT(DISTINCT visitor_id) on the spine CTE.
Replacing the EXISTS with a plain inner join changes the signup count only if duplicate signup_completed events exist; if it does change, say by how much.
Every rate is between 0 and 1.
Re-running the first-session pick with session_id removed from the ORDER BY returns the same labels; if it does not, visitors hold same-timestamp sessions on different channels and the tie rule is doing real work.

Follow-up

The denominator is distinct visitors. If a browser release shortens cookie lifetime, what happens to this rate, and how would you tell that apart from a genuine drop?
A visitor's first session is direct and their signup session is paid search. Your label says direct. When is that the wrong answer for the decision being made?
How do you roll four weeks into a month, and why is averaging the four weekly rates wrong?

A completion rate the owning team can move without fixing anything

Medium · gaming · instrumentation · funnel

A team's target is core-flow completion rate: distinct fct_event.flow_instance_id with a 'flow_completed' event within 30 minutes of its 'flow_started' and no 'error_shown' carrying the same flow_instance_id in between, over distinct flow_instance_id with a 'flow_started' in the window, split by surface and app_version. The same team owns the client that emits those events and the tracking plan that defines them. List the ways this rate rises without any user completing more flows, then redefine the metric and its guardrails so those routes are closed. Deliverable: the hardened definition.

Approach

Work the emission side first, because that is what the team controls: delay minting flow_started until after the first screen so the highest-dropping attempts leave the denominator; stop emitting or rename error_shown; mint a fresh flow_instance_id on each retry so one failed attempt becomes several attempts whose last one completes; move flow_completed earlier in the flow.
Sort those moves by where they are visible. None of them shows in the rate itself; three of them show only in volume, which is why the denominator has to be published on the same chart as the rate.
Re-anchor the numerator on something outside the flow's own instrumentation: require a downstream is_core_action = TRUE event for the same user_id within 30 minutes of flow_completed, so a completion only counts when it produced the thing the flow exists to produce.
Add the guardrail that catches the retry route specifically: mean and p90 flow_instance_id per user per day, with the rule written down that a rising completion rate alongside rising attempts per user is a regression and not a win.
Make definition changes visible instead of forbidden: stamp a tracking-plan version on the series and re-base the history when event semantics or the 30-minute window change, rather than splicing two definitions into one line.

Follow-up

How do you distinguish a genuine instrumentation fix from gaming, given both look like a step change confined to one release?
The jump appears in exactly one app_version. Does that exonerate the team or implicate it?
What do you do with eighteen months of history once the definition is hardened?

Measure whether self-serve help actually answered the question

Medium · proxy metrics · bias · support

A help widget opens in-product. Leadership asks for the percentage of users who got their answer. You have fct_event (event_name, user_id, session_id, occurred_at_utc, is_core_action, properties JSONB carrying article_id), fct_session, and a support ticket table joined on user_id. Nothing records whether the answer was correct or whether the user was satisfied, and no such field is being added. Propose the metric you would publish, state the proxy plainly, name the direction and rough size of its bias, and say which decisions it can and cannot support. Deliverable: the definition plus the bias statement.

Approach

Say first that the target quantity is unobserved and will stay unobserved: nothing in the stream distinguishes a user who was helped from one who gave up, and both leave the same trace, so every candidate metric here is a proxy and the only question is which bias you prefer.
Build the proxy to remove the largest identifiable error: among widget-open sessions, the share with no ticket from that user within 72 hours and at least one is_core_action = TRUE event after the widget opened in the same session. The downstream action requirement strips out most of the silent abandonment that a no-ticket rule alone scores as success.
State the residual bias with a direction and a bound: it still over-counts, because a user who gave up and went elsewhere files no ticket and may still complete an unrelated core action later in the session; bound it using the observable population of widget-open sessions that end within two minutes with no further event.
Publish a directly-measured companion with its own weakness in the same sentence: article thumbs up/down reported with its response rate, and the note that a single-digit response rate missing non-randomly toward annoyed users is exactly why it cannot be the headline.
Write the use statement, because a proxy without one gets reused for the wrong decision: valid for ranking articles against each other and for detecting a week-on-week break, invalid as an absolute deflection rate or as an input to a cost-saved figure.

Worked solution (25 min)

Write the unobservable target in one sentence with the observable set beside it, so the gap is visible on the page rather than assumed away.
Define the published proxy with exact numerator and denominator, including the session scoping of the downstream core action.
Bound the bias from data you already have: compute the share of widget-open sessions ending within two minutes with no further event.
Publish the thumbs signal beside it with its response rate and the direction of its missingness.
Write the two-line use statement covering what the number supports and what it does not.

Expected result: A named proxy with numerator and denominator, a stated bias direction (over-counts success) with an observed magnitude bound, and an explicit list of the decisions the number does and does not support.

Verify it yourself

The proxy and the raw no-ticket rate should differ by roughly the abandonment share computed in step three. If they are nearly identical, the downstream-action requirement is not binding and adds nothing.
A change in support hours or ticket-form availability moves the no-ticket rate with no change in help quality, so the series must carry annotations for support operational changes.
Sessions with consent_state = 'denied' cannot be joined forward to a user_id and therefore cannot enter this metric at all; the denominator definition has to say so.

Follow-up

How would you validate this proxy once, and what would you do if the validation said it over-counts by 20 points?
Support asks for a dollars-saved number from this metric. What do you say, and what would you need before saying anything else?
Ticket volume drops the week support hours change. How do you keep that out of the series?

A conversion rate that fell in one regulatory region

Medium · consent · identity-keys · denominator · geography

Visit-to-signup conversion fell 1.3 points over six weeks. Signups cut by dim_user.country_code put the fall in one regulatory region where a consent banner shipped in week one, but absolute signups from that region are flat. fct_session carries consent_state, visitor_id, is_bot_flagged and session_date and no country column, so the denominator cannot be cut the same way. Using fct_session and fct_event, decide whether behaviour changed or the denominator did, state what these tables cannot settle, and name the one column that would settle it.

Approach

Name the asymmetry before computing anything. The numerator is user-keyed and therefore cuttable by country; the denominator is visitor-keyed and is not. Dividing a region-filtered numerator by an unfiltered denominator produces a quantity that is not a rate, and presenting it as a regional conversion rate is the first mistake available here.
Attack the denominator on the dimension you do have. Compute distinct visitor_id per week and sessions per distinct visitor_id per week: a consent banner that blocks or shortens the identity cookie raises the distinct-visitor count and lowers sessions per visitor, which depresses any visitor-keyed rate with no behaviour behind it.
Split on consent_state. Sessions with consent_state = 'denied' can enter the denominator but can never be joined forward to a signup, so a rising denied share mechanically drives the pooled rate down by roughly its own share. Report the granted-only rate and the denied share as two separate numbers rather than one blended figure.
Cross-check with measures that do not depend on the visitor key at all: absolute weekly signups, which are given as flat, and signups per session rather than per visitor.
State the limit honestly. Without country on the session or on its entry event, the regional attribution rests on the numerator alone, and the correct request is that one column, not a more elaborate model on top of the data you have.

Follow-up

If granted-only conversion is the metric going forward, what selection bias have you accepted, and in which direction does it point?
How would you handle the six weeks of already-published history once the new definition is adopted?
What is the smallest instrumentation change that restores a cuttable denominator without collecting more personal data than before?

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.

Ownership when the result is inconvenient

Sometimes the honest read is that the initiative did not work, and the person who commissioned the analysis was hoping otherwise. Interviewers want to know whether you softened it. Prepare the case where you delivered an unwelcome result, how you presented the uncertainty without hiding behind it, and what the team did next.

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?

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

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'.
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.
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.
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.
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?

Quantify your own impact without claiming the topline you touched

Hard · self-assessment · attribution · communication

You are writing the impact section of your own review. Over the year you ran four experiments, one of which shipped and three of which were flat; you corrected the definition of gross monthly revenue churn so that cancellation is recognised at period_end_utc; and you built a self-serve funnel dashboard. Weekly active accounts rose 14% over the same period. Your reviewer knows the data well. Write the three impact claims you would defend, stating for each what you contributed, what evidence supports it, and what portion of the outcome you are not claiming.

Approach

Recognise what is being probed: whether you apply to your own work the causal standard you would apply to somebody else's roadmap claim. Nearly everyone who would reject 'accounts that do Y retain better' will write 'I drove a 14% increase' without noticing it is the same error with a friendlier subject.
Sort the work by the kind of evidence it can carry. The shipped experiment is the only item with a randomised estimate, so it is the only one where an effect size is defensible, and you claim the interval rather than the point estimate.
Claim the three flat experiments as decisions prevented and price them. Features not built, or built differently, on evidence, with the engineering weeks reallocated as the number somebody else can verify. A defensible null is a delivered decision and should be written as one.
Claim the definition fix as correctness, not as improvement. The old figure was overstated by a specific percentage and appeared in a specific set of recurring documents; the impact is the change it produced in the forecast built on top of it, not a change in churn itself.
Claim the dashboard on usage and displacement: distinct weekly users of it, and the ad-hoc request count for six months before against six months after. If the request log does not exist, record the claim as unverified rather than estimating it upward.
Disclaim the 14% explicitly and once. State that it cannot be separated from seasonality, other teams' launches and a pricing change, and bound your own contribution from above using the shipped experiment's interval converted into headline units.

Follow-up

Your shipped experiment's interval was +0.2pp to +1.4pp on activation. How much of the 14% can that account for, and how do you say so without undercutting yourself?
A peer in the same cycle claims the full 14%. What, if anything, do you do about it?
If you could only keep two of your three claims, which do you drop, and why that one?