Interview guides · Cognite

Cognite Data Scientist Interview Questions & Guide 2026

Cognite Data Scientist interview guide: reported rounds, practice questions with worked solutions and a 7-day prep plan.

The 60-second brief

As a Data Scientist at Cognite, you will operate at the intersection of advanced artificial intelligence, machine learning, and heavy industrial reality. Cognite is a global leader in industrial digitalization, and the data science team is tasked with solving some of the world's most complex and high-impact problems. Your work will directly contribute to redefining how heavy asset industries—such as Oil and Gas, Power and Utilities, and Manufacturing—operate, with the ultimate moonshot of unlocking $100B in customer value by 2035.

How much statistics you need depends on the flavour of the seat. Experiment-facing work wants you deep enough to notice that repeated looks at accumulating data inflate the false positive rate of a fixed-sample test; modelling-facing work wants estimation and honest uncertainty intervals.

Focus areas

Strip CI, retry and synthetic traffic first
Analyse at the account grain, cluster errors
Read NRR on a fixed account cohort

Overview

You will be embedded within the AMER delivery team inside the Global Delivery organization. This is a highly cross-functional environment where you will collaborate daily with Data Engineers, Solution Architects, Project Managers, and client-side Subject Matter Experts (SMEs). Rather than building models in isolation, your role centers on configuring, deploying, and operationalizing digital solutions that integrate seamlessly into the client's industrial landscape, often leveraging the core capabilities of Cognite Data Fusion (CDF).

This position is exceptionally dynamic because it combines technical depth with client-facing consulting. You will lead discovery workshops, translate physical and industrial challenges into data science problems, develop machine learning and physics-based models, and build intuitive front-end interfaces to make your insights actionable for field operators and executives alike.

The Cognite interview process

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

Technical Screening

Much of what gets scored here happens out loud while you type. Nobody can see your reasoning inside a half-written query, so five silent minutes read as being stuck even when they are not. State the plan in plain language first: which tables, what grain you are aggregating to, and the one filter that defines the population. Then write it. The narration doubles as insurance, because a wrong plan gets caught early and cheaply while a wrong query gets caught at the end with no time left to redo it. A timed statistics section, where one exists, is a separate test with its own clock.

What to demonstrate

Whether the query you write matches the plan you just described
What you do with a hint, meaning whether the correction gets absorbed or the first approach gets defended
Whether you can debug your own wrong output by reading the result set and naming which part of the query produced the anomaly

How to prepare

Solve three problems while screen-sharing into a recording, then watch it back and mark every stretch longer than thirty seconds where you said nothing
Practise compressing the plan into one sentence before typing, then check afterwards whether the finished query actually matched it
Time yourself on statistics questions that carry a business reading, such as what a confidence interval does and does not claim, rather than re-reading notes without a clock

HR Introductory Screen

An extra round usually exists because something is still open after the standard loop: a skill the earlier interviews did not sample, a level decision, or two interviewers who disagreed. It is rarely a rerun of what you already did well. Ask the recruiter who you are meeting, what function they sit in, and how long the session runs. That is an ordinary scheduling question, and the answer changes what you should prepare. What separates a strong candidate here is treating the round as a fresh evaluation with its own bar, rather than assuming earlier performance carries you through or sinks you.

What to demonstrate

Whether you can answer well on ground the earlier rounds did not cover, without leaning on what you already said to someone else
Consistency of the facts in your stories: the same sample size, timeframe, team size and scope of your own role as in earlier conversations
How you handle an unfamiliar format live, including whether you ask what kind of answer is wanted before producing one

How to prepare

Ask the recruiter for the interviewer's function, the length, and whether to expect a coding surface, a discussion, or a presentation. Preparing for a 30 minute conversation with a partner team is not the same work as preparing for a 60 minute technical block.
Write out what each earlier round actually covered, then list the two or three areas nobody probed. That gap is the most likely subject of the extra round.
Re-read the numbers in the project stories you have already told, so a second telling does not quietly contradict the first.

Technical and Leadership Rounds

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

What to demonstrate

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

How to prepare

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

Live Coding Exercise

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

Final Executive Interview

A day of back-to-back interviews samples your floor, not your ceiling. Four hours in, the habits that carry a good answer are the first to go: restating the question before solving it, asking what the data would have to look like, checking a number before quoting it. What the day decides is whether the tired version of you is still someone to leave alone with an ambiguous problem. The round that sinks a candidate is usually not the hardest one. It is the one immediately after the round that went badly.

What to demonstrate

Whether the late rounds get the same clarifying questions as the first one, or whether you start answering immediately to save effort
Whether a weak answer stays in the room it happened in, instead of following you into the next conversation as apology or distraction
Whether the quality of your questions holds up, since fatigue removes curiosity about the problem before it removes knowledge of the method

How to prepare

Rehearse the length, not just the content: book four mock interviews of different types in one afternoon with short gaps, because the one you need to observe is the fourth
Put the two or three questions you ask at the start of any problem on a card in front of you, so that under fatigue it is a habit you run rather than a decision you make
Decide in advance what the gap between rooms is for: water, one line of notes on anything you promised to follow up, and an explicit close on the round that just ended so it does not travel
Prepare a different closing question for each interviewer, so the end of a long day does not produce the same one four times

Where candidates lose points

Randomising an experiment at the user level when users share an account

Two problems fire at once. Colleagues in one workspace see each other's work and talk to each other, so a treated user changes the behaviour of a control user in the same account, which violates the no-interference assumption and biases the estimate toward zero. Separately, outcomes within an account are strongly correlated, so the effective sample size is roughly n / (1 + (m - 1) * rho) for m users per account and intra-class correlation rho, not n. With rho around 0.3 and twenty users per account that is a design effect near 6.7, meaning a user-level confidence interval is about two and a half times narrower than it should be and results cross significance thresholds on noise alone. Randomise the account and cluster the standard errors.

Reading consumption metrics before the metering lag window has closed

Usage pipelines land late and correct themselves, which is exactly what is_restated and restated_at record. A dashboard queried on day T sees a partially populated tail for the last several days, so the most recent points always slope downward and always look like a regression. Analysts then explain the artefact, and sometimes ship a change to fix it. Establish the empirical settling time by measuring how much a given usage_date's total moves between first_written_at and its final value, exclude that many trailing days from every reportable figure, and never compare a fresh period against a settled one.

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.

Never asking what decision the analysis will inform

Open with who makes the decision, what the options are, and by when. The answer determines the precision you need, the segments worth cutting, and whether an observational read suffices or an experiment is required.

Questions & practice

What optimization techniques would you use to schedule power grid distribution while minimizing transmission losses?

Medium · statistics and probability

Approach

Say what the estimate is of, and over what population it generalises.
Quantify uncertainty explicitly rather than reporting a point estimate alone.
Write down the assumption the method needs before you use the method.

Follow-up

Which assumption here is most likely to be violated in practice?
How would you explain this result to someone who does not know statistics?

How would you validate the accuracy of a machine learning model when the ground truth relies on subjective assessments from field operators?

Medium · machine learning and modelling

Approach

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

Follow-up

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

Explain how you would structure a Python package to ensure your machine learning pipeline is modular and deployable.

Medium · machine learning and modelling

Approach

Say how the offline result would be validated online before it is trusted.
Frame the prediction: the label, the moment of prediction, and the action it triggers.
Set a baseline first, so any model has something honest to beat.

Follow-up

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

Permutation-test a consumption experiment randomised at account level

Hard · permutation-test · heavy-tails · cuped · experiment-readout

An experiment randomised 900 accounts into two arms. You have one row per account: account_id, arm, consumption_28d (billable units after launch) and consumption_pre (the 28 days before). Consumption is heavy-tailed and the largest account is several percent of the total. Write a permutation test from scratch: winsorise at the pooled 99th percentile as a pre-registered rule, use the difference in arm means of the winsorised outcome as the statistic, and obtain a two-sided p-value from 20,000 relabellings of the account-level arm vector. Report the observed effect, the p-value, and the same test on a CUPED-adjusted outcome.

Approach

Be precise about what the permutation test needs. Under the sharp null of no effect for any account, the outcomes are exchangeable across arm labels, and the test is valid for ANY statistic T(outcomes, labels) provided the identical function is applied to the observed labels and to all 20,000 relabellings. The pooled 99th percentile is a function of the outcome vector alone, so recomputing it inside the loop returns the same number 20,000 times: that is wasted CPU, not a bias, and hoisting it out is an optimisation rather than a correctness fix. Say plainly that capping at all changes the estimand from mean consumption to mean capped consumption; it is not a neutral cleaning step.
The mistake that does invalidate the test is an asymmetry between the observed statistic and the permuted ones, and the easiest way to create it is to derive the cleaning rule from the observed arm labels and then freeze it — winsorise each arm at its own observed 99th percentile, hold those two caps fixed, and permute. The observed value is then computed with caps matched to its own partition while every relabelling is scored with caps belonging to a different one, so the null distribution no longer answers the question the p-value claims to answer. A per-arm cap recomputed consistently inside every permutation is a valid test, but it estimates a contrast whose two sides are capped at different thresholds, so prefer the pooled cap on estimand grounds and pre-register it.
Permute the account-level arm vector, because the account is the randomisation unit. Relabelling anything finer — users, workspaces, requests — generates a null distribution narrower than the design actually supports and returns p-values that are anti-conservative.
Vectorise the null: tile the treatment indicator into a (B, n) matrix and permute along axis 1 with rng.permuted(..., out=...). The statistic is a difference of means, so the treated sum alone determines it and the whole null is one matrix-vector product. Use the two-sided p-value (1 + count(|stat_perm| >= |stat_obs|)) / (B + 1); the plus-one on each side is not cosmetic, it keeps the p-value away from exactly zero and keeps the test valid at finite B.
For CUPED, fit theta = cov(y, x) / var(x) on the pooled data and use that same theta for the observed statistic and every relabelling. Pooled theta, like the pooled cap, carries no label information, so where in the loop you compute it is again only a performance question; fitting theta within arms is what goes wrong, because the adjusted outcome then depends on the labels and an observed-label fit frozen across all 20,000 relabellings breaks the match between observed and permuted statistics. x must be measured entirely before launch, which consumption_pre is. Expected variance reduction is about 1 - corr(y, x)^2; measure the achieved reduction from the two null distributions rather than asserting it.

Worked solution (45 min)

cap_y = np.quantile(df.consumption_28d, 0.99); y = np.minimum(df.consumption_28d.to_numpy(float), cap_y); cap_x = np.quantile(df.consumption_pre, 0.99); x = np.minimum(df.consumption_pre.to_numpy(float), cap_x)
t = (df.arm == 'treatment').to_numpy(); n1 = int(t.sum()); n0 = len(t) - n1; obs = y[t].mean() - y[~t].mean()
rng = np.random.default_rng(11); L = np.tile(t.astype(np.int8), (20_000, 1)); rng.permuted(L, axis=1, out=L); s1 = L @ y; stats = s1/n1 - (y.sum() - s1)/n0
p = (1 + int(np.sum(np.abs(stats) >= abs(obs)))) / (20_000 + 1)
theta = np.cov(y, x, ddof=1)[0,1] / np.var(x, ddof=1); y_adj = y - theta*(x - x.mean()); repeat steps 2 to 4 on y_adj and compare stats.std(ddof=1) between the two runs.

Expected result: A two-sided p-value strictly between 1/20001 and 1, an observed effect expressed in capped billable units per account, and a CUPED null whose standard deviation is smaller than the unadjusted one by roughly sqrt(1 - corr(y, x)^2). The CUPED point estimate stays close to the unadjusted one, since the adjustment removes variance rather than shifting the effect.

Verify it yourself

The permutation null is centred near zero: |mean(stats)| is a small fraction of std(stats).
An A/A run, relabelling arm at random before the test and repeating over seeds, gives roughly uniform p-values with about 5% below 0.05.
std(stats_cuped) < std(stats) whenever corr(y, x) is materially non-zero, and the ratio matches sqrt(1 - r^2) to within simulation noise.
Moving the pooled cap computation inside the permutation loop reproduces the same p-value bit for bit at the same seed — the direct demonstration that a pooled cap carries no label information — while setting the cap to infinity reproduces the uncapped test exactly, so the winsorisation switch is independently verifiable.

Follow-up

The p-value is 0.04 with the cap and 0.31 without it. What do you report, and what did you pre-register?
Colleagues in a shared workspace can see the treated behaviour. How does that change the design and the estimate?
How many accounts would you need to detect a 5% lift given this outcome's distribution?

Describe how you would apply Generative AI and Large Language Models (LLMs) to help operators query complex industrial equipment manuals.

Medium · sql and data manipulation

Approach

Say which table is the grain you start from, and join outward from it.
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.

Follow-up

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

Given a SQL database containing industrial asset tags, write a query to join asset metadata with time-series data and filter out inactive sensors.

Medium · sql and data manipulation

Approach

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

Follow-up

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

Write a Python function to perform a rolling-window calculation on a noisy sensor dataset to identify anomalous temperature spikes.

Medium · sql and data manipulation

Approach

Check whether any join is one-to-many before aggregating, or the sums inflate.
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.

Follow-up

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

Running commitment burn-down and the date consumption crosses it

Medium · window functions · running totals · committed consumption

fct_subscription_period gives committed_amount_cents, term_start_date and term_end_date for each account's current version where pricing_model = 'committed_consumption'. fct_usage_daily gives account_id, workspace_id, sku_code, usage_date and net_amount_cents, at one row per workspace and SKU per day. Inside each account's current term, return the running total of net_amount_cents by usage_date, the first usage_date on which that running total reaches committed_amount_cents, and the fraction of the term elapsed at that point. Accounts that have not reached their commitment must still appear, with a null crossing date.

Approach

Collapse usage to one row per (account_id, usage_date) first. The fact is grained by workspace and SKU, so a raw running total leaves several rows per date and the first-crossing date becomes dependent on the arbitrary order of rows inside that day.
Restrict to the term with usage_date BETWEEN term_start_date AND term_end_date on the account's current row, so no prior term's consumption leaks into this term's burn-down.
Compute SUM(net_cents) OVER (PARTITION BY account_id ORDER BY usage_date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW). Name ROWS explicitly: the default frame is RANGE, which includes all peer rows at the same ORDER BY value and would hand back the whole day's total on each of that day's rows.
Extract the crossing with MIN(usage_date) FILTER (WHERE running_cents >= committed_amount_cents) grouped per account, which returns NULL for accounts still under commitment instead of dropping them.
Elapsed fraction is (crossing_date - term_start_date)::numeric / NULLIF(term_end_date - term_start_date, 0); Postgres date subtraction yields whole days, so the guard matters for same-day terms.
Exclude the trailing days still inside the metering settling window, measured from first_written_at against restated_at, and say how many days you cut and why.

Worked solution (30 min)

Build terms: SELECT account_id, committed_amount_cents, term_start_date, term_end_date FROM fct_subscription_period WHERE is_current AND pricing_model = 'committed_consumption' AND committed_amount_cents IS NOT NULL.
Build daily: join fct_usage_daily to terms on account_id with usage_date inside the term, then GROUP BY account_id, usage_date summing net_amount_cents.
Add running_cents with SUM(...) OVER (PARTITION BY account_id ORDER BY usage_date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW).
Aggregate per account with MIN(usage_date) FILTER (WHERE running_cents >= committed_amount_cents) AS crossing_date, then compute the elapsed fraction with the NULLIF guard.
LEFT JOIN that summary back onto terms so every committed account appears, and verify the last running_cents per account equals a plain SUM over the term.

Expected result: A per-account-per-day series of running_cents inside the current term, plus one summary row per committed account carrying crossing_date and term_elapsed_fraction, with crossing_date NULL exactly for accounts whose final running total is below committed_amount_cents.

Verify it yourself

The maximum running_cents per account equals SUM(net_amount_cents) over that account's term computed independently.
crossing_date is NULL if and only if the final running total is below the commitment.
term_elapsed_fraction falls in [0, 1] for every non-null crossing date.
running_cents is non-decreasing within each account, which fails immediately if a negative credit row was included without comment.

Follow-up

Turn this into an end-of-term overage forecast. What breaks if you extrapolate a linear run rate on a consumption product?
An account crosses its commitment at 40 percent of the term. Is that an expansion signal or a billing-surprise risk, and what would you check to tell them apart?

Walk us through how you would lead a discovery workshop with a client's Subject Matter Experts (SMEs) to define the data requirements for a wind turbine optimization project.

Medium · product sense and metrics

Approach

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

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?

Pick the randomisation unit when workspaces and queues are shared

Medium · interference · randomisation-unit · clustering · switchback

A collaboration feature was tested by randomising 18,000 users 50/50. dim_user_membership shows a median of 9 activated, non-service users per paying account, and weekly successful interactive requests per user carries an intra-class correlation of 0.25 within an account. The reported lift is 4% at p = 0.05. Separately, the same team wants to evaluate a change to the admission-control queue, which is shared by every account in a region. Give the correct randomisation unit for each test, the effective sample size and corrected p-value for the first, and the design for the second.

Approach

Separate the two failures of user-level randomisation, because they need different fixes and only one of them is a variance problem. Interference: a treated colleague changes a control colleague's behaviour inside a shared workspace, which biases the estimate toward zero. Correlated outcomes: users inside an account are not independent draws, which understates the variance. Clustering the standard errors fixes the second and leaves the first entirely intact.
Compute the design effect: 1 + (m - 1) * rho = 1 + 8 * 0.25 = 3.0. Effective sample is 18,000 / 3 = 6,000 users, corresponding to roughly 2,000 accounts. Standard errors are understated by sqrt(3) = 1.73, so a reported t of 1.96 is really 1.13 and the honest two-sided p is about 0.26, not 0.05.
Re-run the first test randomised on account_id with inference clustered on account_id, or equivalently as a regression on account-level means. The account-means version is the conservative one and is usually easier to defend to a non-specialist audience, at the cost of some efficiency when cluster sizes are very unequal.
For the queue change, account randomisation fails too, and for a different reason: treated and control accounts contend for the same finite capacity, so the control arm is mechanically affected by the treated arm and the contrast estimates a property of a mixed system rather than the effect of the policy. Randomise time instead, with a switchback alternating the admission policy at the region-by-30-minute-slot level.
Specify the switchback so it is actually valid. Impose a washout at the start of each slot at least as long as the p99 queue drain time and discard requests enqueued before the switch; randomise slot order rather than strictly alternating, because a fixed alternation aliases with the hourly and weekday traffic cycle; block on hour-of-day so both policies see peak and trough; and cluster inference on the slot, which is the unit that was randomised.
State the power consequence plainly: the cluster count is slots, not requests. Fourteen days of 30-minute slots gives 672 slots, and the variance to plan against is between-slot variance in the outcome, which is far larger than between-request variance and is what makes switchbacks expensive.

Worked solution (30 min)

Compute the design effect 1 + (9 - 1) * 0.25 = 3.0 and the effective sample 18,000 / 3 = 6,000 users, about 2,000 accounts.
Inflate the standard error by sqrt(3) = 1.732, convert the reported t of 1.96 to 1.13, and read off a two-sided p of about 0.26.
Recompute the same contrast on account-level means as an independent check that does not depend on the pilot's rho estimate.
Lay out the switchback: region-by-30-minute slots, randomised order, washout equal to the p99 drain time, blocking on hour-of-day, inference clustered on slot, 672 slots over fourteen days.

Expected result: Design effect 3.0, effective sample about 6,000 users or 2,000 accounts, corrected two-sided p about 0.26. The 4% lift is not significant once clustering is respected and is additionally biased toward zero by within-workspace spillover. The queue change is run as a randomised-order switchback with slot-level clustering and a drain-time washout, not as an account-level split.

Verify it yourself

The design effect depends only on cluster size and intra-class correlation: it must collapse to 1 when rho = 0 and when m = 1. If your expression does not, it is not the design effect.
An account-means regression and a user-level regression with cluster-robust standard errors should give similar standard errors. A large gap indicates very unequal cluster sizes, in which case the means version needs weighting.
The switchback's sample size is slots, not requests. If your power calculation cites a sample in the millions, you have counted requests and will badly overstate precision.

Follow-up

The intra-class correlation was estimated at 0.25 from a 300-account pilot. If it is really 0.40 the design effect becomes 4.2. How do you plan under that uncertainty rather than betting on the point estimate?
Under what conditions is user-level randomisation still the right choice even though users share accounts?
The queue change is expected to leave a four-hour retry backlog. What does carryover of that length do to a 30-minute switchback, and what would you run instead?

Activation drops six points starting at a deploy hour

Medium · instrumentation · release analysis · metric inputs

Seven-day activation, defined as an account's first request with http_status < 400, api_key_id not null and traffic_class <> 'synthetic_monitor' within 168 hours of created_at, fell six points for sign-up cohorts after a Tuesday. A client SDK major version shipped that morning. From fct_api_request (account_id, api_key_id, sdk_name, sdk_version, http_status, traffic_class, request_at) and dim_account (account_id, created_at, is_internal), decide whether activation actually fell or the metric's inputs changed, and state in advance what evidence would convince you of each.

Approach

Decompose the definition and recompute activation under each relaxation: status only, status plus traffic_class, then the full definition. If the entire drop lives in the api_key_id clause, this is an instrumentation question rather than a behavioural one.
Measure the null rate of api_key_id by sdk_version and by hour. A stamping change appears as a step at the deploy boundary confined to the new version; a behavioural change appears as a ramp that grows with adoption and leaves old-version traffic untouched.
Hold the cohort's SDK mix fixed before comparing. New sign-ups adopt the newest version first, so a cohort-level drop can be pure composition even when no individual version changed at all.
Corroborate with a source the release did not touch: whether the same cohorts appear in fct_usage_daily with non-zero billable_quantity, and whether their fct_support_ticket rows with category = 'onboarding' rose.
Write the decision rule down before looking at the answer. An instrumentation artefact predicts unchanged downstream usage and a version-confined null step; a real regression predicts falling downstream usage and more onboarding tickets in the same cohorts.

Follow-up

Old-version and new-version populations are not exchangeable, because new accounts adopt the new version first. How would you build a comparison that is not confounded by cohort age?
What backfill or metric-versioning policy keeps the historical series interpretable once you fix the stamping?

Suggested practice checklist

Four days spend equal time on query work, statistics, modelling and product judgement at deliberately shallow depth, which produces a scored map of where you actually stand. The last three days spend everything on the two areas the role weights most, and close by re-running day one to measure movement.

Day 1 — Breadth pass: query fluency

Solve six prompts spanning aggregation, joins, window functions and date arithmetic in 60 minutes total, stopping at 10 minutes each whether or not it works, and mark every prompt as solved, solved slowly, or stuck.
For each unsolved prompt write the single blocking sentence (I lost the grain, I did not know the frame clause, I could not express the date boundary) instead of reading the solution.
Translate one pandas transformation you know well into SQL and one SQL query into pandas, checking that both return the same row count and the same totals.

Deliverable: A scored six-row table, one line per prompt, saved for the day-seven re-run.

Day 2 — Breadth pass: statistics and inference

Answer ten short questions in writing with nothing open: what a p-value is conditional on, what a 95 percent interval covers across repeated samples, when a paired test is the right one, what the bootstrap estimates, why multiple comparisons inflate false positives, how controlling the family-wise error rate differs from controlling the false discovery rate, what power depends on, what a missed real effect costs a product, the three situations where the central limit theorem does not rescue you (small n, very heavy tails, dependent observations), and what a standard error is the standard deviation of.
Grade yourself against a reference and count only the answers that were exactly right, not the ones that were nearly right.
Rewrite the two weakest answers the following morning from memory in full sentences.

Deliverable: Ten graded answers with an honest count of exact hits.

Day 3 — Breadth pass: modelling

Take one tabular dataset end to end in 90 minutes: a leakage-safe split, a baseline that is not a model (majority class or historical mean), one regularized linear model, one gradient-boosted tree, and a single evaluation metric chosen before you look at any result.
Write why that metric fits the cost structure: precision at a fixed recall for alerting, calibration for anything feeding a price or a threshold, ranking metrics for retrieval, and note that area under the ROC curve is insensitive to class balance in a way that can flatter a rare-positive problem.
Name the leak you were most likely to introduce (an encoding fit on all rows before splitting, or a feature computed after the label's timestamp) and write the check that would have caught it.

Deliverable: A notebook whose first cell states the metric and the baseline, plus two lines on what beat what and by how much.

Day 4 — Breadth pass: product judgement

Answer three case prompts aloud at 15 minutes each, timing how long passes before you state a success metric.
For one case write the first segmentation you would run and the row counts you expect per segment, so that a tiny segment cannot quietly drive the conclusion.
Take a metric definition you did not write, from a public dashboard, a textbook, or documentation you already have open, and list every place two analysts implementing it would diverge: which rows the denominator admits, whether the unit is an account or a person, what the time window is anchored to, and what happens to data that arrives late. Then write the one question that would close the largest of those gaps.

Deliverable: Three recorded case answers plus an ambiguity list for a metric someone else defined, ending in the single question you would ask about it.

Day 5 — Depth, first area

Rank the four areas by how many bullet points in the role description each one covers, pick the top one, and spend the entire day inside it.
Work the six hardest problems you can find in that area and for each write the generalizable move you should have reached for first, rather than the answer.
Re-solve the two you failed the same evening with notes closed.

Deliverable: Six generalizable moves written as instructions to yourself, not as solutions.

Day 6 — Depth, second area, and the seam between them

Repeat the depth protocol on the second-ranked area with the same six-problem structure.
Construct one problem that requires both areas at once, for example a metric redefinition whose effect you must validate with a test whose readout you then have to query.
Solve your own combined problem end to end and note where the handoff between the two areas cost you time.

Deliverable: One combined problem, solved end to end, with the handoff failure written down.

Day 7 — Integration and re-measurement

Re-run the six prompts from day one under the same clock and compare both correctness and time.
Run a 60-minute mixed mock that moves between areas without warning, since switching cost is what breadth passes do not train.
Write the two areas you would still fail on, and the sentence you will use in the interview when you hit one of them.

Deliverable: A before-and-after score table plus a written plan for the two remaining gaps.

Being wrong in public

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

How do you handle highly correlated features (multicollinearity) when modeling sensor arrays in a manufacturing plant?

Medium · behavioural and stakeholder questions

Approach

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

Report an underpowered consumption test to a non-technical executive

Medium · communication · uncertainty · experimentation

An account-randomised packaging change ran six weeks across 900 paying accounts. The effect on billable units per account per month is plus 4.1 percent, with a 95 percent interval from minus 3.2 to plus 11.8 after clustering standard errors at the account and applying the pre-registered winsorisation at the 99th percentile. An executive with no statistical background wants one number this week to decide a full rollout. Produce a three-sentence spoken answer, one chart, and an explicit recommendation of ship, stop or keep running, with the cost of each option stated.

Approach

The interviewer is probing whether you can be decision-useful without either hiding the uncertainty or hiding behind it. Start from the decision rather than the statistics: establish what the executive would do differently at plus 4 percent versus zero, because if the action is identical the interval does not matter.
Translate the interval into consequences in units the executive already reasons about. Multiply both endpoints by the cohort's baseline consumption and contracted rates to give an annualised revenue range, so the answer is a range of dollars rather than a range of percentages.
Price the option to wait. Using the observed variance, state roughly how many additional account-weeks halve the interval width, so keep running becomes a quantified choice instead of a stall.
Offer a cheaper path to the same decision: a lower-variance proximate outcome such as successful billable units on the new SKU, or CUPED using each account's pre-period consumption, quoting the expected variance reduction as one minus the squared pre-post correlation.
Give a recommendation and name the single observation that would reverse it. A strong answer commits; a generic one recites the interval and leaves the decision on the table.

Follow-up

The executive says it clearly works and is just not provable, so ship it. What is your answer?
How much of the interval width comes from clustering and how much from the revenue tail, and what would you do about each?
If you had to ship this week with no more data, which guardrail would you watch for the first fortnight and at what threshold would you roll back?

Disagree with a product manager about an adoption claim

Medium · conflict · telemetry hygiene · adoption metrics

A product manager is about to present that a new SDK release drove a 40 percent rise in requests among adopting accounts, computed from fct_api_request counts grouped by sdk_version. You find the rise is concentrated in traffic_class equal to ci, that rows with is_retry true grew alongside it, and that restricting to interactive non-retry traffic leaves a 3 percent lift. The launch review is in two days. Decide how you raise this, with whom and in what order, and what you propose the claim becomes.

Approach

The interviewer is probing whether you can correct a colleague without ambushing them, and whether your own counter-analysis carries the caveats theirs lacked. Go to the product manager privately before the review. A correction delivered in the room is a status move and loses the argument you are actually trying to win.
Bring a decomposition rather than a verdict: the same accounts and window, requests split by traffic_class with retries held out as their own column, so their 40 percent and your 3 percent reconcile line by line and neither has to be taken on trust.
Reproduce their figure exactly first. If you cannot land on 40 percent with their method, you do not yet know what you are disagreeing with.
Ask whether the continuous-integration lift is itself valuable. An account wiring the SDK into its pipeline has increased integration depth, which is the dominant switching cost in this domain, so the honest claim may be that integration depth rose while interactive usage moved 3 percent. Improving the claim beats deleting it.
Name the mechanism that makes the raw count dangerous: clients retry when the platform degrades, so retry volume climbs exactly when the customer is most at risk. Pull the 5xx rate for the same accounts and window before anyone concludes anything, and note that billable_units is zero on 5xx rows, so request counts and billable quantities diverging is itself the signal.
Close with a standing definition for launch metrics so the next release does not repeat the exercise.

Follow-up

The product manager argues that continuous-integration traffic is real usage and declines to split it out. Is that position defensible, and under what metric definition?
Suppose the 5xx rate for those same accounts also rose 40 percent. What is the claim now?
The review happens and the raw number is presented regardless. What do you do next, and what do you not do?