“Can you pull the numbers
on why churn is up?” That sentence is the job. Not the SQL — the six steps between a request nobody can answer and a decision somebody makes. Our analysts do all six, full‑time in your team, on the Virtus payroll.
Six rungs, each one narrower than the last The bar beside each rung is its width — literally how much of the original question is still in play. An analyst who cannot narrow will hand you a dashboard and call it an answer.
The request 100%
“Why is churn up?”
Unanswerable as stated. Up against what, for whom, measured how? A weak analyst builds a dashboard. A good one asks three questions first.
The real question 82%
Did month‑2 retention fall for customers acquired after the March pricing change?
Now it has a cohort, a window and a suspected cause. It can be answered with a number and a confidence interval.
The grain 64%
One row per customer per month, active = at least one billable session
Defined before the query is written, and agreed with the person who asked. Most analytics disputes are really grain disputes.
The query 48%
Cohort retention over 6 months, split by acquisition channel, excluding trials
Written against the governed model, not a CSV export. Reproducible next quarter by someone else.
The number 34%
Month‑2 retention: 71% → 63% for paid‑search cohorts. Organic unchanged.
One finding, stated plainly, with the slice that did not move — because the contrast is what makes it credible.
The decision 24%
Reinstate the intro tier for paid search only; review in 60 days
If the analysis does not end here,



it was a report rather than an answer. This rung is the one that gets skipped.
Six things they are actually paid for SQL as the primary tool Window functions, CTEs and cohort logic — comfortable enough that the question shapes the query rather than the other way round.
Dashboards people actually open Few, owned, and built around a decision. We would rather delete four dashboards than add a fifth.
Experiment analysis Sample size before the test, not after. Sequential peeking called out. A result reported with its uncertainty.
Definitions and a glossary 'Active user' written down once, with an owner and a changelog, so two teams stop quoting different numbers at each other.
Python where it earns it pandas for the analysis that does not fit in SQL, and a notebook that someone else can rerun.
Telling people things they did not ask The most valuable output of a positive analyst is usually the finding nobody requested.
Seniority The bands map to the ladder A junior analyst starts at rung three and does it well. Paying senior rates buys rungs one, two and six — the framing and the decision.
Junior 1–3 yrs $24–30/hr
Answers well‑specified questions accurately and builds the dashboards. Needs the question framed for them.
Mid‑level 3–5 yrs $30–38/hr
Turns a vague ask into a defensible answer, and pushes back when the data cannot support the claim.
Senior 5+ yrs $38–52/hr
Owns a business area's measurement, designs experiments, and is trusted in the room where the decision is made.

#J-18808-Ljbffr

📌 Hire Data Analyst (Alberta)
🏢 VirtusInfotech
📍 Alberta