Two requests sit at the top of the queue. Huda: “Our paid search budget is up for review. Which traffic sources actually turn visits into orders, and where do people drop off?” Rania: “We acquired a lot of customers in Black Friday months. Do they come back as often as everyone else?” The first is a funnel, the second a cohort retention analysis. Both are built from things you already know: conditional aggregation, dates and CTEs.
\nEach row of web_sessions is one visit, with flags for the steps it reached: added_to_cart and purchased (0 or 1). Summing a 0/1 flag counts the sessions that reached the step, so a funnel is one GROUP BY:
SELECT\n source,\n COUNT(*) AS sessions,\n SUM(added_to_cart) AS carts,\n SUM(purchased) AS purchases,\n ROUND(100.0 * SUM(added_to_cart) / COUNT(*), 1) AS cart_rate,\n ROUND(100.0 * SUM(purchased) / NULLIF(SUM(added_to_cart), 0), 1) AS checkout_rate,\n ROUND(100.0 * SUM(purchased) / COUNT(*), 1) AS conversion\nFROM web_sessions\nWHERE started_at >= '2025-01-01'\nGROUP BY source\nORDER BY conversion DESC;Email converts best overall, 10.9% of sessions, and referral worst at 7.1%. The step rates tell the more useful story. Referral visitors rarely add to cart (13.3%), but those who do check out more often than anyone (53.6%). Email is the reverse. Those are different problems with different fixes: referral needs better landing pages, email needs a smoother checkout.
\nThat’s why a funnel reports step rates, each step divided by the previous one, not only end-to-end conversion. Make each denominator explicit in the column name; “conversion rate” without a denominator is the most argued-about number in any marketing meeting. NULLIF guards the division for a source with no carts.
Swap source for device and a sharper finding appears: desktop sessions that reach the cart check out 56% of the time, mobile sessions only 37%. Mobile brings the most traffic, so that gap is worth more than any change to the ad budget.
A cohort is a group of customers who share a starting period. Here, the month of their first non-cancelled order. Retention then asks: of the customers who started in month M, what share ordered again 1, 2, 3 months later?
\nThree steps, three CTEs. First, give every order a month index, year * 12 + month, so that month differences are simple subtraction even across a year boundary. Second, find each customer’s first month. Third, list each customer’s active months relative to their cohort:
WITH orders_ok AS (\n SELECT customer_id, order_date,\n CAST(strftime('%Y', order_date) AS INTEGER) * 12\n + CAST(strftime('%m', order_date) AS INTEGER) AS month_index\n FROM orders\n WHERE status <> 'cancelled'\n),\nfirsts AS (\n -- grain: one row per customer\n SELECT customer_id,\n MIN(month_index) AS cohort_index,\n strftime('%Y-%m', MIN(order_date)) AS cohort\n FROM orders_ok\n GROUP BY customer_id\n),\nactivity AS (\n -- grain: one row per customer per active month\n SELECT DISTINCT f.cohort, f.cohort_index, o.customer_id,\n o.month_index - f.cohort_index AS month_number\n FROM orders_ok AS o\n JOIN firsts AS f ON f.customer_id = o.customer_id\n)\nSELECT\n cohort,\n COUNT(*) FILTER (WHERE month_number = 0) AS customers,\n ROUND(100.0 * COUNT(*) FILTER (WHERE month_number = 1)\n / COUNT(*) FILTER (WHERE month_number = 0), 1) AS m1_pct,\n CASE WHEN MAX(cohort_index) + 2 <= 2025 * 12 + 12 THEN\n ROUND(100.0 * COUNT(*) FILTER (WHERE month_number = 2)\n / COUNT(*) FILTER (WHERE month_number = 0), 1)\n END AS m2_pct\nFROM activity\nWHERE cohort >= '2025-07'\nGROUP BY cohort\nORDER BY cohort;The final SELECT pivots month numbers into columns with FILTER. The DISTINCT in activity is essential: retention counts people, and a customer who orders three times in a month is one active customer.
Look at the December 2025 row in the query’s output: m1_pct is 0.0. That isn’t churn. Month 1 for December’s cohort is January 2026, and the data ends on 31 December. The m2_pct column shows the fix: a CASE expression checks whether the cohort has lived through that month (cohort_index + 2 is still within December 2025) and returns NULL otherwise. A retention chart with honest blanks is a triangle; one with zeros shows a fake collapse in every recent cohort.
Back to Rania’s question. Change the filter to 2024 and the November 2024 cohort, Cartwheel’s first Black Friday crowd and the largest of that year at 35 customers, retains only 14.3% in month 1, below most 2024 cohorts. Deep-discount customers tend to come for the deal. That’s an answer she can use when planning the next promotion. The exercise asks a related question: what share of each cohort comes back at all within three months?
\nThis closes the analytical toolkit. The final section is about trust: making sure the data and your queries deserve it.
\n","toc":[{"id":"sec-1","text":"A funnel from session events","level":2},{"id":"sec-2","text":"Cohorts: grouping customers by when they started","level":2},{"id":"sec-3","text":"Unknown is not zero","level":2}],"takeaways":["A funnel counts how many units reach each step and divides each step by the one before it, so you can see exactly where people drop off.","Always state the denominator of a rate; checkout rate (purchases per cart) and conversion rate (purchases per session) answer different questions.","A cohort groups customers by the period of their first event, and retention counts distinct customers active N periods later divided by the cohort size.","Months a recent cohort hasn’t lived through yet are unknown, not zero; show them as NULL so nobody reads them as churn."],"further":[{"title":"SQLite aggregate functions and FILTER","url":"https://www.sqlite.org/lang_aggfunc.html"},{"title":"SQLite date and time functions","url":"https://www.sqlite.org/lang_datefunc.html"}],"quiz":[{"q":"Email sessions: 404 sessions, 107 add to cart, 44 purchase. Which number is the checkout rate?
","options":[{"text":"10.9%, purchases divided by sessions","why":"That’s the overall conversion rate. It mixes two steps and hides where people drop off."},{"text":"26.5%, carts divided by sessions","why":"That’s the cart rate, the first step of the funnel."},{"text":"73.5%, sessions that didn’t add to cart","why":"That’s the drop-off at the first step, not a checkout measure."},{"text":"41.1%, purchases divided by carts","why":"Correct. Checkout rate measures the step from cart to purchase, so its denominator is carts."}],"answer":3},{"q":"Why does the cohort query use SELECT DISTINCT cohort, customer_id, month_number before counting?
The December 2025 cohort shows 0% retention in month 1. The data ends on 2025-12-31. What should the table show?
","options":[{"text":"0%, because none of them ordered in month 1","why":"Month 1 for the December cohort is January 2026, which isn’t in the data. Zero would claim they all churned."},{"text":"The same value as the November cohort, as an estimate.","why":"Filling in another cohort’s value invents data and hides that the period hasn’t happened."},{"text":"NULL or blank, because that month hasn’t been observed yet.","why":"Correct. Unobserved cells are unknown. A CASE expression that checks whether the cohort month plus N is within the data range makes this explicit."}],"answer":2},{"q":"To compute “months since first order” across a year boundary, why convert dates to year * 12 + month instead of subtracting strftime('%m', ...) values?
Rania wants one number per cohort: of the customers whose first order was in a given month, what percentage ordered again in any of the next three months (month numbers 1, 2 or 3)?
\nReturn cohort (‘YYYY-MM’), customers (cohort size) and repeat_3m_pct (rounded to 1 decimal) for the cohorts from 2025-01 to 2025-09, which have all three months observed. Sort by cohort. Order matters.
The CTEs are correct. The final SELECT counts the wrong thing.
\n","hints":["What is the grain ofactivity? One row per order, so a customer with five orders appears five times.","Both the cohort size and the repeat count are about people. Count DISTINCT customer_id, not rows.","COUNT(DISTINCT customer_id) FILTER (WHERE month_number BETWEEN 1 AND 3) counts each returning customer once, however many times they ordered."],"explanation":"activity has one row per order, so COUNT(*) counted orders: the cohort size included every order its customers ever placed, and a customer who ordered in months 1, 2 and 3 counted three times in the numerator. Counting distinct customers in both places puts numerator and denominator at the same grain, people. Repeat rates range from about 31% (February) to 67% (June); with cohorts of 16 to 29 customers, a difference of one or two people moves the rate by several points, so report the cohort size alongside it.