How This Product Is Actually Doing: A SQL Investigation
SQL analysis of 325,000+ product events, 90,000+ email events, and 9,381 users
One user segment uses the product six times more than another. Both open their email at almost the same rate.
A product team watching its user count climb every month can still be losing ground underneath, if engagement per user is falling at the same time. This project joins four logs, who signed up, what they did, what emails they opened, and a small separate operational feed, to check both halves of that question together.
Growth was unambiguous: 160 new active users in the first month of the data, up to 1,031 in the last, with no sign of slowing. Weekly engagement told a different story. It climbed for 12 straight weeks to a clear peak, then declined 27% over the following four. The original weekly aggregation had used a date function that resets ambiguously at year boundaries, so the trend was recalculated with weeks numbered from the first event instead, which is what surfaced the decline cleanly.
The most useful finding came from splitting users into the three unlabelled segments the data provides. One segment engages with the product roughly six times more than another, 117.5 events per user against 19.1. If the difference were about interest or attention, you'd expect it to show up in email behaviour too, opens, clicks, engagement with marketing. It doesn't: open rates across all three segments sit within one percentage point of each other. Whatever separates a heavy user from a light one here, it isn't whether they read their email, which points any investigation toward onboarding or the product itself.
Elsewhere in the data: engagement is a weekday habit (Friday sees over four times Sunday's volume), 59.7% of it happens on a laptop or desktop rather than mobile, and the email program already outperforms typical benchmarks on both open rate and click-through.
Why it matters
A rising user count can hide a shrinking one underneath it, if the people already on the platform are engaging with it less over time. This project pulls four separate logs together to check growth and engagement at once rather than trusting the headline number, and finds a real decline the growth figure was hiding, plus a usage gap between user segments that email behaviour alone can't explain.
Engagement events per week, numbered from the first event. Peaks at week 12 (highlighted), then falls 27% by week 16. Week 17 is a partial week.
Left axis: product engagement per user. Right axis: email open rate. Usage varies sixfold by segment; email behaviour barely moves.
Dataset, tools and how it was done+
Dataset: Product event logs: users, events, email_events, and a small job_data feed (325,255 engagement events, 90,389 email events, 9,381 users)
Tools: MySQL · Window Functions · Date/Time Functions
- 325,255 engagement events, 90,389 email events, and 9,381 users, 2013-2014, joined on user_id
- Rebuilt the weekly engagement trend by numbering weeks from the first event rather than using MySQL's WEEK(), which resets ambiguously at year boundaries and had made the original trend misleading
- Device column (26 distinct raw values) rolled up into three categories to make the split actionable
- Email funnel corrected from a per-user GROUP BY, which produces one noisy rate per person, to a single aggregate across all sends
- Segment comparison: product engagement intensity vs. email open rate, split by the three unlabelled user_type values in the source data
- job_data (a separate, 8-row feed) checked for duplicates and language mix only, since eight rows can't support a trend