Crunch in action
Real questions. Real investigations.
Every clip is one question asked in plain English and the investigation Crunch runs to answer it — no edits, no narration. Including the moments it refuses to answer a follow-up, because the data isn't there.
Start here 5
Investigations on your product data
Begin with a question every team has asked — then watch the reveals get sharper: a hidden crossing, a drop with a name, a segment nobody designed for.
The founder's conversation
“Is our growth real — and where is it leaking?”
A three-turn investigation — orient, plan, then pinpoint the funnel leak — the way a founder interrogates the business before a board meeting.
The leak under the flat line
“Our activity looks flat — which step is the leak hiding under that headline?”
The dominant funnel drop-off gets named: listing-view → contact converts at just 18.5% end to end — the leak the flat number was hiding.
The drop with a name
“Our funnel took a hit — where are people falling out, and is contact worse on mobile than desktop?”
The drop gets a name: mobile converts at 20% vs desktop's 60%. A device, not just a number.
The north star, decomposed
“The big number's up — real broad engagement, or a few heavy users carrying it?”
Not broad engagement — p90 (19) is 4.75× the median (4). A thin power core carries the metric.
The channel that looks wrong
“One channel looks weak on conversion — cut it, or is the short window lying?”
The low-looking channel is actually high-value — ranked by true conversion quality (referral 36% · paid 18% · organic 9%), not a short-window rate. Then it declines the 12-month LTV ask: no billing data.
Then your website 4
Investigations on your website
Point the same reasoning at a marketing site: where the funnel really leaks, and which source is actually worth the spend.
The leak is the form
“The landing page isn't converting — are we losing people at the button or the form?”
The leak is the form, not the button — 20k views → 14k clicks → 2.8k submits.
The source worth the spend
“Which source actually sends people who sign up — and is it the one sending the most visitors?”
Volume and quality name different sources: paid sends 16k visits to organic's 8k, but organic converts at 30% against paid's 7.5%. The pooled 15% rate hides the split entirely.
The loss after the read
“Of the people who read our pages all the way down, how many get in touch — and where do we lose the most?”
The loss is after the read, not before it: 4,750 of 16,000 deep readers submit the form (29.7%), against only 4,000 lost on the way to reading deeply. And it refuses to call the zero contact rate real — that's a tracking gap, not disinterest.
The ranking it wouldn't give
“Rank our pages by read-through rate — which ones actually hold people, and which get views then bounce?”
No ranking. The share computes to an impossible 222% — 20,000 scroll events against a 9,000-viewer denominator — so Crunch names the broken denominator and the fix instead of shipping the number.
Ask your own question next
Point Crunch at your data and watch it investigate — live demo or free trial.
See a demo Start free trial