Seven analyses, one question: which part of the business moved?
Prism is not another dashboard. Each view takes a single number apart until the cause is a specific segment, product, cohort or channel you can act on.
A revenue change, split into its four causes
Net revenue = sessions × conversion rate × average order value, less returns. Change any one input and the total moves by a knowable amount. The waterfall shows each contribution in currency, not percentages.
Sessions +6% adds $41k. Conversion −21% removes $151k. AOV +2% adds $11k. Returns +1.4pp removes $9k. Net: −$108k.
Contributions are sequenced, so they sum exactly to the total change — no unexplained residual line.
Any period comparison: week on week, month on month, campaign against baseline.
It surfaces the segment instead of asking you to guess
Prism scans device, channel, customer type, first product and discount status combinations, suppresses statistically thin ones, then ranks what remains by contribution to the total change.
- returning · mobile · paid social18% of revenue−34%
- new · mobile · paid social9% of revenue−11%
- returning · desktop · email21% of revenue+7%
- new · desktop · organic14% of revenue−2%
| cohort | size | M0 | M1 | M2 | M3 | M4 | M5 |
|---|---|---|---|---|---|---|---|
| Jan 2026 | 1,240 | 100% | 46% | 38% | 33% | 30% | 28% |
| Feb 2026 | 1,396 | 100% | 44% | 35% | 30% | 27% | 25% |
| Mar 2026 | 1,502 | 100% | 51% | 44% | 40% | 37% | — |
| Apr 2026 | 1,318 | 100% | 39% | 29% | 24% | — | — |
| May 2026 | 1,611 | 100% | 53% | 46% | — | — | — |
| Jun 2026 | 1,744 | 100% | 55% | — | — | — | — |
Retention by acquisition month, channel, first product, discount status
Cohorts answer whether a good month bought lasting customers. Split by discount status and the answer is usually uncomfortable — and useful.
Revenue net of COGS, shipping, fees and ad spend
Per product and per channel, so you can stop scaling the channel that sells the most and start scaling the one that keeps the most.
- COGS38%
- Ad spend27%
- Shipping9%
- Fees2.6%
- Margin23.4%
- Revenue100
- COGS−38
- Shipping−9
- Payment fees−2.6
- Ad spend−27
- Contribution margin23.4
Cumulative margin curves by cohort, with a payback month
LTV without a payback period is a story. With one, it is a budget: it tells you how much acquisition your cash can carry.
Sliders on conversion, AOV and retention — compounded
Retention lifts repeat orders, which changes the base the AOV lift applies to. Prism compounds the interaction instead of adding three independent lifts.
Effects compound: retention raises repeat orders, which raises the order count the AOV change applies to. Prism models the interaction rather than adding the three lifts.
What the promotion added versus what it shifted
Incremental orders, orders pulled forward from later weeks, and orders that would have paid full price — separated, then priced in margin.
- Incremental orders31%
- Pulled forward44%
- Would have paid full price25%
The promo looked like a $358k week. $111k of it was new demand; the rest was margin given away on orders you were getting anyway.
What each analysis needs from you
Merchants who sign up without these churn and blame the product. Read this before you buy.
| Analysis | Data minimum | You must supply | Caveat |
|---|---|---|---|
| Waterfall | 2 comparable periods | Store connection | Very small stores show noisy conversion |
| Segment discovery | ~500 orders / month | Store connection | Thin segments are suppressed, not shown |
| Cohort retention | 4+ months history | Store connection | Guest checkout weakens customer identity |
| Contribution margin | Current cost data | COGS + shipping costs | Historical costs need historical COGS |
| LTV & payback | 6+ months, ad spend | COGS + ad accounts | Attribution model changes the answer |
| Scenarios | Baseline period | Nothing extra | A model, not a forecast |
| Discount analysis | Promo + non-promo weeks | Discount codes mapped | Needs a clean control period |
If your store is below the volume thresholds, tell us on the contact form. We'll say so rather than take the subscription.
See these seven views on your own numbers.
No trial account needed — start monthly and cancel any time.