What a frozen baseline is, and how to read growth against it
A frozen baseline is four weeks of the store's own numbers, saved and locked before anything changes. Every result after is read against it, per day, from the same source. Convey Growth freezes one in the first week of every install, so a founder can check the growth without taking anyone's word for it.
How it's built.
Four weeks of the old store, from the store's own order record: orders a day, revenue a day, new customers a day.
Locked before we touch anything. Nothing in it gets edited after. If a number turns out wrong, the correction is written next to it, dated.
Per day, not per period, so a four-week baseline and an eleven-week stretch after launch compare fairly.
One per channel. A brand that sells on its website, on Amazon and in stores gets a baseline for each, before a dollar of ads, so a dollar on Meta is read everywhere it lands.
What it showed at a coffee roaster.
Read against four weeks frozen before launch, from the store's own record.
| What | Value | Window |
|---|---|---|
| orders a day since launch | +98% | Jul 28 to Oct 12, 2026 |
| revenue a day since launch | +58% | Jul 28 to Oct 12, 2026 |
| new customers a day, small base | 2.3 to 7.4 | Jul 28 to Oct 12, 2026 |
Observed results from the store's own record, not isolated causal lift.
What it can't tell you.
A baseline shows what changed, not why. A season, a press mention or a price change can land in the same window as the work. That's why every result we publish carries the same line: observed results from the store's own record, not isolated causal lift.
When one cause matters, like whether a new ad works, it gets its own test with its own read. The baseline answers the bigger question: is the store doing more than it did before.
Freeze your own.
Before the next big change, export the last four full weeks of orders from Shopify. Save orders a day, revenue a day and new customers a day somewhere nobody edits, and write the date on it. Every claim about what worked after that date gets checked against it.
Questions.
Why four weeks?
It covers four of every weekday, so one slow Tuesday doesn't set the bar. Much longer windows start to describe a different store.
What if those four weeks were unusual?
Then that's written next to the baseline before the work starts, and every read says so. A baseline only settles arguments if everyone agreed to it before the new numbers came in.
Does it work for a brand on Amazon or in stores?
Yes. Each channel gets its own baseline, frozen the same week, from that channel's own record.
Send us your URL.
Get a free written read in 48 hours. No login, no cost, no call.
Sent. The read lands in 48 hours.