Keep a Change Log: The Small Habit That Makes CRO Compound

A change log is a running record with one entry per change: the date, the page, what changed, the evidence that prompted it, what you expected to happen, the read window you committed to, and - later - what the read said and whether the change stayed. It takes five minutes per entry and it is the difference between a business that learns and a business that merely tinkers. Six months of honest entries will tell you more about your customers than any tool you can buy, because it is the only record written entirely about your store.

Keep the entry short enough that you will actually write it. Date and page. Change, in one sentence concrete enough that a stranger could find it on the page. Reason: the baseline figure, recording or complaint that indicted the old version. Expectation: 'checkout completion up from about 55 toward 65 percent' beats 'should improve conversions'. Read rule: the window and count from the guide on judging changes. Verdict, added later: kept, reverted, or kept-and-watched, with the numbers. That is all. A log with these fields, kept, beats a elaborate system abandoned by March.

The entries that teach most are the changes that did nothing or hurt. Write them as fully as the wins: what you believed, what happened, what you now think. A log that only records victories becomes marketing material for your own past decisions and stops being evidence. There is also a practical payoff: the log stops you re-running a failed idea in nine months when it sounds fresh again, and it stops you reverting a change during a noisy week because you can see the read rule you committed to while calm.

Anywhere durable and boring: a document, a spreadsheet, a notebook by the packing bench. It must be the same place every time, and it must be writable in the moment the change ships - an entry reconstructed from memory a month later is fiction with dates. In a one-person business the log is also your handover document: the day someone else helps with the site, the log is how they learn what has been tried. Treat its honesty as you would treat your accounts: it is a record other decisions will rely on.

Quarterly, read the whole log in one sitting. Patterns surface that no single entry shows: changes to copy outperforming changes to layout three times running; every mobile fix helping, every homepage fix neutral; a hypothesis about price sensitivity confirmed twice. Those patterns are your store's private research findings, and they should steer the next quarter's priorities more than any article - including the ones on this site. This is the compounding the title promises: each entry is small, but the pattern across entries is knowledge a competitor cannot copy.

Open the log with three entries: the baseline itself (numbers, segments, date), the state of the funnel's step rates, and the fix list as currently prioritised. Then work normally. Do not backfill history from memory beyond a line noting that earlier changes went unlogged - the log's authority comes from being written at the time. Within one quarter you will have a record most larger competitors, with their agency handovers and staff churn, genuinely do not possess.

Owners sometimes delay starting a log until they have chosen the right system - a testing platform, a project tool, a shared doc with the right template. The choice does not matter and the delay does. A plain text file with dated entries outperforms an unconfigured platform indefinitely, because the log's value is in the writing at the moment of change, not in the furniture around it. Start in whatever you already open daily. If the log later outgrows it - two people editing, a year of entries - migrate then, with real content to organise. The same logic applies to this whole bench: the instruments serve the habits, and a small store should buy instruments last and build habits first.