If You Can't Remember What You Changed, You Can't Know What Worked

Managing a large account is a stream of small decisions. You raise a budget here. Add three negatives there. Swap a headline. Move a campaign from broad to phrase. Adjust a bid strategy. Pause an ad group that's been limping. On a busy account you might make a dozen meaningful changes before lunch, each one a small bet that things will be a little better after than before.

And here's the question that's supposed to follow every one of them: did it work?

Most of the time, nobody actually answers it. Not because anyone's lazy — because the bookkeeping is brutal. To know whether the match type change helped, you have to remember you made it, remember which day, remember what the numbers looked like before, and remember to go back and look two weeks later when there's enough data to mean anything. Multiply that by a dozen changes a day across several accounts and the honest truth is that most adjustments get made and then quietly orphaned. The needle moves, or it doesn't, and you're rarely sure which of your forty tweaks was responsible.

Some people fight this with a spreadsheet. Some use a project management tool. Most use a combination of memory and hope. None of it is connected to the actual performance data, so even the diligent version is a log of what you did, sitting in one place, while the results live somewhere else entirely.

Now you can just tell the agent.

When you make a change you think matters, ask ROAS Radar to note it. Budget adjustment, keyword addition, ad copy swap, match type change, a new bid strategy, a structural reshuffle — technical or creative, it doesn't matter. The agent records what changed and when, against the same live account data it's already analyzing. The change and its context now live in the same place as the performance it's meant to affect.

Then you tell it when to remind you. Two weeks out, when the test has enough data to be real. The first of next month. After the seasonal spike clears. Whatever window actually makes the comparison meaningful for that change. When the time comes, the agent comes back to you — not with a vague "you changed something around here," but ready to measure the after against the before and tell you whether the bet paid off.

That last part is the whole point. A change log by itself is just a diary. What makes it valuable is the comparison against the control state — what the account was doing before you intervened. Because the agent already has the account's history and already knows your goals, it can frame the result the right way: not just "CPA went down," but whether this specific change plausibly drove it, against this account's actual ROAS target, accounting for what else was moving at the same time.

Stack enough of these and something changes about how you operate. Your account stops being a place where things get tinkered with and starts being a place where things get tested. Every meaningful change becomes a small, tracked experiment with a built-in follow-up date and an honest read on the result. The wins get confirmed and kept. The duds get caught and reversed instead of silently dragging for months. And the incremental gains — the 2% here and 3% there that compound into a genuinely better account over a quarter — actually get found, because for once you can tell which moves earned them.

You were already making the changes. Now you get to find out which ones were worth making.