Impact: Measuring Whether Your PPC Changes Actually Worked | Docs

For · Operators who want to know whether the changes they made actually worked Time · 5 min read Next · Testing timelines and evaluation windows

Most reporting tells you how an account is doing. Impact answers a narrower and harder question: of the specific changes made to this account, which ones did anything. It reads the changes you recorded in your Changelog, measures what followed each one, and says plainly who made the call.

The Short Version

Impact measures every change you record, whether Trellis recommended it or you decided it yourself.
Changes you decided are measured with the same instrument and are never credited to Trellis. The ledger labels each row “Trellis recommended” or “Your change”.
The headline figure is an observed fact, what an entity spent minus what it returned, not a projection.
Impact does not claim that a stopped campaign’s budget moved somewhere better, because that is not something the data supports.
A tracking fix gets a written confirmation and no number, because the honest number does not exist.
Nothing enters the ledger automatically. Recording the change is what puts it there.

Two Lanes, and Only One of Them Is Ours

Every measured change lands in one of two lanes.

Trellis recommended covers changes that started as a recommendation in a Catalyst Audit and that you then implemented and recorded.

Your change covers everything else: the decisions you made on your own judgment, recorded in your Changelog, with no recommendation behind them.

Both lanes are measured by the same code, over the same windows, to the same standard. The difference is attribution, and the rule is one-directional: Trellis may point to lane one when it talks about whether its advice works. It may never point to lane two.

This is not a formality. In a real account, the single largest measured figure came from a campaign the operator stopped against a threshold they had set themselves weeks earlier. Trellis never recommended it. Putting that number in a Trellis track record would be claiming someone else’s decision. Leaving it out of the ledger entirely would mean the ledger described only part of what actually happened to the account. So it is measured, shown, and attributed to the person who made the call.

Where your Changelog entry states your own reason for a change, the ledger leads with your words rather than narrating over them. The reason you wrote at the moment of the decision is better evidence than anything a measurement can infer afterward.

What Gets Measured

A change is classified into one of five kinds, and the kind determines what the row can honestly report.

The entity stopped spending. The figure is net outflow: what that campaign or keyword spent minus what it returned, across its run. Two measured numbers and one subtraction. No forecast, no assumption about what would have happened otherwise.

It kept running and a rate moved. Cost per acquisition, return on ad spend, conversion rate. The row reports the measured change with its range, over a window sized to the kind of change you made.

You added something and volume moved. New keywords, a raised budget, an expanded audience. The row reports the change in conversions or revenue.

A tracking or structural fix, confirmed in place. A written confirmation, and deliberately no number. More on this below.

Not measurable. Sometimes there is not enough data, or the account changed in too many places at once to separate one effect from another. The ledger says so rather than producing a number it cannot stand behind.

If a change turns out to have been costly, the row says so. A campaign that was profitable and got paused anyway shows a negative figure. The ledger is a record, not a highlight reel, and a change that went badly is the one you most need to see.

What Impact Will Not Claim

Two refusals do more for the credibility of this feature than any number in it.

It does not claim your budget moved somewhere better. When a campaign stops, it is tempting to report that the freed spend went to work in a healthier campaign. Trellis tested that idea against real account data and abandoned it. Account-level spend swings substantially week to week on its own, so any rule of the form “total spend held steady, therefore the budget was redeployed” fires on coincidence about as often as on fact. More fundamentally, campaign budgets are not a shared pool: pausing one campaign does not raise another’s budget. That only happens if the second campaign was limited by its budget and somebody raised it. So the ledger reports what stopped, states the rest of the account’s totals as neutral context, and draws no arrow between them.

A tracking fix gets no number. If you repair conversion tracking, more conversions get counted afterward. Almost none of that is new business. It is the same business, finally being recorded. Reporting a conversion increase for a tracking fix would measure the instrument rather than the account, and it would be the most misleading figure this system could produce. So those changes are confirmed in writing and carry no value.

Where Impact does estimate rather than observe, the estimate sits below the facts, is labelled, is bounded to the window that was actually measured, and is never annualized.

Putting a Change in the Ledger

Nothing arrives on its own. Trellis is read-only against your ad platforms, so the sequence is yours to start.

Make the change in Google Ads or Microsoft Ads.
Record it in your Changelog. From a recommendation you can do this in one step from the audit or the dashboard, marking it as implemented or modified with a note on what you did differently. For a change of your own, write the entry directly. Either way the entry is yours, and the note you leave is the part no ad platform holds.
Wait for the window. Every change is measured over a window sized to the kind of change it was, because a bid adjustment and a budget increase do not settle on the same timetable. The window is fixed when you record the change, so a later adjustment to how Trellis sizes windows cannot rewrite what your past changes were measured against.

The single most useful habit is naming the campaign and stating your reason in the entry. Both make the measurement sharper, and the reason is what the ledger will quote back to you months later when you have forgotten why it seemed like a good idea.

Where to Find It

Impact is on the Pro plan, and it is in beta while we finish it. Collection runs for every plan, so the record accumulates from your first recorded change regardless of what you are on, and it is there when you arrive.