Spend changes and what followed
How MeasureScale shows how ad spend and your results move together, and why the answer describes association and never says an ad caused a result.
The impact analysis looks at your daily Google Ads spend beside your GA4 sessions and key events. With a Search Console site, it adds your organic clicks. It shows how they moved together and what happened around big spend changes.
The answer shows association. It does not show cause. Read read it as association, not cause before you act on it.
What it answers
- Do my ad spend and my total results move together?
- Did my total conversions fall when I cut spend?
- What happened to organic clicks when spend changed?
What it reads
- Google Ads: daily cost, clicks and conversions for one account, or for named campaigns.
- GA4: daily sessions and key events. Sessions from the Paid Search channel count as paid. Sessions from every other channel count as non-paid. You name the GA4 property, and you can name one key event.
- Search Console (optional): daily organic clicks, when you name the site.
- Notes and Google updates: your annotations for the site and Google’s published ranking updates. They flag other events near a spend change.
The period
- With no dates, the period is 180 days ending at the newest day that all sources have settled.
- Google Ads settles yesterday. GA4 and Search Console settle about 3 days late.
- A day with no Google Ads activity counts as a day of zero spend.
- A GA4 or Search Console day with no data counts as zero only between that source’s first and last day with data. A missing day at either end is dropped, because the data may not have been collected.
- Days follow the Google Ads account’s time zone. The answer warns when GA4 or Search Console use a different one.
- A period that is too long to read in full is refused. MeasureScale does not drop the newest days.
Two views
How spend and results move together
MeasureScale compares daily spend with four results: GA4 sessions, GA4 key events, organic clicks and non-paid sessions. It allows for the usual weekday pattern and the long trend, then checks how closely spend and each result move together. It tries results on the same day and on the days after. The default delay is 7 days, and you can ask for a longer one.
The answer shows the full table and names the best day. The best day is picked from the table, and the answer says so.
| Label | Needs |
|---|---|
| Strong | The two move together at 0.5 or more on a scale from 0 to 1, clearly apart from zero after the corrections below, with at least 30 effectively independent days |
| Moderate | They move together at 0.3 or more, clearly apart from zero after the corrections |
| No reliable relation | Anything less |
- Neighbouring days look alike, so 100 days of data count for fewer than 100 independent days. MeasureScale allows for that, and gives no result when too few independent days remain.
- Many pairs are tested at once, so the bar for “clearly apart from zero” rises with each test.
- With too short a period for the delay you asked for, no result is given.
- When daily spend hardly varies, a relation cannot show. If the day-to-day swing in spend is under 15% of the average, the answer flags it and no label rises above “no reliable relation”.
What happened around a spend change
MeasureScale finds the days when daily spend moved to a new level.
- A change counts when spend shifted by at least 25% and each side lasted at least 7 days.
- Notes and Google updates within 7 days of a change are attached. The answer then marks the change as mixed up with another event.
For each change, MeasureScale compares the results in the days before with the days after. It then asks how unusual that shift is by comparing it with the same shift on other dates that had no spend change. Those other dates show how much results move on a normal day.
| Label | Meaning |
|---|---|
| Unusual compared with normal ups and downs | Fewer than 10 in 100 of the other dates shifted as much, and there are enough other dates |
| Within normal variation | Results often shifted this much on dates with no change |
| Not enough history | There are too few separate other dates. The shift cannot be judged |
| Cannot be computed | The results before the change were zero, so no percent shift exists |
Other dates close together overlap, so they count as fewer separate dates than they seem. Each extra spend change leaves fewer. A label needs enough separate dates.
For a spend drop, the answer also shows the net view:
- how many paid clicks per day you lost
- how organic clicks and non-paid sessions changed per day
- how total key events changed per day
That answers “did cutting spend hurt my total conversions”. It does not estimate a saving.
Read it as association, not cause
The words “moved with” and “was associated with” are deliberate. MeasureScale never says spend “caused” or “drove” a result.
- Seasons, sales, site changes, news and competitors move these numbers too.
- One spend change gives one before-and-after pair. That is one data point.
- The best day in the table is chosen from the table. The corrections cover the tests run. They do not cover the choice.
- A change picked out of noisy spend can be a false alarm. Read each change with the length of its two sides and its label.
- “Unusual compared with normal ups and downs” says the shift stands out against your own history. It does not say that the ads caused it.
- Your Google Ads conversions and GA4 key events often count the same purchase. MeasureScale shows both and never adds them together.
- GA4 and Google Ads credit conversions differently. The split into paid and non-paid depends on GA4’s default channel grouping.
- The analysis does not estimate the extra conversions that ads bring, or what you would save.
To find out what ads add, run a holdout test. See overlap is not incrementality.