Estimated vs. measured savings in ad accounts
Potential, estimated, reduced spend and measured savings are different numbers. How each is calculated, and why mixing them misleads you.
On this page
If a report about your ad account shows a single "saved" figure with no period, baseline or method, you can't tell what you're looking at.
Four numbers that get called "savings"
| Term | What it means | Typical source | Can it be added to the others? |
|---|---|---|---|
| Potential savings | Spend currently exposed to a problem | An audit | No; it's before any change |
| Estimated savings | A calculated effect of a change, with assumptions | Formula after a change | Only with other estimates, deduplicated |
| Reduced spend | Spend went down after a change | Platform data | No; it says nothing about results |
| Measured savings | Effect observed against a credible baseline | Experiment, holdout, or a sound causal design | Only with other measured results |
How an estimate is usually built
Take negative keywords for search terms that spent without converting. A conservative estimate projects their past cost forward and discounts it, because part of the budget moves to other queries rather than disappearing.
Rules for a conservative estimate
- Use a fixed lookback window (for example, the 14 days before the change).
- Discount for budget that will be reallocated (for example, count 70%).
- Count it only if the campaign's conversions didn't fall after the change.
- Never count the same blocked spend under two findings.
Even with all those safeguards, it's still an estimate.
Worked example: same change, three numbers
Worked exampleIllustrative example; the numbers are not real account data
On September 1 you approve negatives that block search terms which spent $1,400 over the previous 14 days with no conversions. Fourteen days later:
| Number | Calculation | Result | Label |
|---|---|---|---|
| Potential savings (before the change) | $1,400 over 14 days | $1,400 per 14 days exposed | Potential |
| Estimated savings (after 14 days) | $1,400 ÷ 14 × 14 days × 0.7, conversions stable | About $980 | Estimated |
| Reduced spend | Campaign spend fell from $6,000 to $5,300 over comparable periods | $700 less spend | Reduced spend |
| Measured savings | Requires a controlled comparison | Not available | - |
- Potential$1,400
- Estimated$980
- Reduced spend$700
Reduced spend coming in below the estimate is normal: seasonality, competitors, Smart Bidding and reallocated budget all move it. A measured figure needs a campaign experiment or a holdout, designed in advance with enough volume. Until then the honest answer is "estimated: about $980; measured: not available yet."
Before-and-after comparisons (CPA four weeks before vs. after) are a reasonable signal worth reporting, with the period stated and seasonality adjusted where possible. But they're still estimates: a sale, a competitor leaving the auction or a tracking change can produce the same pattern.
Rules that keep savings reports honest
- Every figure states its period, baseline and method, or links to them.
- Estimates and measured results have separate totals. Never add them.
- Saved spend and added revenue stay separate; no combined "profit" number.
- Changes that made things worse are recorded too.
- Brand traffic that grew on its own isn't a saving from your changes.
- A change that was kept isn't proof it helped; it only means it wasn't restored.
- Restoring settings doesn't recover money already spent, and the report shouldn't imply that it does.