Before assuming the account is unstable, check whether the measuring instrument changed underneath the series. Every published answer says "wait for more data" or "widen the window." None raise the possibility that the ruler moved.

On this page

• Why is ROAS different on an account with static conversion values?
• What does the measured data actually show?
• Can a value factor change loosen my cost target without anyone deciding?
• How do I check whether my conversion values changed?
• What should I report instead of ROAS?
• What if my store passes real order revenue into Google Ads?

Why is ROAS different on an account with static conversion values?

If every conversion is written with the same default value, then:

ROAS = value factor ÷ cost per acquisition

Which means a target ROAS is a cost cap, not a return target. And it means any change to the value factor moves reported ROAS with no change in performance whatsoever.

What does the measured data actually show?

Pulled directly across one book of accounts, the value factors written onto a booked job moved between 1.4× and 3.0× inside 12 weeks:

Nobody broke anything. The ruler just changed, repeatedly, and every ROAS series spanning those changes is measuring in mixed units.

A "12-week median ROAS" across that is not a statistic. It is an artifact.

Can a value factor change loosen my cost target without anyone deciding?

Hold a target ROAS constant while the factor rises and you have silently loosened the permitted cost per acquisition in exact proportion.

Real case: a campaign held at target 5.0 while its factor moved 666 → 950. Permitted cost per booking went from $133 to $190. A 42.6% loosening that appears in no report.

How do I check whether my conversion values changed?

Pull value per conversion, per week, per conversion action. If your account writes a static default, that figure should be a flat line. Every step in it is a ruler change, and every ROAS comparison across a step is invalid.

What should I report instead of ROAS?

Cost per acquisition. It is ruler-independent.

But be honest about it: it is not more stable. Across that same book, its coefficient of variation ran 0.17 to 0.82, with individual locations swinging from $185 to $1,698. Neither metric supports a point target from platform weekly data alone.

The correct derivation order is: pin the current factor → choose a target cost per acquisition from settled business data → set target ROAS = factor ÷ target cost.

What if my store passes real order revenue into Google Ads?

Then your ROAS swings are genuinely about performance and mix, and the answers are the ordinary ones: small sample, product mix, promotional periods, new-versus-returning composition. Check the month-boundary throttle before reading any window that spans the 1st.

Related questions

What should I actually be doing to a Google Shopping campaign every week?

Less than most people do. The weekly job is reading the right numbers on a settled window and making at most one change — not optimising something every time you log in. Read the full answer.

Why can't I see all my search terms in Google Ads, and how much of my spend is hidden?

Google withholds terms that were not searched by "a significant number of users," which on most accounts leaves a material share of spend unattributed to any visible query. The fix is not to recover the terms — it is to stop needing them. Read the full answer.

How do I work out what target ROAS to actually set for my store?

Derive it from contribution margin and your fixed per-order costs, not from a benchmark. The formula everyone uses — 1 ÷ gross margin — is a breakeven floor, not a target, and it ignores the per-order costs that decide whether an order is actually worth having. Read the full answer.

Written by Usama Khan, who runs Google Ads for e-commerce brands and home-services companies at AdventurAds. Figures are from accounts under active management and are anonymised. Last reviewed 2026-08-26.