What should I do if one unusually high-value order is skewing my campaign's average ROAS?
What to check before you touch anything
Is this really a one-off anomaly (a single large order) or a recurring, if infrequent, pattern (say, bulk orders once a month)? These are different situations with different fixes.
How much does a single order actually move the aggregate numbers for the period? If you exclude it from the calculation, how much does the overall picture of ROAS and campaign performance change?
Could the unusually high value reflect a data error (a duplicated amount, an incorrect currency conversion, a wrong unit count) rather than a genuine transaction? Before analyzing this as a "distortion," rule out the possibility that the number is simply technically wrong.
Is this product/segment structurally separate from the rest of your keywords and campaign setup, or is it mixed in with typical orders in the same ad group? This determines whether you can even isolate its impact without restructuring.
How do competitors operate specifically in this segment (large/bulk orders)? If it's a niche, low-competition demand pocket, the unusual value may reflect a real, if irregular, market opportunity worth treating separately from the main mass-market competition around typical orders.
Possible approaches
If the anomaly is a technical error in how value was passed, fix it at the tag/feed level first, before drawing any conclusions about campaign performance.
If it's a rare but genuine and recurring pattern (large one-off orders), consider carving that segment out (by product category, by order price range) into a separate campaign with its own, either more conservative or more flexible, target, so it doesn't distort optimization for the main mass-market campaign.
For reporting to the business, it's useful to explicitly show ROAS with and without the anomalous order (two numbers side by side) — that gives a more honest picture than a single blended average that could be misleading about typical campaign performance.
Google generally doesn't provide tools to "manually exclude" individual anomalous conversions from bid optimization — if such cases are recurring, structural isolation (a separate campaign/group) is really the only practical path, rather than trying to "clean up" the statistics after the fact.
If large orders come from a niche, low-competition demand segment, consider deliberately developing that direction separately (dedicated keywords, a dedicated wholesale/B2B offer), rather than trying to force its logic into a general retail campaign optimized for typical, mass-market demand.
Before deciding what to do with the anomalous order, it helps to understand whether it's part of a stable ("within") contribution to results — tied to specific, consistently strong-performing keywords or assets — or a one-off statistical outlier ("mix" effect) not tied to any underlying pattern. Sorting this out by hand, order by order and product by product, is labor-intensive. Our tool (DataMind) shows which keywords and assets are actually driving stronger results than others and which of that is a stable (within), rather than random (mix), factor in your results — so the anomaly is immediately visible in context: is it an exception to the overall picture, or a reflection of a genuinely working, repeatable segment worth developing further?
Related Content
Check the campaign's status in the interface — is there an explicit "Limited by budget" or "Bid strategy constrained by target" flag alongside the overperformance? These are two different root causes producing the same symptom.
Check the campaign's status — an explicit "Bid strategy constrained by target" flag alongside underspent budget (no "Limited by budget" flag, but actual spend running well under the daily limit) is a direct sign that the target is what's holding volume back.
How sharp was the budget increase, and when did it happen? A gradual 10–20% raise and an abrupt 2–3x jump lead to very different behavior — the latter almost always resets part of the strategy's accumulated learning.
What exactly is being passed as "value" — the full order amount (revenue), or an amount already net of discounts, shipping, and taxes? Different systems (your website, your CRM, Google Ads) may define "value" differently, and it's worth cross-checking line by line on a handful of real orders.
How is ROAS being calculated for each product? tROAS optimizes toward conversion value (by default, usually the sale price/revenue), not automatically toward margin. A high-margin product with a lower price and/or lower conversion potential can post a lower "raw" revenue-based ROAS than a cheap, low-margin, high-turnover product — and the system will end up underrating the very product that's more valuable to the business, unless it's told about margin separately.
Does Google Ads actually receive information about whether a purchaser is new or returning? Without that signal, neither a separate goal nor separate reporting is technically possible.
What attribution model does each system use? Google Ads defaults to its own attribution model (usually data-driven), and GA4 may use a different model (or the same model with different cross-channel distribution rules) — the mismatch may be purely a methodology difference, not an error.
The practical way to check this is through Report Editor (or a standard report) broken out by Conversion Action, using the "All Conv. Value" metric. Building a report segmented by specific conversion action and comparing count/value per action makes it easy to spot whether the same purchase is being double-counted across multiple conversion actions (say, between the native Google Ads tag and a GA4 import) — in that case, the account's total "All Conv. Value" will noticeably exceed real revenue, and the breakdown by conversion action will show that several different actions are, in effect, recording the same transactions.
What was the campaign's status during the seasonal peak? An explicit "Limited by budget" or "Bid strategy constrained by target" flag on those specific days points to the target or budget physically preventing the system from taking advantage of the demand spike.
How long, and how consistently, has actual ROAS been beating target? A one-off overshoot from a couple of big orders is different from sustained overperformance across several weeks in a row.
What's the real median gap between click and purchase for this business (the Time Lag report in Google Ads)? Without this number, evaluating "recent" periods will be systematically skewed on volume and distorted on ROAS.
Campaign structure. Is the volume and makeup of keywords/targeting comparable — number of keywords, match types, number of ad groups, audience signal coverage (for Performance Max)? A campaign with a narrow structure physically can't deliver the same volume as one with broad coverage, even with an identical target — this isn't an efficiency question, it's a question of the structure's scale.
Pull up the "Conv. value / cost" and "Value / conv." columns broken out by Conversion Action — if the average conversion value looks suspiciously identical (say, exactly the same number) across many transactions of clearly different size/composition, that's a clear sign a fixed default value is being used, not the real transaction amount.
Campaign structure. Are different product categories currently split into separate campaigns/groups, or is everything running through one shared structure (especially relevant for Shopping/Performance Max)? Without structural separation, it's technically impossible to assign different targets to different categories in the first place.
Is the moment of refund reflected in the passed conversion value at all? If value is passed once, when the order is placed, and never adjusted afterward, ROAS in Google Ads will be systematically overstated relative to the business's real revenue, by the full amount of refunds.
Campaign structure. How broad is the campaign's own keyword/targeting setup? If the structure is narrow from the start (few keywords, tight audience signals), the system physically has nowhere to expand beyond the narrow segment it's already found, regardless of the target.
Campaign structure. Is your keyword/targeting setup broad enough that, once the target constraint is loosened, the system actually has somewhere to expand in a controlled way, rather than jumping straight into a wide pool of low-quality traffic?
Campaign structure. Is the keyword/targeting coverage broad enough that the system could actually find additional volume if the constraint were lifted? If the structure is narrow, the "constrained by target" flag may technically be there, but the real headroom for growth may still be modest.
Is the value rule actually active and applied to the specific campaigns/audiences/locations/devices it was built for? Value rules need to be explicitly turned on at the campaign level, and a rule that's configured but not connected to the right campaigns simply won't fire.
