Identifying Google Ads waste without drawing false conclusions

A row without a conversion is therefore a review task, not a finding.

Why is “cost without a conversion” not a definition?

Seven documented platform mechanics produce rows without a conversion that are nevertheless not waste. They must be checked before any figure is interpreted.

• Interaction time rather than conversion time. Google Ads reports the standard conversion columns by the time of the ad interaction (usually the click), not by the time of the conversion. The available by-conversion-time columns report a conversion at the time it occurred. Call conversions have documented source-specific date treatment.
• Conversion window. The default window for click conversions is 30 days; depending on the source, 1 to 90 days can be set. A window change affects conversions recorded from the change onwards – including later conversions from earlier ad interactions. It is not limited to future clicks.
• Conversion delay. Conversions can be reported up to 90 days after the click. The “Days to conversion” segment breaks the observed delay down into up to 19 rows; the configured window and the coverage of the source limit what becomes visible. It therefore does not show an uncensored, complete conversion-delay distribution.
• Reporting delay. For conversion reporting, Google states a delay of 24 to 48 hours. That range applies to the reporting described there, not as a uniform delay for every source. Yesterday’s figures are therefore not a sound basis for a final judgement on conversion performance; as a reason to look more closely at cost, click or delivery signals, the previous day serves very well.
• Modelled conversions. Google supplements conversions that it cannot observe directly through modelling and includes them in the report only at high confidence. Up to five days may pass before processing is complete and stable. This is a measurement statement about data processing, not a statement about anonymity.
• Retrospective corrections. Conversion adjustments change recorded values and counts. A reversal or an adjustment to zero removes the conversion from the report. In its documentation on conversion adjustments, Google recommends ending the reporting period at least 30 days before today – earlier if the conversion window is longer. This recommendation concerns adjustment reporting; it is not a general waiting period for every waste decision.
• Invalid clicks. Google tries to filter clicks identified as invalid out of reports and payments automatically so that they are not charged. That is not a guarantee of complete removal. What has already been filtered out is no longer available as a saving.

In addition, there are settings that break nothing but shift figures. Secondary conversion actions normally appear only in “All conversions” and not in the “Conversions” column; if an action belongs to a custom goal that the campaign actually uses, it can also enter the “Conversions” column and bid optimisation. The counting method “One” counts one conversion per ad interaction, “Every” counts every conversion after an ad interaction. A row can therefore be valuable and invisible in the wrong column at the same time.

Which six points apply before a finding?

The six points are a checklist, not six gates of the same kind. Where the decision depends on stable conversion evidence, data maturity must be demonstrated; obvious irrelevance and a strategic decision do not need a mature conversion window. As long as a basis needed for the specific decision is missing, the result must not be issued as a “waste” finding.

The same rule applies to every precondition: Not available is not zero. An empty field, missing access or an offline import that was never submitted are data gaps; they are reported as a gap and not counted as evidence against the entity. A grouped row is not automatically unusable: it may be used as a subset if coverage and denominator are stated – the denominator then refers to exactly that subset. Missing rows do not justify silently treating the observed subset as the whole account. An unknown total spend is not estimated.

Whether your business conversion goal passes this check is decided beforehand in the conversion tracking audit. Without that judgement, every performance-based waste claim remains an unverified conclusion.

Where can spend arise – and who owns the method?

Waste is not a search terms topic alone. The diagnosis runs across several sources; the method for each area has its own dedicated page.

For the “irrelevant demand” area, the search terms and negative keywords method is authoritative; this page only carries the judgement on whether a finding reaches the evidence threshold.

How maitiq documents changes: the “automation” row is answered by a decision log. Every maitiq proposal names the action, the rationale and the data basis with the old and the new value where such a comparison of values exists. You decide manually, or a configured limited autopilot approves and applies eligible proposals within its authorised limits; the implementation is a separately logged step within valid authority – no anonymous change arises in the process.

What do the reports not show?

The search terms report shows search terms that a significant number of people used and that triggered an ad. Terms with too little activity are omitted for data protection reasons. The search terms insights group terms with lower spend, fewer conversions or clicks and no recognisable category together as “Other search terms”; they are available for Search, Performance Max and Shopping campaigns at account and campaign level and for periods from March 2023. Performance Max has a search terms report of its own, likewise from March 2023, which does not include store visits and store sales. Negative keywords there affect Search and Shopping inventory; Display and Video inventory is managed via excluded content keywords in the content suitability settings.

One rule follows for every evaluation: a missing row is “not available”. It is not evidence that no cost arose there, and it must not shrink any denominator.

How does the evidence ladder classify a finding?

The evidence ladder is a proposed framework on this page, not the complete list of states that maitiq operates in production. It assigns a finding to one of five states and is neither a scoring system nor an automatic mechanism. The levels are not a mandatory sequence: a strategic decision may be necessary without a monetary materiality threshold, and an unverifiable evidence situation is not a higher severity but a different question.

All thresholds on this ladder are set in the account: together with the people responsible, in account currency, against the approved plan. Publishing a CPA, click or cost threshold as generally valid is out of the question, because a figure from another account creates exactly the false certainty this process is meant to prevent. “Not verifiable” is a fully valid, visible result and not an intermediate step.

A proposal can arise at the “material” level and then names the source rows it rests on. A strategic question – brand, competition, market entry or business boundary – is treated as a decision with named accountability and does not need a monetary materiality threshold; the same applies to an exclusion based on obvious irrelevance. “Not verifiable” leads to a question for you, not to an action.

What belongs in the decision sheet?

A finding without provenance is an opinion. The review schema on this page keeps source and judgement separate and only lets a row into the assessment once all evidence relevant to that decision from section A and section B is available.

Section A – provenance. Source row identifiers, account and entity, entity type, period in the account time zone, account currency, cost field, clicks, deduplication key, coverage status of the data source.

Section B – measurement. Business conversion goal and its selected actions with their respective primary or secondary status, counting method, conversion window, observed delay in the account, consent and modelling status, maturity judgement, reconciliation with the business source.

Section C – assessment. Campaign role and approval, relevance to the business with rationale, account-specific materiality rule, conflict check against positive keywords, evidence level.

Section D – decision. Proposed action, scope of the action, expected effect with formula, person responsible and approval, verification date, rollback where available, residual risk.

If a value relevant to this decision is missing in section A or B, the row stays at “not verifiable” and moves onto a clarification list; a performance judgement is then not possible. Genuinely inapplicable conversion, consent or modelling fields are recorded as “not applicable” with a reason, never as zero. An empty field is never normalised to zero, and a calculation without referenced source rows is not valid.

How does the scenario calculation work?

For a confirmed row, a transparent linear projection is permissible:

deduplicated observed cost in the evidence window / calendar days actually covered × 30.4

The factor 30.4 is a rounded approximation of 365 days / 12, the average length of a month, and serves only comparability with a monthly period. It is not a cap on your costs and not a forecast. The calculation is an extrapolation under the stated assumptions, not a promised saving.

The result cannot be calculated if any of these conditions applies:

1. no referenced source row identifiers;
2. overlapping or double-counted rows;
3. mixed or missing account currency;
4. an unclear or interrupted period, or one defined outside the account time zone;
5. no covered calendar days or a denominator of zero;
6. the coverage status of the source is unknown, or the measurement is not mature for a performance judgement.

In the output block of your calculation, record observed cost and the linear scenario separately and, next to every figure, name the source, period, unit, deduplication rule and assumption. The calculation is an estimate: the scenario assumes comparable delivery and that an action prevents exactly this spend. Demand, auction, seasonality and budget shifts can change the result. Non-overlapping cost rows reported separately may be summed, even if the terms are variants; overlapping observations are deduplicated. When estimating demand volume, however, variants are not added up into a market size. A known zero with a valid denominator remains zero. “Saving” is only permissible after an approved change and mature follow-up observation – and even then causality remains an open question.

How maitiq takes on this review

This framework is this page’s method, not an algorithm automatically enforced by every maitiq workflow: in the audit, the check depends on the data supplied or connected, and recurring implementation checks run at the cadence configured for the account. Whatever reaches the evidence threshold comes to you as a proposal with the action, rationale and data basis. You decide manually, or a configured limited autopilot approves and applies eligible proposals within its authorised limits; every implementation is logged separately within valid authority. Explicit approval to exclude specified keywords already covers the associated guarded exclusion action. A proposal states which evidence supports it – and where the evidence for a confident judgement is still missing.

How do you protect valuable demand?

A positive conflict exists when a planned response would also block known or plausible valuable demand. Negative keywords can limit irrelevant search queries, but they do not cover similar variants the way positive keywords do. That is why, before every exclusion, the check asks whether a tighter keyword option, a narrower scope or observation instead of exclusion is the right answer. The check does not protect every possible piece of demand; it makes the remaining risk visible and writes it down.

Where does this judgement end?

A generally valid threshold from which cost without a conversion counts as waste is deliberately absent here – as is a Swiss benchmark, an industry average or an estimated saving in francs or as a percentage. The conversion tracking audit and the budget and governance methods remain separate steps, and no campaign is judged without account access, knowledge of its role and business context.

Frequently asked questions

Sources and how to read them

The linked pages are Google’s official platform documentation: they describe how the reports and conversion columns work and where their limits lie. They support the platform statements in this guide but are not an independent measure of success or of the market. Information about how maitiq works is available at maitiq.com.

Have maitiq review your specific case.