What a Paid Search Ad Account Record Proves

Three tiers, sorted by the evidence rather than by the dispute

Most writing about paid search evidence is organized around the dispute: click fraud in one place, agency negligence in another. That is the wrong axis for the question an attorney actually has, which is whether the record in front of them carries the weight the pleading puts on it.

So I sort it by the evidence. Tier one is what the account record shows on its own, exported by one person with access, without asking anyone for anything. Tier two is what it shows only beside records held by the other side or a third party. Tier three is what it never shows, however complete the export and however early the preservation.

Tier three is the part worth reading. Intent, causation between a spend change and a revenue change, what a manager considered before making a change, and profit are not in the account, and no volume of data puts them there. Whether any of it is actionable is a question for counsel. What follows is what the record shows.

Tier one: what the account record shows on its own

One person with account access, asking nobody for anything, can establish all of this:

• The change log. Change history records the email address attached to a change, the entity, the field, the old and new values and the timestamp, covering edits through the web interface, Google Ads Editor, automated rules and the Google Ads API.
• Configuration, as it stands and as it changed: budgets, bid strategy, match types, negative keyword lists, location targeting options, campaign types, and the settings on each conversion action.
• Spend and delivery, segmented by campaign, ad group, keyword, ad, device and audience.
• The Search terms report, for queries that clear Google's privacy volume threshold.
• Auction Insights, reporting impression share, overlap rate, position above rate, top of page rate, absolute top of page rate and outranking share on search campaigns.
• The Invalid clicks column, a count of clicks the platform filtered before billing.
• Auto-apply recommendation subscriptions, and the billing and transaction record.

The strongest tier-one finding is usually internal inconsistency rather than any single row: the manager did a thing in one campaign and not in the identical campaign beside it. That finding needs no external benchmark.

One question worth asking on day one: did the advertiser run Search Ads 360? Its change history reaches back to May 21, 2016 or later.

Tier two: what the record shows only alongside someone else's documents

Much of what an attorney wants is a joint of two records, only one of which sits in the ad account:

• Web server and CDN logs, the only independent observation of the traffic itself and the only place anyone sees IP addresses, user agents, timing and session behavior. Without them an invalid traffic question stops at a count.
• Agency invoices against platform billing, two records kept by two parties. Reconciling them is frequently the whole question in an overbilling matter.
• Tag manager version history, often the only record of when a conversion tag was changed, broken or removed.
• Order and CRM records, the only way to turn a reported conversion count into a count of sales.
• Correspondence. The account shows a campaign was paused; it does not show that someone asked for it.
• The contract stack, and any third-party detection vendor's output.

That last one carries a warning. A vendor's classification of clicks and the platform's invalid-click count are not the same measurement: the platform filters before billing on signals the advertiser never sees, and the vendor measures after the click lands on signals the platform never sees. Google publishes its own reasons its click count diverges from third-party tracking, among them repeat clicks, stripped referrer headers, browsers without JavaScript, and clicks its own protection filtered out. Read plainly, that is the platform stating billed traffic and independently measured traffic are different populations. Differencing the two numbers is not a damages method.

Tier three: what the account never shows

Four things are simply not in the record, and a page that does not say so is not being straight with you.

Intent

The change log has no field for reason, purpose or authorization. It records that a budget went from one figure to another at a time, under an email address. It does not record why, and an email address on a row is not a state of mind.

Causation between a spend change and a revenue change

The account records the spend and it records the reported conversions. It connects neither to the other. Cost per click is an auction outcome that moves when competitors move. Seasonality, a competitor entering or leaving, Google's own product changes (match type behavior, automated bidding defaults, attribution model changes) and the advertiser's own pricing, stock levels and site changes all shift measured performance with no party doing anything. A sequence in the data is not a cause, and no ad account holds a record of what would have happened instead.

What was considered

Nothing logs deliberation. Google also states that not all account-level settings changes, and not all changes made by Google representatives during consultations, appear in the history, so advice given on a call can leave no trace in the client's own record.

Profit

The account holds cost. It does not hold margin, cost of goods, returns, refunds, chargebacks or fulfillment cost. Quantifying lost profit is work for a forensic accountant or an economist; what I supply is the factual predicate underneath it.

Why several of the account's numbers are not measurements

A number on a screen invites an attorney to treat it as a business record. Several of the most quoted ones are estimates, subsets or relative positions.

The conversions figure is a blend. Google's documentation states that in the "Conversions" column it reports both modeled and observed conversions, and the standard interface has no toggle separating them.

The Search terms report is a subset. Queries below Google's privacy volume threshold are withheld. Aggregate click and cost totals still include the withheld queries, so search terms rows do not sum to campaign totals. That gap is expected behavior, not evidence of tampering.

Auction Insights is relative, not absolute. It shows nothing below a 10% impression share, excludes search partner traffic, and Google's own example notes an advertiser showing 100% impression share may share those auctions with a competitor whose report shows 50%.

Two Google products will not reconcile to zero. Google Ads counts clicks and dates conversions to the click; Analytics counts sessions and dates conversions to the conversion. Discrepancy is the baseline condition. No published, sampled figure exists for how large a normal discrepancy is, which cuts both ways: an opposing expert calling a gap normal has no benchmark either.

The retention clock decides which tier a question is still in

The windows run from the date the account is examined, not the date of the conduct. When one closes, a tier-one question quietly becomes a tier-two question: the answer is no longer something your client can export, it is something another party has to be asked for.

Sources, all read August 14, 2026: Google Ads Help, "About change history" and the Google Ads Data Retention Policy, whose 37-month, 11-year and 3-year figures took effect June 1, 2026; Google Ads API documentation for change_event and change_status; Microsoft Learn, "Reporting Data Retention Time Periods"; and Search Ads 360 Help. Microsoft's six-month figure is documented for its reporting API; the window in its web interface is not documented in a source I can cite, so I verify that in the account.

One clock is easy to miss: Analytics retention is a setting a party controls, with event and user-level data held for 2 or 14 months on a standard property, and age, gender and interest data held two months whatever the setting says.

One caution on reading that table. Google's language for data outside the reporting window is that it is not accessible through the interface or the APIs, which is a statement about access rather than about deletion, so I do not assert the data was destroyed.

Where there is no standard for the record to be measured against

There is no published, citable professional standard of care for paid search management. No licensing board, no accrediting body, no professional association with practice standards and disciplinary authority. That absence is a finding rather than a gap to paper over: an opinion that a manager fell below "industry standard" has no published document behind it. Neither does the opinion on the other side.

What holds weight instead is narrower and more durable: the terms the parties actually agreed; the platform's own documentation of what a setting does; and the account's internal inconsistency, where a manager applied a practice in one campaign and not in an identical one, which shows the practice was known and not applied without appealing to any outside benchmark.

Two related absences are worth naming. I have located no court opinion ruling on paid search damages methodology specifically, and none addressing modeled conversions or attribution modeling as an evidentiary question. That is open ground. Separately, Google Ads clicks and invalid clicks hold Media Rating Council accreditation, in a letter dated March 31, 2026. Accreditation means an auditor examined the process against a published standard; it does not create a click-level ledger anyone can inspect.

The demands that come back empty

Some requests cannot be satisfied by anyone, and knowing which ones saves a round of motion practice that produces no document.

• Three years of change history. Google states data older than two years will not be available. Past that line a cooperative party has nothing to produce.
• An API pull to recover an old change. The field-level record is a thirty-day product.
• An API export that matches the interface export. Google's developer documentation states a change event may not include every row from the change history in the web client.
• A preservation letter as a preservation act. Serving one does not change a retention setting or extend a platform window. Exporting does.
• Screenshots standing in for the record. A screenshot preserves a rendering at one moment, without the metadata, the full date range, or any ability to re-segment.
• An account-takeover theory built on the change log. Google states expressly that change history does not track password changes.

The first thing I do on a paid search matter is establish which tier the question belongs to and whether the window on it is still open. That answer is usually quick and occasionally ends the inquiry, which is a good outcome for everyone but the invoice.

Frequently Asked Questions

Published 2026-08-14