Help · How Adintel decides
How the audit decides
Checks decide every finding by rule, most money lines come from the account itself, findings rank by severity, and the nightly check says what changed and why.
Reviewed 24 October 2026
Who decides a finding?
The checks do, by rule, over the account's last 90 days: whether there is a finding, its severity, whether it is a quick win or strategic, which campaigns and figures it rests on, and its estimated impact. The nightly check is written in Adintel's own plain wording. When you write a report, and for the first report after you connect an account, a language model rewrites each finding's title, description and recommendation in plain words, working from the check's facts, and may not change those facts. Any finding the model fails to rewrite keeps the check's own wording.
Where do the thresholds come from?
Where a threshold is about money, it comes from the account itself, not from a number fixed for everyone; the one fixed amount is the $40 a region must have spent before it is judged. Something should have converted by now when its spend would have bought at least three conversions at the cost per conversion of the rest of its campaign (or of the account, when the campaign has too few conversions), and it recorded none. Waste is judged among the largest spenders, where it matters. Comparing ROAS between ad groups, asset groups or products runs only on an account that tracks real conversion value, and a check that depends on whether the business sells online or generates leads runs only once its campaigns and conversion actions make that clear.
How is a region judged?
Within each campaign, from Google's report of where its spend and conversions were. A region where people were located is judged when it took at least a tenth of the campaign's spend over the 90 days and at least $40, in the account's currency at the latest stored exchange rate. It is a finding when it converted far less than its spend would have bought at what the campaign paid per conversion for people located elsewhere: so far below that chance would explain it less than one time in a hundred, even allowing for chance in what the rest of the campaign converted. A campaign bidding for value is judged on conversions, its return on ad spend beside them, because order sizes vary too much between regions to judge value that way. People reached through their interest in a place are reported apart. Where a campaign takes them, Google's default, they are a finding only when they convert worse than the people located there. The finding names the lever: a lower location bid adjustment or an exclusion with manual bidding, an exclusion with Smart Bidding or Performance Max, which do not use location bid adjustments. Spend Google places in no location is named and never given to a region.
How are findings ordered?
By severity first, then estimated impact: a critical finding is never ranked below a lesser one, however large the lesser one's figure. Quick wins, findings you could act on today, also get their own list at the top; strategic ones need planning, such as a restructure or a bidding change. All findings follow, by area.
When does a check not run?
When the account has too little data for it, such as too few days, too little spend or no conversions. The region checks also wait until Google's region figures for the whole 90 days have been read. A check that does not run is skipped and listed under Checks not run on this account, with its reason, never reported as a pass. Areas no audit can see are listed under Not assessed in this audit. An audit also needs a complete sync from the last 48 hours; without one it says so and does not run, rather than judging an incomplete copy of the account.
How does the nightly check compare two nights?
Each day, after the morning read, every monitored account that has synced again is checked and compared with its last nightly check, subject by subject: a keyword, an ad group, a campaign. A finding's subjects are compared only when both checks ran the same version of its check in full, so a change to a check is never read as a change to the account. What is new, no longer reported or now critical is stored with the first of these causes that applies: Fixed from Adintel, A change in the account that bears on it, Adintel corrected its copy, Business context edited, Crossed the check's line, and No change recorded. Nightly checks are kept 35 days; the record of what changed between them, 400 days.
Still need a hand? Email support@adintel.io and a person replies.