First, split the fall into clicks and conversion rate, and check it against the business's own records
Conversions are clicks multiplied by the conversion rate (CVRShare of visits or clicks that convert.), so a fall has to come from one side or both. Put the affected campaigns' clicks, conversions, conversion rate and cost per conversion side by side for the days since the fall and an equivalent stretch before it, with whole weeks on each side so the days of the week match. If clicks fell and the rate held, the problem sits upstream of the click, and the impressions dropped diagnosis or the low click-through rate diagnosis is the place to start. If clicks held and the rate fell, the problem sits after the click: in what is counted, in who is clicking, or on the page. The conversion rate calculator and the CPA calculator give comparable rates when the reports export raw counts.
Then compare with the business. The change history tutorial makes the same comparison before it reads a single edit: if orders, bookings or leads held steady in the shop system or CRM while reported conversions fell, the measurement broke; if they fell too, something real happened to the traffic or the site. Read the account by conversion time for this comparison, because business records are dated by when the order arrived, while the standard columns date each conversion by its click, as the conversion windows and attribution guide explains.
Finally, decide whether the fall is a signal at all. A metric moves every week for reasons nobody controls, and the control charts framework separates that routine variation from a real shift by building limits from the metric's own history. If the fall sits inside the ordinary spread, the right response is to watch, not to act. If conversions fell in every channel at once, not only in Google Ads, the diagnosis for results down across all channels is the better starting point.
| Business records | Google Ads clicks | Google Ads conversion rate | Where to look first |
|---|---|---|---|
| Held | Held | Fell | Conversion settings, open windows, tracking |
| Fell | Fell | Held | Impressions and click-through |
| Fell | Held | Fell | Bidding, search mix, landing page, price and stock |
| Fell in every channel | Any | Any | The cross-channel diagnosis, then the site |
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Conversions fell with clicks; conversion rate steady | Impressions, impression share and click-through for the same days | Fewer clicks: impressions or click-through fell first | Follow the impressions or click-through diagnosis |
| Conversions fell on the day of an edit to conversion actions or goals; orders steady | Change history for conversion actions and goals; primary and secondary status; counting | Primary conversion actions, goals or counting changed | Restore the intended primary action, or reset targets on the new basis |
| The latest days look worst; older days recover in later reports | Conversions by conversion time; window and model settings in change history | Recent days still filling, or a new window or attribution model, read as a fall | Judge only closed periods; treat a settings change as a break in the series |
| Conversions fell after a release, container publish or banner change; orders steady | Conversion action status; tag preview; consent state; release log | Tracking partly broken by a site release, tag or consent change | Repair the tag, trigger or consent setup through the tracking diagnosis |
| Conversions fell after a strategy, target or budget edit; status reads learning | Bid strategy status; change history for bidding edits | A bid strategy, target or budget change | Follow the bid strategy diagnosis; avoid further edits while it settles |
| Clicks steady, conversion rate falling, new queries in the report | Search terms report and n-grams before and after the fall | The search mix shifted towards queries that rarely convert | Add negatives, tighten matching, give converting themes their own keywords |
| Conversion rate fell across every campaign and source at once; orders down too | Site analytics by landing page and step; price, stock and offer changes | The landing page, offer, price or stock changed after the click | Diagnose on the conversion rate page; reverse or explain the change |
Fix, verify and prevent each cause
Fewer clicks: impressions or click-through fell first
A fall that starts before the click is not a conversion problem, even if conversions are where it was noticed. Fewer impressions, a lower share of the auction, or a weaker click-through rate on steady impressions all leave fewer people reaching the page while the page converts as before.
Fix: follow the impressions dropped diagnosis, which reads change history, lost impression share by budget and rank, competitors and demand in order, or the click-through diagnosis when impressions held and clicks did not.
Verify: clicks return to their earlier level and conversions follow without any change to the conversion settings.
Prevent: read clicks and conversions together in every review, so a fall in volume is caught as a traffic problem rather than misread as a conversion problem.
Primary conversion actions, goals or counting changed
The conversion action clean-up tutorial explains the unit: a Conversion ActionA Google Ads definition of one action to count, such as a purchase or lead, with its source, category, counting method, value and conversion window. is primary or secondary, primary actions fill the Conversions column and are what bidding optimises towards, and secondary actions appear only in All conversions. Moving an action from primary to secondary, making a different action primary, switching a campaign to a custom goal or changing counting from every to one changes what the Conversions column counts and what the cost per conversion (CPACost of each conversion.) divides by, with no customer behaving differently. Google's page on conversion tracking describes the action's settings; the tutorial is the site's account of how they interact. A GA4 import made primary in place of the native tag also changes the count, because, as the comparison of Google Ads and GA4 figures sets out, the two systems attribute and date conversions differently.
Fix: read change history for edits to conversion actions and goals in the days before the fall; if the change was a mistake, restore the intended primary action; if it was deliberate, accept the new basis and reset targets from figures counted under it.
Verify: All conversions, which still holds the secondary actions, shows the same activity as before, and primary conversions move with orders again.
Prevent: run every change to primary actions through the conversion goal change checklist, and warn whoever reads the reports before it goes live.
Recent days still filling, or a new window or model
Standard columns place each conversion on the date of its click, so recent days keep filling in until the conversion window closes; the account's Conversion LagThe time between an ad interaction and the conversion it leads to, which leaves the most recent reporting days incomplete. decides how long that takes. The windows guide also explains the two settings changes that look like a fall: a shorter window lowers totals, because conversions that arrive after it closes are no longer credited, while a new AttributionAssigning conversion credit to touchpoints. model mainly moves credit between campaigns, so one campaign can lose conversions while the account total barely moves.
Fix: read the period by conversion time, or leave out the days whose window is still open; check change history for window or model edits and, where one was made, treat its date as a break in the series.
Verify: the apparent fall shrinks or disappears in later reports, or the account total holds while credit has moved between campaigns.
Prevent: compare only closed periods with closed periods, change one setting at a time, and note every window or model change in the change history audit record.
Tracking partly broken by a site release, tag or consent change
A tracking fault rarely zeroes conversions; it more often cuts some of them. A consent banner that withholds the granted state, a Tag Manager trigger tied to an address that moved, or a template release that dropped the tag from one page type all reduce recorded conversions while orders carry on. The conversions not tracking diagnosis walks the chain from tag to container to Consent ModeA Google tag feature that adjusts what tags collect according to the visitor's consent choices. to click identifier to import, and the website migrations guide covers the larger version of the same break.
Fix: follow that chain from the first link, and protect bidding while the fix is made using the tracking outage response SOP.
Verify: the conversion action status shows recent activity, preview shows the tag firing in the granted state, and reported conversions move with orders again.
Prevent: check conversion action statuses and compare conversions with the business's records on the day after every release, container publish and banner change.
A bid strategy, target or budget change
The Smart Bidding strategies guide records that Google lists changing the strategy type, changing conversion actions and large changes to targets or budgets among the changes that can start a learning period, during which performance can fluctuate and the status column shows Learning. A target tightened beyond what the campaign has achieved can also make the strategy sit out auctions it used to win.
Fix: follow the diagnosis of results that dropped after a bid strategy change, which separates learning, an over-tight target, a capped budget and thin data.
Verify: the status settles and conversions on closed days return towards their earlier level.
Prevent: change targets in small steps, one setting at a time, and avoid bidding edits in the same week as a conversion clean-up.
The search mix shifted towards queries that rarely convert
When clicks hold and the rate falls, the people clicking may have changed. A broader keyword, a new match type, a removed negative list or a new competitor term can bring searches that click and do not buy. The Search Terms ReportThe Google Ads report of the actual searches that triggered an account's ads. shows which searches triggered the ads, and the search term n-gram analysis finds the recurring words that carry the change across many cheap queries.
Fix: add negatives for the themes that do not convert, narrow the match types that collect them, and give converting themes their own keywords; the irrelevant search terms diagnosis traces why an unwanted search matched in the first place.
Verify: the share of clicks from converting themes rises and the conversion rate recovers on closed days.
Prevent: review search terms weekly for the campaigns that carry most of the spend, and compare n-grams before and after any keyword or match type change.
The landing page, offer, price or stock changed after the click
If the conversion rate fell in every campaign and every source at once and orders fell with it, the change is probably on the site. A price rise, a best seller out of stock, a new delivery charge, an ended promotion or a broken form step changes what every visitor does after the click. The conversion rate dropped suddenly diagnosis owns these causes and splits the drop by device, browser, source, landing page and funnel step.
Fix: diagnose and repair on that page, then reverse, soften or explain the commercial change.
Verify: the site's step rates and the Google Ads conversion rate recover together.
Prevent: ask for advance notice of price, stock and offer changes and record them in the same dated Change LogA dated record of changes made to tracking, accounts, reports, lists or rules, with who made each and why, so later movements in the figures can be traced. as account edits and site releases.
Prevention: a weekly conversion check that knows what changed
Most of the causes above were visible before anyone noticed the fall: an edit in change history, a release in the web team's log, a consent banner update, a new set of queries. A weekly check that puts the figures beside the changes catches them early and stops the wrong fix being made.
| Weekly check | What it compares | What it prevents |
|---|---|---|
| Conversions by conversion time against orders or leads | The account against the business's own records | A tracking or settings change mistaken for lost sales |
| Clicks, conversion rate and cost per conversion on closed days | This week against the metric's own control limits | Reacting to routine variation, or missing a real shift |
| Change history for conversion actions, goals, windows, models and bidding | Edits against the dates the figures moved | A clean-up or target change read as a market change |
| Conversion action statuses after any release or banner change | Recording actions against expected ones | Partial tracking loss running for weeks |
| Search terms and n-grams for the largest campaigns | New themes against converting ones | A drift towards searches that click and do not buy |
| Price, stock and offer changes from the business | Commercial changes against conversion rate | A site-side fall blamed on the ads |
Run it in that order. Measurement comes first, because every later reading depends on it: a conversion rate calculated on a broken count says nothing about the searches or the page. Then the account's own edits, because they are dated and the quickest to confirm or rule out. Then the traffic and the site. Keep one Change LogA dated record of changes made to tracking, accounts, reports, lists or rules, with who made each and why, so later movements in the figures can be traced. that holds account edits, site releases, tag and consent changes and commercial changes together, with dates and owners, so the weekly reader can see everything that moved, not only what the account records. When a shift is confirmed, annotate its date on the control chart and recompute the limits from the new level, as the framework advises, so next week's reading is made against the process that exists now.
Two habits make the check reliable. First, keep the business's own figures beside the account's in the same report, so the comparison takes seconds rather than an export. Second, agree who is told before any change to conversion actions, windows, models or bidding goes live, so that a planned measurement change never arrives as a mystery in the next report. The collection of Google Ads troubleshooting pages covers the neighbouring symptoms when the check points elsewhere.
A bicycle parts shop traces a fall to its own clean-up (illustrative)
Cuvrellan Cycle Parts
Cuvrellan Cycle Parts, an invented online bicycle parts shop, sees Google Ads purchases fall from 212 one week to 131 the next, while clicks hold at around 4,800 and its own order system records 236 web orders against 241 the week before. Cost per purchase in the account rises from £19 to £31, and the manager's first instinct is to loosen the target. Change history shows a conversion clean-up on the previous Monday: the tag-based purchase action was set to secondary and an imported GA4 purchase made primary. All conversions still shows the tag-based action recording 208 purchases for the week. The two most recent days are also still filling: by conversion time the week reads 149, not 131. The shop restores the tag-based action as primary and keeps the import as secondary for comparison, notes the change and its date in the change log, and leaves the bid strategy and target untouched. Over the following three weeks, read by conversion time against the order system, primary purchases sit between 205 and 219 a week while web orders run between 229 and 244.
Common mistakes
- Reading the latest days as a fall. Standard columns date conversions by click, so recent days are still filling in; read by conversion time or leave them out.
- Skipping the business's own records. Orders that held while conversions fell point at measurement, not at the market; one comparison saves a week of the wrong fixes.
- Changing the bid strategy first. A new strategy or target adds a learning period on top of the cause and makes both unreadable.
- Treating a clean-up as a loss. A deliberate change to primary actions changes the count; reset targets on the new basis instead of chasing the old figures.
- Blaming the ads for a site change. A fall across every source at once usually lives on the page, the price or the stock.
- Reacting to routine variation. A dip inside the metric's ordinary spread needs watching, not a fix.
