First, confirm the drop is real
Before diagnosing anything, make sure the fall exists in the data rather than in the way the report is framed. Compare like with like: the same number of whole days, the same days of the week, and a comparison period that is not itself unusual. A comparison against a holiday week or a sale will show a fall that is really the end of a peak. Remove any filters left on from earlier analysis, and check that the view is not limited to one network or one device.
Then segment. Split the campaign by network, device and day. A fall concentrated on search partners, or on one device, points somewhere quite different from a fall that is spread evenly across everything. Note the first day the line broke, because that date is the key you will use in change history. If the campaign had no impressions at all from the start, or a status column shows something blocked, the ads not showing diagnosis covers paused levels, disapprovals and targeting exclusions; this page assumes the campaign was healthy and then declined.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Impressions break on a specific day with no outside event | Change history filtered to the campaign around that date | A budget, bid strategy, target, keyword or targeting edit | Reverse or soften the edit, or accept it if it was intended |
| Ads stop earlier in the day than they used to | Lost impression share (budget); status "limited by budget" | Budget reduced, or the same budget now funds more | Restore the budget or narrow what it pays for |
| Impressions slide gradually after a target change | Bid strategy target history; bid strategy status | Target CPA or ROAS tightened, so automated bids fell | Loosen the target in small steps toward recent actuals |
| Lost share attributed to rank climbs while budget loss is flat | Lost impression share (rank); auction insights overlap and outranking | Competitors winning auctions the campaign used to win | Improve ad and landing page quality or bid more on the terms that pay |
| Impression share is steady but impressions fall | Impression share trend; keyword planner trend; prior-year data | Fewer people searching: season or market shift | Adjust expectations and budgets; widen coverage only if economics allow |
| Particular queries vanish from the search terms report | Negative keyword lists and recent additions | A new negative or shared list blocking wanted queries | Remove or narrow the conflicting negative |
Fix, verify and prevent each cause
An edit in change history
Change history records who changed what and when, at every level. When the impression line breaks on a particular day, the explanation is very often an edit made the same day or the day before: a budget lowered, a bid strategy switched, a target moved, keywords paused, a location removed.
Fix: decide whether the edit was meant to reduce volume. If it was, the drop is the intended result; if it was a mistake or had a side effect nobody expected, undo it or replace it with a smaller version.
Verify: impressions return toward the baseline over the next several full days, and nothing else changed in that window to muddy the reading.
Prevent: record the reason for each material edit, and keep one owner per campaign as described in the account management hub, which treats change history as a management tool rather than an audit trail read only after something breaks. When several edits fall in the same window, the change history audit tutorial classifies them and lines them up against the daily results to find the one that caused the fall.
Budget reduced or budget loss rising
The help centre documents that once the daily budget is exhausted the ads stop serving until the next day, and that daily spend can vary around the average. A budget cut shows up as ads stopping earlier and as lost impression share attributed to budget rising. The same pattern appears without any budget edit when the campaign gains keywords, locations or a higher bid that make each day's money run out sooner.
Fix: restore the budget where the returns justify it, or reduce what it funds so the remaining money covers the hours that matter.
Verify: lost share due to budget falls and the hourly report shows delivery through the day.
Prevent: review pacing on a fixed day each week using the method in budgets and pacing.
Target tightened
With target CPA or target ROAS, the target governs how hard the system bids. A stricter target makes it skip auctions it judges too expensive, and impressions fall even though no keyword or budget changed. The help centre describes a learning period after significant changes, so the fall may arrive gradually rather than overnight.
Fix: compare the target with the campaign's recent actual cost per acquisition or return, and move it back toward that figure in modest steps, waiting between each.
Verify: the bid strategy status settles and impressions recover alongside conversions rather than at their expense.
Prevent: set targets from margin and recent performance, never from a wish, and follow the change discipline in the Smart Bidding guide.
Losing Ad Rank to competitors
When lost impression share attributed to rank rises while budget loss stays flat, someone else is winning the auctions. Auction insights shows which domains overlap with you, how often they rank above you, and whether a new name has appeared. Ad Rank depends on the bid, ad and landing page quality and the expected effect of assets, so a competitor can overtake you by improving quality as well as by bidding more.
Fix: decide which terms are worth defending, improve relevance and landing page experience there, and raise bids only where the extra CPCAverage cost of each click. still pays.
Verify: rank-attributed loss falls on those terms within a couple of weeks.
Prevent: read auction insights monthly and work on the components described in the Quality Score guide; the auction insights tutorial sets out that monthly reading step by step. If the main effect of the competition is a higher price per click rather than fewer impressions, the rising CPC diagnosis is the better next page.
Search demand fell
Sometimes nothing in the account or the auction changed: fewer people are searching. The signature is a fall in impressions with impression share roughly unchanged, because you are taking the same slice of a smaller pie. Seasonal categories, a news cycle ending, or a product being replaced all produce it.
Fix: confirm the shape against the same period in a previous year and against keyword planner trends, then adjust budgets and expectations rather than bids. If the business needs more volume, widen coverage with additional phrasing, following the match-type choices in the match types guide.
Verify: share stays stable and the seasonal pattern matches the prior year.
Prevent: keep a demand calendar for the account so the annual dip is expected rather than diagnosed every time.
A new negative over-blocking
A negative keyword added to clean up traffic can quietly block wanted queries, and a shared negative list applies the damage to every campaign that uses it at once. The help centre notes that negatives do not match close variants in the way positive keywords do, so a broad negative may still block more than intended.
Fix: find negatives added around the date of the fall, test core queries in the Ad Preview and Diagnosis tool, and remove or narrow the term that conflicts.
Verify: the blocked queries reappear in the search terms report.
Prevent: check new negatives against active keywords before attaching a list, using the conflict review in negative keywords.
Prevention: change logs and impression-share alerts
The causes above share a pattern: they are either introduced on a known day or detectable in a column that already exists. A prevention routine turns that into an early warning instead of a post-mortem. Keep a short change log outside the account that records each material edit with its reason and its owner, so that anyone reading a chart can match a break to a decision. Set an automated rule or a scheduled report that flags a campaign whose impressions fall well below their usual weekly level, and pair it with the impression share columns so the alert arrives already split into budget and rank. Review auction insights once a month for the campaigns that carry most of the spend. Maintain a demand calendar so seasonal troughs are labelled in advance. And make one change at a time on any campaign that matters, waiting long enough to read its effect, because two simultaneous edits turn a five-minute diagnosis into a guessing game.
An accountancy firm's impressions halve in a week
A small accountancy firm runs a search campaign that normally records about 1,800 impressions a week. In the second week of May it records 860. Change history shows one edit: on 6 May the target CPA was lowered from 60 to 42 after a good month. Lost impression share due to budget is 3% and unchanged; lost share due to rank has risen from 18% to 51%. Auction insights shows no new competitor. The firm also checks demand: the prior year shows a post-deadline dip, but only of around 15%. The target change explains most of the fall. Moving the target to 52, then to 56 two weeks later, brings weekly impressions back to around 1,600 while cost per acquisition stays below the old target.
Impression share describes coverage of your own eligible auctions rather than your presence in the market, a distinction the impression share versus share of voice comparison draws out.
Common mistakes
Comparing a quiet week with a peak and calling the difference a drop. Reacting to the drop with a budget increase when the budget was never the constraint. Reading auction insights before change history, and blaming a competitor for an edit made in-house. Removing an entire shared negative list to find the one term that over-blocked.
