A paid-ads account rarely fails in one visible way. It leaks through ten specific, checkable signals: budget pacing drift, creative fatigue, CPA creep, ROAS below target, frequency overload, audience overlap, Quality Score erosion, impression share loss, silent disapprovals and attribution mismatch. Three or more flagged on one account is worth acting on now.
Why do these signals go unnoticed in the first place?
They go unnoticed because checking for them is still a manual habit, not a system. According to ppc.land, about 87% of agencies still pace budgets by hand, and Fluency's research found that 71% of ad-ops teams say manual work puts campaigns at risk. Both numbers describe the same gap: a strategist who is meant to catch these ten signals is also the person running the client call, writing the report and covering three other accounts.
None of the ten signals below need a dashboard to notice. They need someone to look, on a schedule, before the pattern compounds. That is the part that quietly stops happening once an agency grows past the point where one strategist can hold every account in their head.
The signals are not hard to spot; the difficulty is finding the hour each week to look for them across every account, on time.
What are the ten signals a paid-ads account is losing money?
Each signal below is a specific, checkable condition, not a vague warning. Run through the account against each one and mark it clear or flagged.
1. Budget pacing drift
Pacing drift happens because spend is usually reviewed weekly, not daily, and automated bidding can front-load a budget when early response looks strong. Nobody notices the account is running hot until the Monday report, by which point it has overspent against the plan for three or four days straight. It costs the agency an awkward budget conversation with the client and, on the other side of the same coin, underspend that quietly leaves delivery on the table. Spot it by comparing day-by-day spend to the pro-rated daily target: a gap of 10% or more before Thursday is the point worth flagging.
2. Creative fatigue
Fatigue sets in because a strategist managing a dozen accounts refreshes creative on the ones asking loudest for it, and the quiet, steady account gets left running the same ad for months. Click-through rate falls as the audience sees the same creative too many times, and every extra day it runs without rotation adds roughly 8 to 15% to the cost per click. Spot it by watching CTR across three or more consecutive days: a falling trend with no creative change behind it is the flag.
3. CPA creep
Creep happens as the platform's algorithm keeps optimising toward whichever audience segment is currently cheapest to reach, and that segment shifts as the initial, best-converting audience saturates. The account keeps spending at the same rate for a worse cost per acquisition, and the agency bills the same management fee for a weaker result. Spot it by tracking CPA week on week: a rise with no bid or budget change logged behind it is the signal.
4. ROAS below target
This one is less about the metric and more about what happens after it drops: nothing. Most accounts have no rule that forces an escalation when return on ad spend falls below the number the client agreed to. The account keeps running, the client keeps paying full price, and the first person to notice is usually the client, in a review call the agency did not see coming. Spot it by checking whether ROAS has sat below target for seven days or more with no escalation logged anywhere.
5. Frequency overload
Frequency climbs when an audience is narrower than the budget it is being asked to absorb, or when a campaign runs long without the audience being expanded or refreshed. The same people see the ad again and again, CPMs rise because the platform has to work harder to keep finding room for the same impression, and conversion rate falls as the audience goes ad-blind. Spot it on Meta specifically: a frequency above 3.5 with no creative refresh or audience expansion behind it is the threshold worth acting on.
6. Audience overlap
Overlap creeps in when a campaign is duplicated to test a variable, and the new ad set is built without checking whether it targets people already reached by another ad set inside that same campaign. The two ad sets end up bidding against each other in the same auction, which inflates the campaign's own CPMs for no reason. Spot it with the platform's own audience overlap tool, or by comparing targeting parameters across ad sets within the same campaign by hand when that tool is not available.
7. Quality Score erosion
Quality Score drifts down as an account grows unevenly: new keywords get added faster than ad copy is written to match them, or a competitor simply improves their own relevance and resets the curve everyone is graded on. Because Google's ad rank formula multiplies bid by Quality Score, a lower score means paying more for the same position, not a smaller effect but a direct tax on every click. Spot it by checking Quality Score on the account's top-spend keywords: below 6 is worth flagging.
8. Impression share loss
Share is lost passively, not dramatically: a competitor raises their bid on the client's own brand terms, or launches a new campaign targeting them, and the account does not respond because nobody is watching branded search specifically. The cost is a strange one: competitors capture clicks, and sometimes customers, using the client's own brand name, on the exact terms that should have been the cheapest, highest-converting traffic in the account. Spot it by checking Search Impression Share on branded terms: below 70% is the flag.
9. Silent disapprovals
Disapprovals go silent because platform notifications sit in an inbox nobody checks daily, and an account with many active campaigns can have one or two ads pulled without the overall spend number moving enough to draw attention. The campaign keeps running at reduced reach, or stops entirely, while the client is billed the same management fee for an account that is quietly doing less than it should. Spot it by checking the disapproved and limited status across every active ad, not only the top spenders.
10. Attribution mismatch
Mismatch is structural, not accidental: Meta counts a conversion on a 28-day click window, Google on 30, and most in-house dashboards default to 7. Three systems, three different counts, describing the same account. Nobody chose this on purpose, but nobody reconciled it either, so the client sees three numbers that do not add up and starts to doubt all three. Spot it by pulling the same date range from all three sources side by side and looking at the size of the gap.
None of these ten signals require new software to check; they require the same ten minutes spent on every account, on a schedule that does not slip.
How many flagged signals mean the account is actually at risk?
One flagged signal, on its own, is rarely urgent. Healthy accounts trip a single signal from time to time: a Quality Score dip after a keyword expansion, a frequency spike during a short flight, a pacing gap in a genuinely slow week. Three or more flagged signals on the same account, at the same time, is a different pattern. That is not one unlucky week, it is an account that has gone unchecked for long enough that several small drifts compounded together.
Treat one flag as a watch item and three or more as the trigger to act, because the risk is in the pattern, not in any single number.
What I saw building the ad-ops agent
When I designed, built and deployed an ad operations agent for a paid-ads agency, the thing that surfaced was not a technical limit, it was how much of this checking lived in one strategist's memory rather than in any process. The strategist who caught frequency creeping past 3.5 caught it because they had seen the same pattern before on a different account, not because a dashboard told them. That kind of pattern recognition does not scale by hiring more strategists; it scales by teaching a system to hold the checks a good strategist runs without thinking, so the account gets watched on a Friday afternoon the same way it gets watched on a Monday morning.
You can read the full mechanism, including how the agent keeps the money-affecting decisions deterministic or human, in the ad operations agent case study.
Common questions
Do these signals apply to both Google Ads and Meta accounts?
Most of them, yes. Quality Score erosion and impression share loss are Google-specific, and frequency overload is a Meta metric, but budget pacing drift, creative fatigue, CPA creep, ROAS below target, audience overlap, silent disapprovals and attribution mismatch apply to any paid channel with a reporting dashboard.
How often should an agency run this check?
Monthly, at minimum, on every active account, and weekly on any account already carrying a flag from the month before. A quarterly cadence is long enough for pacing drift or creative fatigue to compound into a client conversation nobody wanted to have.
What if only one signal is flagged?
Log it and watch it at the next check. A single flag is common on accounts that are otherwise healthy, and treating every flag as urgent trains a team to stop trusting the checklist.
Does a flagged signal mean the strategist made a mistake?
No. Most of these ten signals are the predictable result of one person holding too many accounts in their head, not a judgement error. The fix is rarely "check harder," it is building a way to check that does not depend on nobody being busy that week.
Download the printable checklist version of this audit to run against a client account in about 20 minutes, or leave it with a client directly. The ad-operations build below is what most agencies reach for once running it by hand, every month, on every account, stops being realistic.