Every Google Ads audit I run starts in the same place, and it is not the campaigns. It is conversion tracking. Smart Bidding, budget pacing and Quality Score are all downstream of one thing: whether the account can see, accurately, which clicks turned into revenue. Get that wrong and everything built on top of it — automated bidding, ROAS targets, reporting — is optimising toward the wrong number with total confidence.
The frustrating part is that broken tracking rarely looks broken. The account keeps spending, conversions keep showing up, and the dashboard looks fine. The faults below are the ones I find most often, roughly in order of how much damage they do.
1. Duplicate conversion actions inflating value
Two conversion actions counted as one
A common pattern: a business tracks an actual sale with a correct, dynamic value, but also tracks a secondary action — a qualified lead, a booking request — with its own fixed value, and both feed into the same reported conversion value. Reported ROAS ends up higher than the real figure, sometimes by a wide margin. It usually happens because the secondary action was added later, by a different person, for a genuine reason — often to give Smart Bidding more signal — and nobody went back to check what it was doing to the blended number.
In Google Ads, break conversion value down by individual conversion action rather than looking at the account total. If two actions are both contributing meaningful value to the same customer journey, you likely have this problem.
The fix is rarely to remove the extra signal — it often still helps Smart Bidding find volume. The fix is to report on sales value and lead value separately, so decisions are made on the real figure rather than the blended one.
2. Tracking that broke quietly, weeks ago
A change elsewhere on the site silently breaks the tag
A CMS update, a cookie consent banner, a new checkout provider, a developer "cleaning up" the header — any of these can stop a conversion tag firing without triggering an error anywhere obvious. Spend continues. Conversions drop to zero, or near enough, and because nobody is watching daily, it can take weeks to notice.
The tell is usually in the trend, not a single number: a sudden, sustained drop in conversions with no corresponding drop in clicks or spend, starting on a specific date. Cross-reference that date against anything that changed on the website around the same time.
This is the most expensive fault on this list, because the algorithm does not just under-report — it actively throttles spend once it stops seeing results, and can take real time to recover once tracking is fixed.
3. Micro-conversions counted as primary conversions
A newsletter signup, treated the same as a sale
Google Ads treats every conversion action as equally important unless told otherwise. A business tracking a purchase and a newsletter signup as two "conversions" of equal weight will see Smart Bidding chase whichever one is easier to get — usually the newsletter signup, since it asks less of the visitor.
This is easy to miss because overall conversion volume looks healthy. It is only when volume is split by action that the pattern shows: plenty of low-value signals, very few high-value ones.
Set primary and secondary conversion actions explicitly, and only include genuinely commercial outcomes — a purchase, a qualified lead, a booked call — in whatever the bid strategy is optimising toward.
4. Phone calls that are not tracked at all
The majority of enquiries, invisible to the account
This is especially common in home services and other local trades, where most enquiries are phone calls rather than form fills. Without call tracking, the account has no idea those conversions happened, and optimises entirely around the minority of leads that came through a form.
The result is bidding that looks rational but is working from a fraction of the real picture — usually underinvesting in exactly the campaigns producing the most calls.
Call tracking, ideally with recording, closes this gap and is one of the highest-impact single changes available to a business whose customers mostly ring rather than click "submit."
5. An attribution model nobody agreed to
The account is optimising to a definition of success no one chose
Google Ads defaults to a data-driven attribution model, which can differ meaningfully from how a business actually thinks about a sale — particularly for longer sales cycles, where an enquiry today might not close for weeks. If GA4, the CRM and Google Ads all define "conversion" differently, the reported numbers from each will disagree, and decisions get made on whichever number happened to be open on screen.
The tell is usually a business asking why three different reports show three different numbers for the same period, with no one able to say confidently which is correct.
This is not usually a technical fault so much as a decision that was never made deliberately. Pick the definition of a conversion that matches how the business actually recognises a result, apply it consistently across every system, and treat the other numbers as supporting context rather than competing truths.
What this costs, in practice
None of these faults are exotic. They are common enough that I would be surprised to find an account with none of them. The cost is rarely obvious in the moment — it shows up as a ROAS that looks fine but is not real, a bidding strategy that cannot find its footing, or a budget that keeps getting throttled for reasons nobody can quite explain.
The starting point is always the same: before changing a bid, a keyword or a budget, confirm the account is measuring the right thing. Everything else depends on it.
