Are your Google Ads conversions real? A 6-point audit from one season of real data

Why does this matter before anything else?

Google's bidding optimises toward whatever you call a conversion. If a fifth of those conversions are bots, the algorithm learns to buy more bots — and the dashboard reports that as success. Every decision downstream, including how much to spend, inherits the error.

The six checks

1. Is your cost per conversion close to your cost per real lead?

Define a lead as a person with a name and a number who came from an ad — a form email carrying an ad flag is enough. In our audit the gap was two to four times. That gap is the size of the problem.

2. Is Search Partners or Display switched on?

One day on Search Partners showed a 49.10% click-through rate — 164 clicks from 334 impressions — against 7.8–14.4% on every other day. Real people do not click one ad in two. That single anomaly backed a refund claim, and both networks were switched off permanently.

3. Does every conversion action actually record?

The phone-call goal in this account pointed at a conversion action that did not exist. For seven weeks, every call from an ad was invisible — so the account was simultaneously over-counting fake form fills and under-counting real calls.

4. Are your own keywords bidding against each other?

Four exact- and phrase-match copies of the best broad keyword spent ₹2,674 for zero conversions at roughly twice its cost per click. They were competing with the keyword that earned everything.

5. Are you reading search terms by cost, not impressions?

Sorting by impressions once hid 75% of a week's spend. Sorted by cost, the junk is obvious: wrong neighbourhoods, competitors, the wrong audience. This account now carries 147 negative keywords, all found that way.

6. Is the constraint budget, or ad rank?

This is the expensive one. Doubling the daily budget from ₹1,200 to ₹2,500 bought 17% more clicks at 23% lower cost per click — and 53% fewer conversions per day. Most impressions were being lost to ad rank (58–69%), not budget, so more money bought cheaper, worse traffic. Check impression share lost to rank before raising a budget, and never raise more than 50% in one step.

What did the audit add up to?

The honest part: with attribution broken for most of the season, nobody can say exactly how many bookings the ads produced. What is certain is that the dashboard's version of events was wrong, and the owner spotted the fraud before the data did.