Measurement mistakes are the most expensive ones because you cannot see them: the account looks like it is working, the reports produce figures, Smart Bidding bids with confidence... towards something that is not a sale. These are the ones we have found most often, and in every case the client was convinced their measurement was correct.
Mistake 1 · Everything is a primary conversion
Case: a software platform (SaaS) with 9 conversion actions, all of them primary: sign-up, demo requested, click on "pricing", click on "contact", 75% scroll, video viewed, PDF download, chat opened, newsletter. The account "converted" 400 times a month at €6 each. Real demos: 11. Smart Bidding had learned that the ideal customer was somebody who scrolls.
How to spot it: Goals → Conversions: count the primary ones. More than two or three per objective is suspicious. Segmenting conversions by action (Module 7) exposes the split.
How to fix it: primary only for what you would pay to get (a demo, a qualified sign-up); everything else, secondary. And accept that the real CPA will go from €6 to €200: nothing has got worse, it is just true now.
Mistake 2 · Navigation conversions inflate the value (L1)
Case: a B2B services company measured "see phone number" and "click on email" as primary, with the same value as the form. They were 70% of the conversions. When they were moved to secondary, the campaign went into learning, "conversions" fell by 70%... and real form fills went up by 30% within two months, because the bids stopped chasing browsers. Side effect observed: CTR dipped slightly (fewer clicks from window-shoppers), which frightened the client until they saw the leads.
How to spot it: actions of the "click on..." or "visit to..." type set as primary; the share of value those actions represent of the total.
How to fix it: the L1/L2 rule — navigation actions (L1) as secondary or, if you keep them, with a value that adds up to less than half the total. Change it gradually (10-20% of the weight at a time) so you do not reset learning all at once.
Mistake 3 · Duplicated conversions
Case: a shop with the conversion tag placed by its CMS plugin and by Tag Manager. Every purchase counted twice. The target ROAS was set at 6 on duplicated data: the real one was 3, below break-even. They had been losing money for four months with a report that said the opposite.
How to spot it: suspiciously even conversion counts; Tag Assistant shows the event firing twice; the action's diagnostics warns about a duplicated tag; orders in the shop ≈ half the conversions.
How to fix it: a single tag (the Tag Manager one), exclude the bad-data period in Smart Bidding and recalculate the target on clean data.
Mistake 4 · The thank-you page that vanished
Case: a home renovation company redesigned its website in August. The new
form did not redirect to /thank-you. Conversions at zero from day 3. Nobody
looked at Google Ads until September: Smart Bidding, with no data, had lowered
bids until the campaign was almost switched off, and after that the target CPA
"did not work".
How to spot it: conversions at zero with normal clicks; a "no recent conversions" status; an end-to-end test would have caught it on day one.
How to fix it: fix the tag, exclude the period in the bidding strategy (so it does not learn from the hole) and wait for learning. And the lesson: measurement is checked weekly (Module 6).
Mistake 5 · Judging with the wrong column
Case: a monthly report for the board using the "All conversions" column (primary + secondary + 3-second phone calls). Budget decisions taken for a year on a figure that was triple the real one.
How to fix it: "Conversions" to judge; "All conv." to diagnose. And in any report that leaves the account, the definition of what is being counted, in writing.
The five-minute measurement audit
- How many primary actions there are and what they are → Mistakes 1 and 2.
- Even conversion counts, or double what the CRM says → Mistake 3.
- Actions at zero or "no recent conversions" → Mistake 4.
- Which column the reports doing the rounds use → Mistake 5.
💡 Ninja trick: these five mistakes are the whole reason the L1/L2 methodology, the Agent's %L1 calculation and Lead Rating exist: so that what reaches Google is true conversions with their real value. When measurement is honest, Smart Bidding is a formidable ally; when it is not, it is a formidable ally of the lie.
What you should remember
- Primary = what you would pay to get; navigation is secondary.
- Navigation conversions never more than half of the value.
- One tag; even conversion counts are an alarm.
- Conversions at zero with normal clicks = broken measurement, not the market.
- Judge with "Conversions", not with "All conv.".