The conversion action that teaches Google to buy your worst lead

Most underperforming Google Ads accounts are not badly managed. They are precisely managed toward the wrong definition.

Here is the mechanism.

What the account is actually being told

Automated bidding takes the conversion actions you marked as primary and works to get more of them for the budget. If those actions all count as one conversion each, it has been told that every one of them is worth the same.

So it looks across your keywords, your placements and your audiences, and it finds the ones that produce conversions cheapest. A newsletter signup is cheaper than a quote request. A quote request is cheaper than a booked job. Within a few weeks, the money moves toward the cheap end.

The account gets more efficient every month on a number that has stopped meaning anything.

Check it in five minutes

Open Google Ads. Goals, then Conversions, then Summary.

Look at three columns.

Conversion action. Count the rows marked primary. If there is more than one, and the second one is a phone-click or a page view or a scroll depth, that is being bid on with the same weight as your real one.

Value. If it reads "no value" or shows the same number on every row, the bidding is running on a count.

Include in "Conversions". This is the switch that decides whether an action feeds bidding at all. Secondary actions get reported and not bid on. Plenty of accounts have this backwards.

Then look at the conversion count against the number of jobs you actually booked last month. If the account says 240 and you booked 11, you already know what it has been buying.

What a fix looks like

Give each action a value that reflects what it is worth to you.

You do not need perfect numbers. You need the ranking to be right and the gaps to be roughly the right size. If a booked job averages $4,000 and closes about 1 in 4 quote requests, a quote request is worth around $1,000 and a newsletter signup is worth a small fraction of that. Put those numbers in. The bidding will re-sort itself around them within a few weeks.

Then close the loop. When a job actually closes, that outcome should get back into the account, so the bidding learns from revenue rather than from the form. That is an offline conversion import or a server-side event, and it is the part most accounts never build.

Where this stops working

Two places, and you should know both before anyone sells you the fix.

If your lead volume is low, values will be noisy. Under roughly 30 conversions a month, an account does not have enough signal for value-based bidding to sort anything reliably. Below that line, structure and negatives do more than bid strategy will.

If your close rate is unknown, the values are a guess. We can rebuild what the account counts and what it feeds back. We cannot guarantee your close rate or your deal size, and any value model is built on top of numbers that are yours.

That is the honest limit. The measurement gets fixed either way, and the measurement is usually where the money was.

The one number to write down

Take last month's ad spend, divide it by the number of jobs you actually booked from ads, and write it on a sticky note.

That is your real cost per job. It is almost never the number in the platform, and it is the only one worth optimizing.

• Google Ads
• tracking