Diagnosing a sudden drop in conversions.

I helped diagnose a sudden drop in conversions in a Google Ads campaign. The campaign normally produced about 25 conversions (leads) a day. Conversions dropped to 1 or 2 a day and hadn't recovered.

Campaign setup.

The campaign used the Target CPA (cost per action a.k.a. cost per lead) bid strategy. The target cost per lead was $50 and the actual cost per lead was about $43.

Leads came in as:-

• Calls from a click-to-call button when the ad was shown on a mobile phone. (An ad extension / asset).
• Calls from the number shown on the landing page.
• Contact form submissions.

The client tracked conversions from contact form submissions with the Google Tag. They used CallRail to track the calls.

Eliminating the obvious.

The change history showed no obvious reasons for the drop in conversions. Changes to targeting, budget and bid strategy often cause turbulence but they'd done none of these things.

The client told me that they'd made no changes to the landing page either. When I pressed them on this they told me they'd made one small change...

A few days before the performance drop they'd removed the tracking number from the landing page.

All calls from their tracking numbers go to a call centre that handles enquiries for the whole business. The manager from the department served by the problem campaign didn't like this. She had IT replace the tracking number with the direct number for her department.

Conversions dropped about 4 days after they changed phone number so they'd ruled it out as the cause. But, they were wrong.

Finding the root cause.

Removing the tracking number broke part of the conversion tracking. (The form and click-to-call button still worked.)

Broken conversion tracking is a big problem when you're using a conversion-based bid strategy. (In our case Target CPA.)

A conversion-based bidding algorithm asks 2 questions every time your ad could be shown:-

• Should we show an ad to this person?
• How much should we bid for a click from this person now?

The algorithm calculates the answers based on how likely it is that the person will call or fill in our form. (Become a conversion in Google-speak).

• If the probability of a conversion is high it'll bid high and our ad will show at or near the top of the page.

• If the probability is low it'll bid low. Our ad might not show. If it does show it'll be low down on the page and less prominent.

Conversion tracking tells the algorithm if it got the right answer or not. If a conversion is recorded the algorithm knows it got it right. If a conversion isn't recorded then the algorithm knows it got it wrong.

Over time it learns from this feedback and gets it right more often.

When IT removed the tracking number they stopped reporting some conversions. The algorithm thought it was getting it wrong so it stopped doing what was working before.

This change isn't instant. In the same way it takes a while to learn what works, it takes a while to learn what's not working. This is why there was a lag between breaking the conversion tracking and the drop in conversions.

My explanation made sense to the client. But they wanted to know why all their conversions dropped, not just calls from their landing page.

A bidding algorithm has 2 modes:-

• Explore where it's trying to figure out what works.
• Exploit where it's using what works to get as many conversions as possible.

Before they broke the conversion tracking most of the ad spend went on exploit mode. You could see this because they were getting a steady flow of leads.

Breaking the conversion tracking meant the algorithm thought what it was doing before wasn't working. It changed back to explore mode to find something that would work.

In explore mode the algorithm doesn't bid as high for each click because it doesn't know if the click will turn into a conversion. Bidding lower means our ads showed less often, and when they did show, they were lower on the page.

Ads that show lower on the page get clicked less often. The drop in clicks meant fewer people completed the enquiry the form.

And, generally only the highest ad on the page has a click-to-call button. That meant fewer conversions from calls.

Fixing the problem.

We fixed the problem by setting up a dedicated tracking number pool for this department. Calls will come direct to the department but will also register as conversions.

In time the algorithm will learn from this.

The client agreed to cut the learning time by increasing the target CPA from $50 to $100 and raising the budget. Once the campaign is back on track we'll reduce the target CPA and budget.

It appears to be working.

The day after we made the changes we got 11 conversions, the next day 9 and the next day 10. I expect it’ll take a week or so to get back to the previous performance.

Hope you found this useful.

Related articles

Testing Google Ads for an RV Campground

Why negative keywords are becoming less effective in Google Ads

What’s the right bidding strategy for a brand new Google Ads account?

You can't run profitable Google Ads if your pipes are leaking

The campaign looks good but the ads aren't profitable

Spent $35,000 on Google Ads, made no sales

Topics you'll find on this site