Module 10 - Custom columns and CPA

Why can blended conversion numbers be misleading?

Because a single cost per conversion figure can mix cheap, low value actions in with the actions you actually care about, making the real cost look far better than it is. On a client account I looked at over the last 30 days, the graph looked like it was scaling nicely, with the number of contacts going up per page view, showing Google was starting to optimise well. The issue was that this client had originally set up page view as one of their conversion events.

Why would you deliberately track page views as a conversion?

Because for some businesses, simply getting more people onto the site is genuinely valuable in its own right. This particular client is a travel business, and people browsing for holidays typically look at 15 or 20 different websites before deciding, so at this stage the priority is getting more people onto the site, and optimising for page view can be a reasonable strategy while contact tracking, including phone numbers, is being built in alongside it. You could add in a micro conversion like time on site as a step before moving up to a firmer conversion like contact.

Why not just remove page view as a conversion event?

Because doing that outright would break the campaign, since Google would need to start relearning from scratch with only contact as the goal. In this case, the blended cost per conversion looked like fifteen hundred and fifty pounds divided by 222 total conversions, working out to six pounds ninety eight, but most of those 222 were page views rather than contacts, so the real cost per contact was being masked. Looking at raw numbers, one campaign showed 28 contacts for 35 pounds, working out to around one pound twenty five per contact, while another showed 49 contacts for around fifteen hundred and fifty pounds, closer to 30 pounds per contact, and it's important to be able to see that difference clearly.

How do you build a custom column to see the real cost?

By adding a custom column in Google Ads that divides total cost by the number of a specific conversion action, rather than relying on the blended total. In this account there was also a metric like time on site, showing 173 people who'd spent 30 seconds or more on the site as a landing page view, which again isn't the actual goal. To isolate the real cost of a phone call, go to columns, modify columns, scroll to custom columns, and add a new custom column, in this case named "cost per telephone call". Add performance cost as the numerator, then choose the specific conversion action, in this instance the s.media telephone click, as the denominator, and change the number format from a percentage to a monetary value.

What does that column show once it's applied?

Once applied and sorted from lowest to highest, it makes the winning campaigns obvious. The top campaign showed 28 calls at a cost of 35 pounds, giving one pound 25 per call, close to the estimate made earlier by eye. Another showed 18 calls for a cost of 645 pounds, and the average across this particular set of campaigns worked out to 25 pounds 60 per call. From there, you can start optimising using that column rather than the blended cost per conversion figure.