Click-through Rate

Ad reports used to come back saying 300,000 impressions and nothing else. It's a big figure and looks like things are working, but impressions rise whenever budget gets spent. Double the money and you double the impressions, so that number reports how much was spent and nothing else.

Which meant the people reading the report and the people running the ads kept talking past each other. The report says 300,000 impressions; site visits are flat against last month. The ad did appear in front of people, and nobody responded to it — both true, and an impression count can't distinguish between them.

A number that counts response was what was missing. Click-through rate (CTR) turns appearances into a ratio of how many were clicked, which makes the creative and the audience assessable independently of how large the budget was.

Divide clicks by impressions. 3,600 clicks on 300,000 impressions is 1.2 percent. Three things to watch.

1. Remember the denominator counts events, not people. Appearing three times to the same person is three impressions. Campaigns that repeat to the same audience accumulate impressions quickly and see the rate fall as they run.
2. Fix the unit you group by. Mixing one campaign's average with one creative's value inverts the conclusion. If creative A runs at 3.1 percent and B at 0.4 percent, viewing them combined at 1.2 percent hides the creative that should be turned off.
3. Read it after the sample accumulates. Twelve clicks on 1,000 impressions is also 1.2 percent, and there a single click moves the figure by 0.1 points. Decisions to pause or keep a creative wait until impressions pass a few thousand.

Read it against the placement as well. Ads above search results commonly run 2 to 5 percent; a banner on the side of a website commonly runs 0.1 to 0.5 percent. The same 1 percent is poor in the first place and very strong in the second. So the number doesn't get compared across different placements — it gets compared across creatives in the same placement.

"CTR is 3.4 percent and sign-ups are flat. We need to look at the landing page."

Getting the click worked; the leak is in the step after it.

"Creative B is at 0.4 percent. It's past 8,000 impressions, so let's pause it."

The sample has accumulated and the value is low, so the call can be made.

"This is a display banner. Don't compare it against search numbers."

Different placements have different baselines.

Ravi runs four creatives at a time at Pagedeck and uses this number to pause two of them. Last quarter the creative showing a product screenshot came in at 0.6 percent while the one written as a sentence about the customer's problem came in at 2.8 percent. Moving the budget onto what survived tripled the visits the same spend produced.