Measure an ad-matched page

A matched page feels right. The question worth paying for is whether it sells more than the page your ad pointed at before. A campaign test answers it for each ad, in orders and revenue.

How a campaign test is fair

For every ad, Proa takes only the visitors that ad sends and splits them between the original page and the page matched to that ad. Both groups clicked the same ad, so the only difference between them is the page. That is what makes the result about the page, not about the ad's audience.

Each ad gets its own experiment, and a visitor only ever enters one of them. In the wizard, each ad's row shows its link as you type.

Set one up

1. In Experiments, choose New, then Experiment, then Campaign test.
2. Basics: name it after the campaign.
3. Where it runs:
   • Ad parameter: comes filled with utm_content. Keep it if your ads already carry it. If not, choose the one they use from Presets or type it.
   • Page path: the landing page all the ads point at.
   • Ads pointing at this page: one value per ad (Add ad for each). Each row shows the link for that ad.
4. Variants: Same variants in every ad starts off, so each ad gets its own matched page. Each ad has the original page as its control and one variant: the version matched to that ad. The variant already carries the ad's name, and you can rename it.
5. Measurement: choose the metric that decides the test (conversion rate, revenue per visitor or average order value). See Reading results.
6. Review: check it and create the draft. Nothing goes live from the wizard.
7. Each ad has its own experiment in Experiments, named after the test and the ad. Open each one and choose Open the editor on its variant. Build the matched page for that ad. See Using the editor.
8. When every variant is ready, choose Start test at the top of each ad's experiment. Starting a test publishes its variants.

Before you start

• Orders must reach Proa. Connect revenue attribution through the Proa app or the order webhook. See Revenue attribution.
• Each ad is its own test. Two ads means two smaller samples, so each takes longer to reach a result than one ad would. If traffic is modest, start with your biggest ad.
• One link needs a campaign test. If all the traffic comes from a single link, such as one influencer, use a Campaign test, not External tool. With one link, nobody would see the original page.
• Keep the ads running. A test only reads the traffic its ad sends. If an ad stops running, its test stops getting visitors from it. The test stays running in Proa.

Reading the result

Proa shows no winner until each variant has enough visits and attributed orders, and the report says how far it is. When it is ready, it shows the result against the original for each ad: which page sold more, by how much, and how sure the reading is. See Reading results.

If the matched page wins, keep it: end the test with that variant as the winner and roll it out. See The life of an experiment.