Estimate the traffic required for success
The Adobe Target Traffic Estimator provides feedback that lets you know whether you have sufficient traffic for your Automated Personalization (AP) activity to succeed.
Because Automated Personalization activities use multiple offer combinations, it is important to know how much traffic is required to provide meaningful results. The Traffic Estimator uses statistics about your page and the number of experiences being tested to estimate the amount of traffic and the test duration necessary to make the activity successful.
The Traffic Estimator determines if there is enough traffic to generate personalized models by comparing the estimated page impressions and typical conversion rate for the pages. Ideally, for a successful activity, the correct sample size ensures that personalized content is ready within 50% of the activity duration or 14 days, whichever is less. This process allows sufficient time for obtaining personalized content and learning which content to deliver.
Remember that Target randomly serves experiences until the personalization algorithms are built. The check mark icon beside each offer shows when the model for that offer is ready and Target is able to begin delivering personalized content. Because lift is expected only after the models are ready, the visual indication lets you set the right expectation. Use the Traffic Estimator in the Visual Experience Composer (VEC) to get a guideline of when the models are ready.
Use the Traffic Estimator
-
From the Experiences page of the Visual Experience Composer in an Automated Personalization activity, click the Traffic icon.
The Traffic Estimator opens. You can click Traffic again to hide the Traffic Estimator.
-
Specify the typical conversion rate (or the conversion rate that you expect from this activity), estimated activity impressions per day, and test duration.
table 0-row-2 1-row-2 2-row-2 3-row-2 4-row-2 Metric Description Number of Offers This metric is calculated automatically based on the number of experiences being created as a part of your activity, after any exclusions. Typical Conversion Rate This metric is expressed as a percentage, based on your estimation or past data from your analytics system. Estimated Visits Per Day This metric is the number of visits per day from visitors who are able to view the activity, based on the targeting criteria. This metric could be based on your analytics data. This number must be visits, not unique visitors. Test Duration The number of days you want the activity to run. The Traffic Estimator uses these metrics to determine what adjustments are necessary to run a successful test.
Near the top of the Traffic Estimator, the values you entered are calculated and the results are shown.
As you change the numbers, the estimate changes. For example, if you are testing many combinations and your conversion rate and impressions are too low, the Traffic Estimator shows how long the test must run to be successful. Or, if your traffic is low, the Traffic Estimator could suggest a lower number of offer combinations so you can run the test the desired number of days.
If you do not have sufficient traffic, consider the following:
- Consider using an Auto-Target activity instead of Automated Personalization to create experiences with several offer changes in one experience variation.
- Reduce the number of offer combinations within your Automated Personalization activity.
- Increase the duration of the activity.
Adjust the numbers until the Traffic Estimator indicates that you have sufficient traffic, then design your test accordingly.
If the traffic is sufficient, the Traffic icon shows a green check. If it is insufficient, the icon shows a red warning label.
Frequently Asked Questions about the Traffic Estimator
Consider the following FAQs as you work with the Traffic Estimator:
Why are personalized models not built even though my AP activity has enough traffic?
In certain circumstances, your traffic is large enough for a personalized model to be built, but that traffic could inform Target that there is no meaningful difference between the personalized model and random. Although the model is built in Target and tested, it is not deployed because the model is not any better than random.
A possible reason for the model not being better than random could be that the offers are not different enough from each other. If so, you can try making the offers more visually different if the messaging is similar, or you can try changing the messaging itself.