Skip to main content
All CollectionsOptimize
Contextual Auto Optimize
Contextual Auto Optimize

Contextual Auto Optimize offers sophisticated automatic optimization, powered by a Contextual Multi-Armed Bandit algorithm.

Updated over 7 months ago

As with Auto Optimize, you add multiple pieces of content to the optimization, and the system’s machine learning algorithms automatically allocate traffic to the best performing items.

Learn more in Auto Optimize.

The difference with Contextual Auto Optimize is that it also considers the different context of each visitor when making decisions. Contextual Auto Optimize works fast to show the best performing items to ensure that you’re gaining the maximum return for the items within the optimization.


Key information about Contextual Auto Optimize

  • You can set the goal you want to achieve for the items within the optimization.

  • Works on items in the optimization being seen, and not on them being clicked. This means reporting may differ from other areas of the system, and that items in the test don’t need to be clickable for the Contextual Auto Optimize to work.

SmartBlocks used with Contextual Auto Optimize must be metered (dynamic).


Use cases

  • Recommendation SmartBlocks targeting differently priced products, using the device type context to optimize which to show.

  • Different content that appeals to different types of people, for example, families with babies vs families without babies, and using weekend or weekday to optimize the right one to show.

  • Different content for different stages in the shopping journey, for example, first site visit to multiple purchaser, and use number of sessions to automatically optimize when content is switched over.


Create and manage a Contextual Auto Optimize

To access Contextual Auto Optimize, you must have the Advanced Testing and Optimization module. If you’re interested in adding the module to your account, or would like to discuss the module, contact your Account Manager.

Before setting up a Contextual Auto Optimize, first you should create the content that you want to include in the optimization; this could be multiple SmartBlocks to be added to a Slot, Trigger programs or Experiences.

Once you’ve created your content:

  1. Go to Optimize in the left hand navigation menu.

  2. Select CREATE OPTIMIZATION.

  3. Expand the Location drop-down menu, and select where you want to add the optimization:

    • Experiences

    • Website slots

    • Email slots

    • Triggered email slots

    • Triggered email programs

  4. Select the option (Slot, program, or Experience) you want to use, then select CONTINUE.

  5. Drag the items (SmartBlocks, Trigger programs or Experiences) that you want to use from the left into the decision tree on the right. Position the items on top of one another.

  6. When you have a group of multiple items, select the top row to open the Optimization panel on the left.

  7. Select Contextual Auto Optimize.

  8. Expand the Select the goal for this Optimization drop-down menu and select from:

    • Increase Conversion

    • Increase Average Order Value

    • Increase Identification Rate

    • Decrease Bounce Rate

    • Increase Site Visits

    • Increase Average Time On Site

    • Increase number of items in purchase

    • Increase revenue

  9. Expand the Context drop-down menu and select the context you want the optimization to work in.
    Learn more in the section Contexts.

  10. Select SAVE.

The Contextual Auto Optimize is now set up and ready to go. If the Slot, Trigger or Experience Rule is live, then the Contextual Auto Optimize becomes live as well.

Your new Contextual Auto Optimize now appears in the Optimize Center:

  • The coloured dot indicates if it’s active.

  • Select the row to return to the optimization setup.


Contexts

Contexts are the different data points the Contextual Auto Optimize machine learning algorithm can use to influence its decisions about which content to show.

For example, if testing content A and B against each other, the system also takes into account the context selected, such as day of the week. It may find that content A performs better Friday to Sunday, but content B performs better Monday to Thursday. It therefore shows different content on different days — the best performing piece towards the selected goal.

It’s only possible to select a single context per optimization. For each context that’s selected, multiple Multi-Armed Bandits (the algorithm behind Auto Optimization) are created in the system, one for each element of the context. Therefore, if too many are selected, then there are too many variables and the system is unlikely to reach a successful outcome.

The default context is Auto. For this, the system selects the context that has the highest correlation to the outcome of the goal you select, based on all recent visitors to your website. The context the system selects to use is based on mathematics, and how strongly each context correlates with the selected goal. Auto uses the context that has the strongest correlation.

The system may select a second context if correlation is high for multiple contexts. Auto is therefore likely to select the context that has the biggest impact. However, as Auto is based on all visitors, if you’re deploying content that is targeted for a specific context type, or on a page or location targeting a specific type of visitor, you might want to manually select the context.

The contexts available vary per goal and website, because we only include contexts that are deemed relevant by the system algorithms. In some cases there may not be any contexts available; in this case, we’d suggest using Auto Optimize instead.

Available contexts:

These are updated regularly in the system, so this list may not be exhaustive.

  • Daylight/Dark

  • Weekend/Weekday

  • Operating system

  • Device type

  • Country

  • Number of purchases

  • Number of sessions

  • Day of the month

  • Day of the week

  • Number of products browsed

  • Number of pages viewed

  • Purchase price prediction


Reporting

The reporting for Contextual Auto Optimize is the same as for Auto Optimize.

Learn more in the section Auto Optimize: Reporting.

Did this answer your question?