Get started with Optimizely Web Recommendations

Optimizely Web Recommendations is not currently included in the packaging for new customers. It is not a standard solution and requires great effort to configure. If you are looking for personalization for Optimizely Web Experimentation, contact your Customer Success Managers to find the right tool for you.

Optimizely Web Recommendations lets you add recommendations to any page on your site and measure the impact with the Stats Engine. See How Stats Engine calculates results for more information. You can introduce visitors to products and content based on your chosen algorithm and measure results.

Before you build a Recommendations experiment, do the following:

• Create a catalog and a recommender for your experiment.
• Configure the Recommendations algorithm and decide where it displays.
• Preview and verify your results.

Create a catalog

A recommendation experiment needs a list of items stored in a catalog to recommend to visitors. You must create a catalog before you create a recommendations experiment.

Your type of business determines the contents of the catalog. For example, retail sites may include all items for sale. B2B companies may include a list of blog posts or help articles.

1. Go to Implementation > Catalogs and click Create New Catalog.
2. Enter a Name and Description for your catalog.
3. Add the events and related tags you want to include in Catalog Events.
4. Click Create Catalog.
5. Click More (...) to download a CSV copy of your catalog or to archive it. It takes at least 24 hours for Optimizely to assemble it for download.

Set up a recommender

Your recommendation experiment also needs a recommender, which is the combination of logic, filters, and user inputs that tells Optimizely what items to recommend from your catalog.

1. Go to Implementation > Recommenders and click Create new Recommender.
2. Enter a Name and Description for your recommender. The recommender must have a unique name since all recommenders for all catalogs appear beside each other when you select one for your experiment.
3. Select a Catalog for this recommender. The recommender does not automatically use all the events in the catalog; instead, you attach specific events to this recommender to use later.
   After you select a catalog for your recommender, you cannot change it to a different catalog. You must start over by creating a new recommender.
4. Select an Algorithm for your recommender. See Recommendation algorithms for more details.
5. Choose the specific catalog events for you algorithm inputs.
6. Add the Boost Events you want if the algorithm supports them.
7. (Optional) Add a filter to customize your recommender further.
   
   You can change the algorithm after you set up the recommender. However, this removes all events and filters, and you must start over.
8. Click Create Recommender.

To create an extension to display your recommendations or to archive your recommender, click More (...) next to the recommender's name. Select the appropriate menu item and follow the prompts.

Create a filter

You can add filters to your recommenders. Optimizely applies the filter to any catalog items returned by the algorithm. Items that do not meet the filter conditions do not display in the recommendation's results.

1. Leave the Exclude all recommendations by default box unchecked without adding any conditions to get a list of every item in your catalog by default. Alternatively, check the box and do not include any conditions if you want the recommender to return no results by default. 
2. Select the items you want to include in this filter.
3. Specify if the filter applies to items that meet all the conditions you set or to items that meet any of them, if your filter has multiple conditions.
4. Select the values you want the filter to evaluate. These values reflect the field names on your catalog items acquired from the tags. Choose from the values listed or add a custom fixed value not included in that list.
   

   If you select custom, type in any literal fixed value.

   
   

   If you choose a tag, select which item the tag refers to. The Key Item is the page the visitor views, while the Recommendation Item is the item your recommender returns.

   
   

   This only applies when you use the co-buy or co-browse algorithms. The collaborative filtering and popular algorithms do not support the use of a key item, so this drop-down does not appear in those cases.

5. Select the comparison operator your condition uses. This lets you add another row of conditions to your filter.
   This row's use depends on your selection in step 3. If you selected any, only one condition row must be evaluated as true to return a match. If you selected all, all conditions must be evaluated as true to return a match.
6. Click Add condition group to add a new condition group to your filter.

Optimizely evaluates the condition groups to look for a match. If there are no matches, the recommender takes the default action specified in step 1.

Preview your results

You can preview the results the recommender generates before creating your experiment. If the results are not what you expect, you can refine the recommender until the results are correct.

1. Go to Implementation > Recommenders and find the recommender. Click Previewer. The Previewer has a set of summary statistics for your site's traffic over the last seven days:

   • Total Recommended Items – The number of rows of recommendations calculated when the algorithm was last run.
   • Partial Recommendations – Items that have fewer than ten recommendations, usually due to missing data.
   • Full Recommendations – Items that have a full set of 20 recommendations each.

2. Enter an Item ID in Preview Item Recommendations to preview the other items recommended to your visitors when they view that item. This is useful for algorithms that make recommendations based on items the visitor has already viewed. For context-free algorithms, such as Popular, changing the Item ID does not change the output because the output is not tied to any specific item.

Recommendation algorithms

When you set up your recommendations, you can choose from one of four algorithms:

• Co-browse – Recommends items similar to those a visitor has recently view. For example, if a visitor views several different tables during the current session, the algorithm recommends other tables.
• Co-buy – Recommends complementary products to those a visitor has previously bought. For example, if a visitor has bought a table, this algorithm recommends chairs.
• Collaborative filtering – Recommends based on the combined browsing history of the user and other similar users. This algorithm recommends different items for each user but remains consistent for the duration of an individual user’s viewing experience. For example, "You browsed similar products as this group of website visitors, and they tended to like them."
• Popular – Recommends crowd favorites on the homepage. Use it to introduce best-selling items to new visitors. For example, "Items that other visitors viewed or bought most often."

The co-browse algorithm is sensitive to the order in which a visitor views items. The co-buy algorithm is not; it checks if the same person bought the items.

Types of algorithms

A recommendations algorithm produces an ordered list of pairs (item, relevancy_strength) with 20 results. Optimizely’s algorithms use the collective behavior of past visitors for the results. The input to each type of algorithm is different:

• Visitor-based algorithm – Looks at the visitor ID and recommends based on each visitor’s specific viewing history. The recommendations generate specifically for each visitor and the results are the same on every page. For example, the recently viewed and collaborative filtering algorithms. 
• Item-based algorithm – Only looks at the page the visitor is viewing. The results are the same for each visitor but vary from item to item. These are described as similar items or people who liked this also liked. For example, the co-buy and co-browse algorithms.
• Global algorithm – Does not consider the current visitor or page. These recommendations appear on high-level pages (such as the home page) where they appear as most popular options or check out these best-sellers. An example is the popularity algorithm, which displays items in descending order of views, clicks, purchases, or the event your recommender uses. However, this does not personalize the results.

The input is the key that looks up a recommendation's output: the visitor-based algorithm takes a visitor ID as the input key, while an item-based algorithm uses the item ID. The global algorithm does not need a key at all. The output is the target. In an item-based recommender, both the key and target are item IDs; in a visitor-based recommender, the key is a visitor ID, and the target is an item ID.

Glossary of Recommendations terms

• Catalog – The collection of all items that appear in your visitors' recommendations. 
• Events – Optimizely uses visitor events, such as product page views, to construct the catalog. Additionally, events generate the recommendations and initiate scoring the different items in relation to the user or other items.
• Item – A unique product or a piece of content with its metadata that Optimizely recommends to your visitors. For retail customers, items correspond to specific products or a website URL. Items have unique IDs; catalog listings with the same unique ID are the same item. Different pages may have different sets of tags and represent different item types.
• Recommender – Consists of an algorithm, the specific inputs to that algorithm, and any business logic filters.

Recommenders take input from your visitors, run it through the algorithm and filters, and output a set of recommendations that display for your visitors.