Marfeel Recommender: Setup and configuration guide

The Marfeel Content Recommender delivers personalized, engaging content to your audience. By combining different recommender engines with content restrictions, you can tailor recommendations to match user interests and your site’s goals.

This guide walks you through configuring a Marfeel Recommender step by step, from choosing the right ranking profile to setting up content restrictions and tuning the CTR signal. For a broader understanding of how feeds, engines, and curations fit together, see the core concepts overview.

Steps to Configure a Marfeel Recommender:

1. Creating a Recommender experience
2. Choose a ranking profile
3. Set Published and Visits from
4. Apply Content Restrictions
5. Configure the ranking CTR signal

1. Create a Recommender Experience

Every Marfeel Recommender is configured as a Marfeel Experience. You can set one up by navigating to Experiences screen and clicking on New Experience. The experience creation popup opens up showing available experience templates, where you can filter by recommender ones using the checkbox in the Family section of the sidebar. Marfeel offers a set of default templates for Recommender experiences, but you can set your own customized ones in Experiences settings.

Once you have selected the template you want to start with, the experience edition screen shows up. You can configure all the experiences common options here (Format, Targeting, and Delivery). Recommender specific options are configured in Content tab.

2. Choose a Ranking Profile

Marfeel offers several pre-configured recommender engines, each designed to prioritize different ranking signals. Selecting the appropriate engine depends on your content strategy and what you aim to achieve.

Available Recommenders

1. Simple rankers, no usage of Ranking Signals
   1. Most Read : Highlights the articles with the highest number of views in the selected Visits from period. Read more.
   2. Latest : Features the most recently published articles on your site. Read more.
   3. Oldest : (Amplify only) Displays articles in publication order, starting from the oldest ones. Read more.
2. Algorithm based rankers, different weightings of all Ranking Signals
   1. Popular: Prioritizes articles with high traffic while also factoring in engagement and other signals. Read more.
   2. Trendy : Showcases articles with a sharp increase in traffic over short periods. Read more.
   3. Similar : Recommends articles closely related to the one being viewed. Read more.
   4. Highest CTR : Surfaces articles that have performed well in terms of click-through rate. Read more.
   5. Personalized : Tailors recommendations to the user’s interests, preferences, and behavior. Read more.
   6. Balanced : Provides a balanced recommendation approach, combining all signals without heavily weighting any particular one. Read more.

3. Set Published and Visits from

Each feed answers two separate questions: which articles can be recommended, and how they are ranked. The feed form groups its fields under Articles and Ranking to keep them apart, and a one-line summary above the preview describes the selected feed as you edit it, for example: “Recommends articles published in the last week, ranked by visits in the last 24h.”

Published

Published, under Articles, sets how recently an article must have been published. Options: Any time, Last 24h, Last 3 days, Last week, Last month, or Custom with a whole number of days.

How strictly the period applies depends on where the recommender runs:

Recommender	Behavior
Recommender experiences	Articles published in the period come first. If there aren’t enough, older ones fill the gaps (see the Fallback Mechanism).
Amplify and Newsletters	Only articles published in the period can be recommended.

Published replaces the Days Since Published restriction: it saves the same “Days since published <” condition, so feeds configured before keep working and show their value in Published.

Visits from

Visits from (previously Time Window), under Ranking, sets the period of visits the recommender uses to rank articles: a short period (last 24 hours) highlights what’s trending now, a longer one (last week) favors sustained popularity. It doesn’t limit how old the recommended articles are. It’s hidden for the Latest profile, which ranks by publish date.

Example: a news site that wants this week’s articles ranked by what’s trending today sets Published to Last week and Visits from to Last 24h.

4. Apply Content Restrictions

Content Restrictions control which articles are eligible for recommendation. You can restrict recommendations based on various dimensions such as domain, section, author, topics, content type, and more.

Available Restrictions

1. Host : Limit recommendations to articles within the same domain or a specific host. This is ideal for multi-site setups.
2. Folder
3. Section
4. Author
5. Topic
6. Tag
7. HTML Language
8. Content Type : New, Recent, Evergreen or Not Editorial.
9. Has Image : Ensures that only articles with images are recommended, preventing broken images in modules that display thumbnails. True by default.
10. Visibility
11. Days since last update

How to Apply Restrictions

1. Same Value : Restrict recommendations to content that shares the same attributes as the current article (e.g., same section, author, or topic).
2. Specific Values : Choose specific hosts, sections, authors, etc., to include or exclude. You can use operators:
   1. Equals : Include articles that match the specified values.
   2. Not Equals : Exclude articles that match the specified values.

Example 1: Restrict to Same Section

To recommend articles from the same section as the current article being read by a user:

• Set the Section filter to Same Value.

Example 2: Exclude Specific Authors

If you want to exclude articles by certain authors (i.e. News Agencies):

• Set the Author filter to Not Equals and specify the authors to exclude.

Fallback Mechanism

If the number of articles matching your restrictions is fewer than requested, the recommender removes some restrictions to fulfill the request. The filters are removed sequentially in the following order:

1. Publish Time
2. Tags
3. Topics
4. Authors
5. Sections
6. HTML Language

Bear in Mind:

Marfeel Recommender will not recommend articles below a certain traffic threshold, deeming them as not relevant. For an article to be considered relevant enough to be recommended, it needs to fulfill any of the following minimum conditions.

• 50 visits during last 7 days
• 10 visits during last hour
• 5 visits during last 5 minutes

5. Configure the Ranking CTR Signal

The Ranking CTR (Click-Through Rate) signal evaluates how well articles perform based on user clicks in recirculation modules or specific contexts. Configuring this signal boosts articles with higher engagement, making your recommendations more effective. This matters because:

• Articles with higher CTRs are proven to attract user interest.
• Focusing on CTR can increase overall engagement and click-through rates.

Warning This is an advanced configuration parameter. Most of the times using Module Name = same is the desired option. It gets specially useful for reordering modules.

You can specify which CTR data the recommender should use by setting filters based on all the dimensions supported by Marfeel Recirculation module like module Name, origin and destination URL, Author or Section among others. For additional options that are not available in the UI, see the advanced configuration reference.

Using Module Name

Boost articles with higher CTR in the same recirculation module: the recommender will consider the CTR of articles within the same module, promoting those with higher CTR.

• Set Module Name to Same Value (Module Name = same).

If you have a recirculation module named “Related Articles” on your article pages, setting Module Name = Related Articles will prioritize articles that have performed well in that specific module.

Using Origin URL to promote most clicked articles in the Home Page

Recommend articles with the highest CTR from a specific origin page: the recommender will prioritize articles that have the highest CTR when users navigate from the specified origin URL.

To recommend articles that have the highest CTR from the homepage:

• Set Origin URL = homepage.

This configuration will boost articles that users frequently click on when visiting your homepage.

Using Destination Author

Recommend articles with the best CTR for a specific author. The recommender will promote articles that have high CTR and are authored by the specified author.

If you want to recommend articles with the best CTR from the same author as the current article:

• Set Destination Author = author.

This will prioritize articles written by the same author that have performed well in terms of CTR.

Troubleshooting

A new article does not appear in recommendations yet

Newly published articles are not available for the recommender immediately. They first need to be ingested into Marfeel’s content pool, which takes a few minutes after publication in your CMS.

If an article still does not appear after ingestion, check in this order:

1. Whether it meets the minimum traffic conditions.
2. Whether your active Content Restrictions (especially Days since last update) or the Published period exclude it.
3. Whether the Fallback Mechanism is removing other restrictions before the one you expected, causing older content to surface in its place.

What ranking profiles are available in Marfeel Recommender?

Marfeel Recommender offers simple rankers (Most Read, Latest, Oldest) and algorithm-based rankers (Popular, Trendy, Similar, Highest CTR, Personalized, Balanced) that use different weightings of ranking signals.

How does the fallback mechanism work for content restrictions?

If the number of articles matching your restrictions is fewer than requested, the recommender removes filters sequentially in this order: Publish Time, Tags, Topics, Authors, Sections, HTML Language.

What is the minimum traffic threshold for articles to be recommended?

An article must meet at least one of these conditions: 50 visits during the last 7 days, 10 visits during the last hour, or 5 visits during the last 5 minutes.