impact.com's Discover Recommendations leverages the power of machine learning to recommend ideal partners you can reach out to and recruit for your program. Our Data Science team has put together a series of well-established machine learning algorithms to analyze millions of data points across the impact.com partnership ecosystem, resulting in partner recommendations that you can trust.
See the Recommendations deep-dive section below for a detailed technical explanation of how impact.com curates these recommendations.
When in Discover, you can access the recommended partners from the left navigation menu by selecting Recommendations.
By default, all the partners that impact.com recommends will be shown. You can filter your recommendations using the Recommendations filter. Select the filter, then select which types of recommended partners you want to view. See the Recommendations filter reference below for a description of the different types of recommended partners.
Recommendation type | Description |
Conversion Path Introducers | These are the top 5 partners (not yet in your program) that play an introducer role to at least 25% of the conversion paths they participate in. |
High AOV Potential | These are the top 5 partners (not yet in your program) that generate the highest 90-day average order value across all programs they participate in. |
High Revenue Potential | These are the top 5 partners (not yet in your program) that generate the highest 90-day average revenue across all programs they participate in. |
Partners from Similar Brands | These are partners (not yet in your program) that our machine learning algorithm sees as most similar to your brand and have the highest propensity to be present in programs from brands like you. |
Partners You Will Love | These are the top 10 partners (not yet in your program) that are most likely to have productive partnerships. |
As mentioned in the introduction, impact.com leverages the power of machine learning to recommend partners to you. The inputs of the machine learning algorithm include, but are not limited to:
Partner attributes (Location, promotional methods, retail vertical, etc.)
Partner audience demographics
Partners’ historical performance metrics such as payout history and revenue
Brand and program similarity with other brands and programs
Brand and program attributes (Location, retail vertical, products, product categories, etc.)
Brand’s and program’s target audience
Product taxonomy and items sold
These machine learning algorithms are then applied to the entire partner universe so all partners get a propensity score. This score helps us sort and list partners based on how relevant they are to your program.
The relevancy sort order is based on the machine learning propensity score, but is also adjusted by various productive and relevant factors:
Productive Rate Given Brand Send—We check how often a partner generates payable actions when a brand proposes a contract to them. Those that are more productive are given a higher relevancy sort order.
Partner Size—A re-ranking is given based on the size of the partner, which helps find quality relationships.
Low Payout—If the partner has a low payout (less than $200 or less than $1,000 in 90 days), then they are given a lower score.
Country Match—We ensure that the partner can work with brands in your country.
Verifying that the partner will work with small brands—If any partner has significant evidence of not working with a small brand, we don’t recommend them to small brands.
Not an Existing Relationship—We will only recommend partners with whom you are not in an existing relationship.
Partner Tags—If your brand belongs to a specific industry, we check if that industry aligns with partner tags. For instance, we look for words like “personal finance” and “investment” for the finance industry.
We hope this deep-dive has strengthened your trust in the recommendations that impact.com provides. Our Data Science team is continuously tuning the machine learning model and considering new inputs to further optimize the quality of the recommendations.