Marketers often struggle to find the perfect incentive balance to meet their goals. Either they use too many, or they aren't using them at all.
Powered by Emarsys AIM, Incentive Recommendation is a landmark product on our journey towards truly 1-to-1 personalization. We analyze your campaign launch lists to identify which customers are more likely to purchase when offered an incentive, and then assign the most efficient incentive to each one.
In this article
- What is Incentive Recommendation?
- How does it work?
- What incentives are available
- How quickly does the algorithm learn?
- How does this affect my email campaign design?
- What reporting is available?
- Supported functionality
What is Incentive Recommendation?
Incentive Recommendation is an AI-based feature which uses a learning algorithm to decide whether or not a contact requires an incentive and if so, what level of incentive.
In other words, instead of sending the same incentive to the whole launch list, or grouping them in broad categories, you simply define the range of incentives you want, and we decide which one to send to each customer on a case-by-case basis.
For even more flexibility, you can use Incentive Recommendation in the Automation Center in conjunction with other AIM products, such as Send Time Optimization.
How does it work?
You upload a series of incentives in the form of images or HTML code snippets to Emarsys and assign a value to each one.
We then use behavior data to predict your customers’ responses – for example, whether 5% off would be enough to influence their decision, or if they would require a full 25% off before considering a purchase. Based on this behavior history, and taking into account the additional settings you apply to the smart incentives, we decide which incentive to display to each contact.
If we decide that a contact requires no incentive, this content is either removed from the email (you should design your email to allow for this, see below), or is replaced by some other content of your choice.
Additionally, you can use a control group to measure the effectiveness of this feature. If you choose to do this, a percentage of contacts are selected at random and all are sent a fixed incentive instead of a smart incentive.
At the same time, you decide how conservative or aggressive the campaign should be. Being more aggressive means that more contacts receive more valuable incentives, incurring higher costs on your side.
To help your decisions, the Details box provides graphs, which show our predictions of costs and revenue for your selections. As you move the slider, the diagrams change accordingly.
Incentive Recommendation only decides which incentives to show to which contacts. It is up to you to provide the voucher code or landing page that will allow you to process the incentive.
What incentives are available
Currently, you can define one-time percentage reductions, fixed-amount coupons and free shipping offers, but other discounts are also planned for future versions. Once created, the incentives are added to the email campaign as placeholders in the email.
How quickly does the algorithm learn?
Incentive Recommendation is powered by a learning algorithm, which means results will become better and better over time.
The algorithm needs 10-15 campaigns sent to each contact before it can recommend with confidence; during this learning period results may fluctuate.
Expect a longer learning curve with leads and inactives, due to the limited amount of contact data and less interaction with your campaigns.
Learning is much faster with active and defecting customers and you can expect even better results, thanks to their higher engagement and the larger amount of data available.
For our math-savvy users, here is how the algorithm actually works:
- We track how incentives perform in terms of how many people make a purchase upon receiving incentives. This forms the basis of a general model that calculates how much higher incentives boost the likelihood of a purchase.
- For each individual contact, we combine the predicted purchase probability with a cart value prediction to calculate expected revenue and cost of each potential incentive.
- Revenue and cost are weighted according to the strategy set by the customer so that we can choose the best incentive for each contact.
For more information on calculations, see Incentive Recommendation Metrics.
How does this affect my email campaign design?
The goal of Incentive Recommendation is to make sure the revenue you miss out on incentives is not wasted. If our algorithm decides that a recipient has a good probability of making a purchase, the incentive placeholder will be removed and they will not see that content.
The missing incentive will be replaced by an empty (white) pixel; you should therefore add the incentive placeholder in such a way that the email will still look good if it is not there. But you also have the option to replace the incentive placeholder with some other content intended for recipients who do not need any incentive.
Here is an example of an email where the following incentives and alternatives have been defined:
- 10% discount
- 20% discount
- Free shipping
- Alternative content A (with some product images)
- Alternative content B (without any specific content)
For more information on how to add the incentives to your email content, see Adding and configuring the incentives.
What reporting is available?
You can see the aggregated metrics for incentives across all the campaigns that used them on the Incentive Recommendation Dashboard, which you can access by going to Add-ons > Incentive Recommendation.
Currently, incentives reporting is not available on a campaign level.
- Incentive Management - create different kinds of incentives based on your incentive strategy.
- Strategy Console - define the incentive distribution strategy: aggressive or conservative.
- Monitoring and Measurement - aggregated reporting shows the revenue and margin contribution of each campaign.
- Reminder Campaigns - increase the chance to convert by re-sending.
- Incentive Alternative - content for customers who don’t need to be incentivized.