What is an up-sell repeat purchase model?
An up-sell repeat purchase model scores customer accounts for their likelihood to purchase a product that they have bought from you before. In order to score accounts against this model, the accounts must have purchased the target product of the model from you at least once before.
What do I need to be able to create up-sell models?
What is an example use case for an up-sell repeat purchase model?
An up-sell repeat purchase model is valuable when you want to identify accounts that are likely to purchase the same product from you again. You want to use this model in a campaign that has the purpose of driving existing customers to purchase the same product from you again.
Creating an up-sell repeat purchase model
First you need to do some preparatory steps, and then you can build an up-sell repeat purchase model.
Preparation step 1: Identify the campaign you want to run
Before you build a model you first must decide what the purpose of your campaign is. In the cross-sell first purchase article, we created a model that will help us predict which customers will purchase a Display product for the first time. But now we want to know which customers are likely to purchase the same product again.
A Repeat Purchase model is set up the same way as a First Purchase model, with one important difference: you must select the timeframe in which the customers have not purchased the product. This is called a purchase lapse period.
Let’s say the Product Manager for Displays thinks that if companies buy our Displays and like them, those companies will buy more again within six months. So the Product Manager wants to figure out which companies are likely to buy again.
Preparation step 2: Build the segment that represents the audience you want to target
Since our campaign is targeting existing customers to make a purchase of Displays again, our marketing manager has built a segment with the following criteria:
- Customers who have purchased a Display product.
- Customers who are located in the United States, Canada, or Mexico (Use this and other firmographic factors to limit the customer base to a certain market. This is helpful if you want to run different campaigns in different markets that still target the same product.)
Start creating an up-sell repeat purchase model
Now it is time to train the new up-sell repeat purchase model and score your segment against that model.
Go to the Models tab in Rev.Up and click Create Model.
Choose the model type by clicking on "Customers that will Purchase Again Next Month".
Choose the target segment
Choose the segment that you built in Step 2. In our case that segment is called "NA Customers - Displays Repeat Purchase".
Choose the product that you want to sell
Choose the product that you want to sell to your customers who have purchased it in the past. In our case, this product will be Displays.
You must also choose the timeframe in which the customers in your segment have not purchased the product (the purchase lapse period). You can use this setting as a way to train the model on how often customers should make a repeat purchase of a product. For example, some products may be purchased monthly, while other products only need to be purchased every 6 months. This setting will vary based on the product.
In our case, our marketing manager thinks the product should typically be purchased every 6 months.
Choose the kind of model to be trained
In the next screen, select one of two options that define the kind of model to be trained:
Likely to Buy: The model scores and rates your customer accounts using ratings from A to D on their likelihood to make a repeat purchase of the selected product (Displays). Accounts with the rating A are the most likely to buy, accounts with the rating D are the least likely to buy.
Likely Amount of Spend: The model measures your customers' likely amount of spending when making a repeat purchase of the selected product (Displays).
Make selections about how you want to train the model
In the final screen, make selections about how you want to train the model. There are several options and statistics displayed on this screen.
- Accounts to be Scored: The number of accounts in the segment that will be scored. The best practice is to create a very specific segment, but the system will filter out any accounts that don’t meet the scoring criteria (the accounts must fit the segment's criteria, have previous transactions, and must have purchased the target product in the past). This is why this number could be lower than the total number of accounts in your segment.
Records Selected: The number of account records that will be used to train the model.
How did this system come up with this number? Briefly, the system is looking at accounts in our segment (accounts who have purchased a Display product in the past). It also considers accounts that have never purchased the Display product. Such accounts are considered for all eligible periods.
Eligible period: This depends on what you selected as your purchase lapse period. The model looks at whether or not a previous purchaser had the purchase lapse specified (6 months in this case). If so, the record will be included in training. Then the model looks ahead to see if the purchaser made a repeat purchase and labels the account accordingly.
The main thing to pay attention to: The number of records selected should be at least 5,000.
Repeat Purchases: The number of customers who made a repeat purchase of the target product. In our case, this means the number of customers who have purchased the Display product once and then purchased the same product again. These accounts represent desirable events that the model uses to compare against other accounts that have not repeatedly purchased the target product.
The main thing to pay attention to: Make sure the number of Repeat Purchases is at least 50. The modeling job will fail otherwise. An optimal model should have at least 200 repeat purchases.
- Remember that our segment criteria for building the model included customers who have purchased the Display product at least once. Therefore, the number of Repeat Purchases means that there are 364 accounts that made a repeat purchase of a Display product.
Now that we understand how the accounts used for training the model are determined, we can use the following settings to refine the training data used by the model. Adjusting these settings will change the number of records shown in Accounts to be Scored, Records Selected and Repeat Purchases.
Only use successes where spend in the month/quarter was: Using this option removes accounts with purchases that do not meet the criteria from the model training data. You can use this setting to set the total amount of spend that you are looking for when a customer makes a repeat purchase. You can adjust this to target accounts that are likely to spend more or less in a repeat purchase of your product.
Only use successes where quantity bought in the month/quarter was: Using this option removes accounts with purchases that do not meet the criteria from the model training data. You can use this setting to specify the total amount of units purchased that you are looking for when a customer makes the repeat purchase. You can adjust this to target accounts that are likely to purchase at least (or at most) a certain number of units in the repeat purchase of the target product.
When modeling, only use historical records from the last ... months: Using this setting trains the model with purchases only from the selected time range. Use this when something significant changed in the past (with Account, Transaction or Product tracking), and you do not want to use older data for training the model.
An example of a significant change is when the way the product is packaged and sold has changed.
Let me model on a similar segment, not exactly the one I want to sell into: This setting uses a separate segment for model training, but the model will still score the records in the original target segment you created and selected. If you have insufficient data to train the model using your target segment, you can use a similar larger segment in its place.
For example, if you wanted to break into the Baltic market but don’t have many Baltic accounts in your system yet, you could train the model on a segment of Western European accounts but assign model scores to the Baltic accounts. You can do that by selecting your Baltic segment on the screen that asks you “Which segment do you want to sell into?”, and then select your Western European segment here.
Let me model on similar products, not exactly the ones I want to sell: This setting replaces the actual target product that you want to sell with a different product for the purpose of training the model. This setting is useful when you have insufficient data to train using the target product. In this case, you can use a similar one with a fuller purchase history in its place.
For example, if you just started selling Displays and don’t have much history of transactions involving Displays in your system, you could select another product that is similar and has more history.
Start model training job
After you have made these selections, click Model. The model will take 2 - 4 hours to create. Once the training is complete, you can come back and review the model.
How do I know that my model is good and I should publish it?
Rev.Up checks to make sure that the up-sell model will be good to use once it has been trained. The model creation wizard will not allow you to build a model if there are not enough records available to use to train the model.
Because of this, you are not required to review the up-sell model as closely as Account Fit models (Rev.Up does not perform as many verifications for Account Fit models).
To ensure that you have a good up-sell model, make sure to closely follow all of the steps in this article.
How do I activate my model and use the output to build a segment?
Click Activate on the model dashboard to see what your model predicts for your target Accounts.
You will be able to see the Rating buckets, and change their bounds to tune your Ratings. (This functionality is available in the Ratings tab at any time even after activating the model.)
Click Activate Configuration to create a Scoring Action in your Process & Analyze queue. The next time you run Process & Analyze, your model ratings and scores will be updated in My Data. You can use them to make new segments, rule-based models, or plays.