Likelihood To Buy
The Likelihood To Buy (LTB) model predicts near-future repeat purchasing
behavior based on past transactions, email, and browsing behavior. This model
helps marketers to predict the buying behavior of customers in the following
time intervals:
- Today
- This month (whether the customer will make a purchase within the month of the
scheduled campaign)
- 1 month ago (whether the customer made a purchase a month earlier than the
scheduled campaign)
- 2 month ago (whether the customer made a purchase two months earlier than the
scheduled campaign)
Use cases
Revenue or margin optimization
- Identify the customers that are likely to buy.
- Give lower discounts to the customers that are likely to buy.
Alternatively, you can give higher discounts to the customers that are
unlikely to purchase without a high discount. Thus, you can stop
cannibalizing margin on the former while earning incremental revenue on the
later.
- Improve lift by targeting the customers with a higher likelihood of buying
through expensive campaign mediums like direct mail.
Reactivation optimization and churn prevention
Identify customers who are not trending in likelihood to buy and
create programs to avoid churn.
Customer comprehension
- Understand the customer by pivoting likelihood groups with other dimensions
such as demographics.
- Understand how transactional, email, and web engagement variables are
drivers of repeat purchases.
Target audience
The LTB model predicts the behavior of buyers. A buyer is a customer with a
transaction in the past three years. Hence, anyone who hasn’t bought an item in
the last three years isn’t included in this model.
How does the model work?
For each contact, the model calculates:
- the purchase behavior features such as lifetime value, recency, frequency, and
aov
- the email behavior features such as send, open, click frequency, volume, and
recency
- the web behavior features such as browse or abandoned cart counts, session
duration, session recency
CDP trains and tests the model on historical data, where the outcome is known.
Post training, the model is deployed to predict future engagement.
The output of the model is the likelihood to buy for each customer.
This probability is segmented in 10 equal-sized deciles, 1 being the most likely decile to yield
purchases:
| Decile |
Description |
|---|
| 1, 2 |
High |
| 3, 4, 5 |
Medium |
| 6, 7, 8, 9, 10 |
Low |
| NA |
Non buyer |
| NA |
Not qualified |
Using the LTB model
CDP displays the output from the LTB model in:
- 360 > 360 Profiles
- Actions > Campaigns and Actions > Campaigns+
- Analytics > Metrics
You can add the deciles: 1 - High and 2 - High as additional filters for:
- Pushing information to Facebook as it is an expensive channel. Hence, you can
run daily campaigns that send information to Facebook to target high LTB
customers.
- Sending direct mails to audience lists through ad-hoc campaigns. Thus, you can
limit the audience to high LTB customers.
To review the continued effectiveness of the model, you can use a dashboard to
compare the likelihood to buy deciles from one month earlier to recent
purchases (in the last one month).
To learn more about how the Likelihood To Buy (LTB) model can enhance your
marketing workflow, contact us.
