Platform · Purchase prediction
A purchase propensity model that tells you who will buy next, and what.
Most of your list isn’t ready to buy this week, and a few people are. LiftSignals scores every customer’s chance of buying in the next 30 days, predicts the product they’ll want next, and suggests the next best action, so you send fewer messages to the people who are ready.
No credit card · Read-only access · Works with Shopify, WooCommerce, BigCommerce, Amazon, Square or a CSV
Definition
What is a propensity model?
A propensity model estimates how likely each customer is to take an action, here, to place an order in the next 30 days. It turns a list of customers into a ranked list, so the people most likely to buy see the right product at the right time and everyone else isn’t pestered.
HOW A SCORE READS
Propensity 0.72 = a 72% predicted chance of ordering in the next 30 days
Customers are ranked from highest to lowest each night. Category affinity scores sit alongside, for example Footwear 0.41 and Accessories 0.12, so you know what to show, not just whom to send to.
| Batch-and-blast | LiftSignals purchase prediction | |
|---|---|---|
| Who gets it | Everyone on the list | The customers most likely to buy |
| What they see | Your bestsellers | The product each customer is likely to want next |
| When | Your campaign calendar | When each customer usually engages |
| Result | More sends, more unsubscribes | Fewer sends, more orders per send |
Why sending to everyone backfires
Most of your sends land on people who aren’t ready to buy.
01
Volume trains people to ignore you.
Every irrelevant email teaches a customer that your messages can be skipped, including the one that matters.
02
Bestsellers aren’t everyone’s next purchase.
The customer who just bought a pair of running shoes doesn’t need another pair. They might need socks.
03
You discount people who were already buying.
Without a propensity score, the coupon goes to the customers who would have checked out this week at full price.
How LiftSignals does it
Five predictive models working on one customer record.
Purchase propensity
Probability of ordering in the next 30 days, for every customer, rescored nightly.
Next product
The product or category each customer is most likely to buy next, from what similar customers bought after the same items.
Market basket
Which products are bought together, so bundles and cross-sells come from your own orders, not a hunch.
Next best action
One suggested move per customer: nudge, cross-sell, wait or win back, based on the scores together.
Send time
Each customer’s likely best day and time to hear from you, shown on the customer profile.
PROPENSITY TO BUY · NEXT 30 DAYS
Illustrative read-out for a store with 8,400 customers. Your scores are computed from your own order history.
What to do with it
Four ways to spend fewer sends and make more orders.
01
Send the launch to the people likely to buy
Email a new product to customers with high propensity and high category affinity first. Widen the audience only if it sells.
02
Put the right cross-sell in the order confirmation
Use next product and market basket to choose the one item worth suggesting after each purchase.
03
Hold the coupon for people who need it
High-propensity customers get full-price reminders. The discount goes only to those who are likely to buy with it and unlikely without.
04
Retarget only the ready
Send your top propensity band to Meta and Google and exclude the bottom band. Less spend on people who weren’t going to buy this month.
Proof on every play
Then prove the targeting beat sending to everyone.
Every audience you send carries an automatic holdout: a random slice of those customers who don’t get the message. 30 days later you see what the campaign caused, not what it claimed, in dollars, with the confidence interval printed next to it.
Is it right for you?
Who purchase prediction is for.
A good fit if you
- Sell a catalog with repeat or related purchases
- Send email or SMS several times a month
- Want product recommendations from your own orders
- Have at least about 500 customers and 90 days of orders
Not the right tool if you
- Sell a single product with no cross-sell
- Need real-time on-site personalization widgets
- Want recommendations driven by browsing behavior only
- Have only a few hundred orders so far
FAQ · purchase prediction
Purchase prediction questions, answered.
See how propensity fits a retention program in customer retention strategies.
What is a purchase propensity model?
A model that estimates each customer’s probability of buying within a set period. LiftSignals uses 30 days and ranks every customer by that probability each night.
What is next best action?
A single recommended move for each customer, such as a cross-sell, a gentle reminder, a win-back or simply waiting, chosen from all of that customer’s scores together.
What is market basket analysis?
Analysis of which products tend to be bought together. LiftSignals reports pairs with how much more often they appear together than chance, using your own orders.
How is this different from Shopify product recommendations?
Shopify’s on-site recommendations suggest related items on product pages. LiftSignals predicts what each customer will buy next and when, so you can use it in email, SMS and ads.
How often do scores update?
Every night, from your latest orders.
How many orders does purchase prediction need?
Propensity scores need about 500 customers and 90 days of orders. Market basket pairs need enough orders containing both products to rule out chance, so pairs with too few orders stay hidden until they have them.
RELATED CAPABILITIES
Customer lifetime value →
Predicted 12-month value for every customer.
Audience sync →
Send scores to Klaviyo, Meta, Google and more.
Discounts and returns →
Who needs a coupon and who sends it back.
See who is ready to buy this week.
Connect your store this afternoon. Scores arrive overnight.