What Is Purchase Intent Prediction? Who's Ready to Buy

Two Visitors, Two Different Predictions

Two people land on a coffee gear store at once.

Visitor A searched for one grinder by name. They open that page, pick it, and reach checkout in under three minutes.
Visitor B came from a social ad. They view five grinders, go back and forth, add one to the cart, then wander off to the blog.

A prediction would call A likely to buy and B less likely. An offer might win B's order. The same offer to A just cuts margin on a sale you already had.

You can't read minds. Behavior is another story. Purchase intent is how ready a visitor really is to buy. Purchase intent prediction is your best guess at it, built from what you can see.

What Signals a Prediction Reads

Most predictions blend small clues from the visit:

Where the visitor came from, like a brand search or social ad.
Which products they viewed, and how many.
How long they stayed on each page.
Cart activity: adding, removing, or leaving items there.
Whether they started checkout.

Simple setups use rules, like "added to cart, then idle for two minutes." Advanced ones use a model trained on past visits. Either way, it's a likelihood, not a sure thing.

Why Intent Prediction Matters on Shopify

Most visitors don't buy. Across 958 stores and 24 million visitors in our data, the average conversion rate was 2.2%. Only about 8 of every 100 sessions reached the cart. For more, see our Shopify visitor intelligence guide to purchase intent.

The useful question isn't just "who will buy?" It's "who might buy with a small nudge, and who needs none?" Getting that split right lets you save discounts for walk-away customers. See how dedicated buyers and walk-away customers differ.

Shopify's built-in reports look back. They show how many sessions added to cart, reached checkout, or converted. Prediction judges one visitor while they're still browsing. Acting on it, like showing an offer to just one person, usually needs an app. Compare options in our list of the best Shopify conversion rate apps.

The Two Mistakes Every Prediction Makes

No prediction is perfect. It can go wrong in two ways:

A ready buyer is tagged as unlikely. They get a discount they didn't need. You lose margin.
A walk-away customer is tagged as ready. They see nothing and leave. You lose a sale you might have won.

Example: Your store gets 1,000 visitors a day, with a $60 average order. Your setup flags 200 visitors as unlikely to buy and shows them 10% off.

10% of $60 is $6 off each order.
Say 20 of those 200 were ready buyers who would have paid full price.
If all 20 buy with the offer, that's 20 × $6 = $120 a day given away.

That $120 is the cost of the first mistake. Whether the setup pays off depends on how many real walk-away customers it wins back. Only a test can show that.

So don't take a prediction on faith. Check it with an A/B test. Model-based predictions also have less to learn from on very low-traffic stores.

How Growth Suite Predicts Intent

Growth Suite, our app, tracks timestamps, products viewed, cart activity and checkout events. It uses them to sort visitors into dedicated buyers, who are likely to buy, and walk-away customers, who may need a nudge.

Offers go only to visitors who look less likely to buy. The app waits for engagement signals, like time on page or cart activity, before showing one.
The offer adapts. More interested visitors get a smaller discount and a shorter timer. Less engaged visitors get a bigger discount and a longer timer, within limits you set.
Each offer uses a unique, single-use code. When the timer ends, the code is deleted on the server, so the countdown is real.
Cooldown periods help prevent offer fatigue, when shoppers see so many offers they tune them out.
The A/B testing module lets you judge results by conversion rate, AOV or total revenue.

Myth: A Long Visit Means a Ready Buyer

It's easy to read time on site as intent. A ten-minute visit can mean someone is comparing prices, stuck on sizing, or just passing time. A two-minute visit can be a buyer who already knows what they want. Time only means something next to other signals, like cart activity.