Home/Solutions/Purchase Intent Prediction

What is a purchase-intent prediction system?

A plain-language guide to purchase-intent prediction: what it is, what signals it uses, and how Quantellix.AI's DXI score puts it into practice.

Short answer: A purchase-intent prediction system uses AI to read a visitor's on-site behavior and output a single propensity-to-buy score. Quantellix.AI's version is DXI, built from click, scroll, search and transactional signals.
Quantellix.AI·Reviewed by the Quantellix.AI product team·Updated September 2026

What is purchase-intent prediction?

Purchase-intent prediction is the use of AI to estimate how likely a specific website visitor is to buy, based on their behavior during the session — rather than waiting to see whether they convert. Quantellix.AI's implementation is DXI, a 0-100 propensity-to-buy score produced by its LNM-AI model.

Signals used to predict purchase intent

  • Click and scroll behavior across product and category pages.
  • On-site search terms and search depth.
  • Session length, return visits and visitor loyalty.
  • Transactional signals once a visitor reaches checkout.

Quantellix.AI's data shows buyers engage roughly 4.3x more than non-buyers across these signals, and typically run 1-2 on-site searches before purchasing — the kind of pattern a purchase-intent model is trained to detect.

4.3x
buyer engagement multiple
1-2
avg. searches before purchase
Top 20%
visitors flagged for retargeting

From score to action

A purchase-intent score is only useful if a team acts on it. With DXI, that typically means: retargeting the top 20% of scored visitors, adjusting messaging for visitors with a low Transactional sub-score, and setting a realistic conversion goal based on where top-performing sessions already land.

Purchase-intent prediction accuracy benchmarks

In head-to-head testing on real ecommerce datasets, DXI-based scoring reached 99.8% accuracy on visitor conversions and 98.6% on ad-impression conversions, ahead of general-purpose LLM and AutoML baselines tested on the same data. See the full DXI methodology page for how the underlying score is calculated.

Frequently asked questions

What is an AI purchase-intent prediction system?
A purchase-intent prediction system is AI software that analyzes a website visitor's behavior — clicks, scrolls, searches and session length — and outputs a single score describing how likely that visitor is to buy. Quantellix.AI's version of this score is called DXI.
What signals go into a purchase-intent score?
Common signals include click and scroll depth, search terms used on-site, time on page, return visits, and transactional behavior at checkout. Quantellix.AI's LNM-AI model combines all of these into four sub-scores before producing the final DXI number.
How accurate are purchase-intent prediction systems?
Accuracy varies by vendor and data quality. In Quantellix.AI's own benchmark testing across real ecommerce case studies, its DXI-based scoring reached 98-99.8% accuracy on visitor-conversion and pricing-efficiency metrics.