They looked at the same thing a few times and left.

Inferred

Repeat views of one product or plan, no cart, and no completed decision.

Example eligibility rule

Require at least three views of the same item across two sessions in 14 days, with no cart, trial, or purchase. Offer the save or compare path on the third qualifying view rather than waiting for cursor-exit detection.

Three PDPs, a pricing page, a doc, then exit. The inference is interest. The failure might be confidence, effort, or a bad page. Intervention means a light, skippable next step, not a takeover.

Avoid this assumption

Exit intent is not consent. A third view is only a hypothesis. This case offers a low-pressure return path. "They keep coming back to the same product. They still won't buy." addresses the confidence gap on the page itself.

Test a quiet return path, such as save, compare, or ask, against doing nothing. The signal is uncertain, so the response should remain light and easy to dismiss.

Reach

How many eligible people can you reach in the chosen channel and time period?

Cost

What will content, delivery, QA and ongoing maintenance cost?

Return

What incremental outcome would cover that cost, and what is it worth?

Planning inputs, not measured results. Estimate reach × incremental outcome rate × value per outcome, then subtract cost. Use a range and a consistent time period.

Example format: Experience

A dismissible "save this" or "compare" on the third view, not on exit.

Compare lighter and deeper responses

Choose the lightest response that solves the need.

Assist: Offer useful help

Nothing in-session. A later note only if they already gave you a channel.

Adapt: Change a meaningful part

A dismissible "save this" or "compare" on the third view, not on exit.

Transform: Tailor the journey

On return, that item first, with proof at the point of commit.

What to watch out for

• A modal on first product view.
• Treating comparison research as abandonment.
• Retargeting ads that follow them louder than the site did.

Control

No treatment.

Treatment

Third-view save or return-path, with a holdout.

Success measure

Later conversion of that item.

Guardrail

Scroll depth and bounce. If the chip disrupts the page, remove it.

Among visitors who met the third-view rule, compare the save or compare option with no intervention. Measure later conversion of that item and track dismissals, saves never reopened, and cross-item purchases.

Win or lose, record what worked, for whom and why. What you learn becomes the next signal.

• Klaviyo Behavior-triggered email and SMS with real-time customer profiles and segments, built for ecommerce.
• Insider One Cross-channel engagement platform (web, app, messaging) that builds a unified profile from attributes, real-time behavior, and precomputed predictive scores to drive personalization and campaign targeting.
• Nosto AI commerce platform combining personalized search, recommendations, merchandising, and content/A-B testing under one data layer.
• VWO Testing Testing platform with audience/behavior targeting and a visual editor; personalization runs as a separate module from Experimentation. As of September 2026, VWO and AB Tasty merged their brands under the new parent brand Wingify, with each platform still running independently during the transition.

They keep coming back to the same product. They still won't buy. Answer what's stopping them. Don't just say buy again.

They left with items in the cart. Remind them of the actual items. Discount last, not first.