What browsing behaviour can tell you before a cart exists

A brief visit could be curiosity, an accidental click or a quick answer to a specific question. Repeated visits could reflect comparison, uncertainty or interest. A page view alone cannot tell you which explanation is right.

Look at patterns in context:

• A single product visit gives you little reason to change the experience for that visitor
• Visits to the same product on different occasions could make product details or a buying guide worth testing
• Visits across one category could point towards a comparison or category-level guide
• Repeated visits to delivery, sizing or returns information could suggest an unanswered practical question

Check the underlying measurement before acting. Reloads, duplicated events, staff activity and automated traffic can distort a pattern. A browser or device identifier also does not necessarily represent one person.

Write the observation separately from the explanation: “returned to the category” is something you may have measured. “Ready to buy” is an interpretation.

Begin with a change you can explain in customer terms.

For a hypothetical outdoor retailer, repeated visits across hiking backpacks might lead to this question: would a simple capacity and fit comparison help shoppers choose?

That question can support several tests. Choose one first.

Improve the website

Make the comparison easier to find on the category page. A homepage link to a useful buying guide may also help people find their way back.

You can test clearer content for everyone without building a personal browsing profile. Where permitted and technically supported, you could separately test showing a relevant category guide to a returning browser. Keep the normal navigation available and provide a sensible default when recognition is unavailable.

Send a relevant email

For a reliably identified contact with the appropriate permission, test one helpful email linking to the comparison. Explain the products clearly and avoid language that makes an unsupported claim about what the recipient wants.

An email address in the CRM does not automatically connect a website visit to that contact. Nor does it establish permission for this message. As one implementation example, Klaviyo’s browse-flow documentation describes product-view tracking for identifiable browsers. Verify the equivalent behaviour in your own setup.

Test permitted retargeting

If your existing advertising setup supports it, a permitted audience could receive an ad for the comparison guide. Decide the audience window, exclusions, frequency controls and spending limit before launch.

Audience creation and ad delivery are separate steps. Google Analytics documents how audiences can become available to linked advertising products when personalized advertising is enabled. Check your actual connections, consent settings and platform requirements before using that path.

Define the question before adding more tracking. For the backpack example, a useful starting set might be:

• The product or category viewed and when the event occurred
• A permitted browser identifier and, where reliably available, the contact match
• Relevant consent and channel eligibility
• Recent cart, order and campaign activity
• The guide or product information needed for the experience

Keep detailed browsing events in the system that needs them. Your CRM may need only a time-limited category-interest signal, its source and its last update. Avoid filling customer records with every page view.

Map which system supplies each fact and how quickly updates arrive. Test what happens when a customer signs in, changes devices or withdraws permission. If the match is uncertain, leave the visit unmatched.

Start by checking native features and existing integrations. A new platform becomes a consideration when a specific requirement cannot be met reliably with what you have.

An A/B test compares alternatives. Personalization changes the experience according to a rule or segment. You can test whether a particular personalization rule actually helps.

Here is a hypothetical pilot for the backpack retailer. These numbers illustrate a test design; they are not recommended benchmarks.

1. Audience: eligible, reliably matched email subscribers who visit the backpack category on two separate occasions within seven days
2. Exclusions: recent purchasers, people in a higher-priority cart journey and anyone whose permission or required data is missing
3. Change: randomly assign eligible contacts to one comparison-guide email or a control group receiving no additional browse email
4. Contact limit: at most one pilot email per contact, with the normal campaign limits still applied
5. Review: set a four-week pilot window and a required sample size before launch, adjusting the design to the buying cycle and available traffic

Choose one primary result, such as purchase rate across everyone assigned to each group, measured over the same follow-up period. Clicks can help explain what happened, but the people who clicked are a self-selected subset.

Track unsubscribes, complaints, delivery failures and conflicting messages alongside the main result. Avoid changing the website and retargeting at the same time for these groups, because that makes the email result harder to interpret.

At the planned review, decide whether to keep, change or stop the test. If the sample is too small or the data is unreliable, record the result as inconclusive.

Treat tracking, email and advertising eligibility as separate checks. Explain how browsing data will be used, collect only what the experience needs and set a retention period.

The Office of the Privacy Commissioner of Canada’s behavioural-advertising guidance addresses clear purposes, consent choices and limits on collection and retention. The applicable requirements depend on the audience, jurisdiction and implementation. Get qualified advice for your circumstances; this guide is operational information, not legal advice.

Before releasing an email or updating an audience, check whether the customer has purchased, entered another journey or changed permission. Decide how quickly audiences must update and what to do during a data delay.

Have an owner, a pause control and a record of why someone qualified. Test the stop conditions as carefully as the entry conditions.

Repeated visits to a service, pricing or demo page may suggest a useful content question: are people missing implementation details, an example or a clear explanation of the next step?

Improve that information first. Where reliable identity and permission exist, the pattern could help choose content in an existing nurture journey or add context to a CRM record. Do not turn anonymous visits into named prospects or treat browsing as a request for a sales call.

Once an enquiry arrives, give the person replying the original request and relevant, appropriately held marketing context. Keep observed behaviour separate from assumptions about intent, and leave unknown needs or timing visible. Let the enquiry process take priority and avoid overlapping follow-ups.

A useful progression is clearer website content, one consented journey, then better coordination between the website, marketing platform and CRM. Each step should answer a real customer question and produce evidence for the next decision.

If you need to bring suitable visitors in first, read the guide to customer acquisition.

If your immediate gap is after an item reaches the cart, start with the abandoned-cart automation guide.