TapCub is live — analytics, insights and live chat, free on one platform
Growth & Marketing

Buyer Intent Signals: What to Track on a Website and How to Score It

Eight first-party buyer intent signals on a website, a worked example that measures each signal's conversion lift, and a signal-to-score table built from your own numbers.

Emma Collins Published Nov 20, 2026 11 min read Growth & Marketing
Sample data: 600 of 20,000 visitors requested a demo (3.0%). Visitors are assigned to their highest signal; chat and pricing are 15% of visitors but 65% of requests.Sample data

Buyer intent signals are visitor behaviours that statistically precede a purchase or enquiry: viewing pricing, returning within days, opening a chat, reading several product pages, starting a form. Their strength is specific to your site, so measure it: in a sample of 20,000 visitors, the 15% who opened chat or viewed pricing produced 65% of demo requests. Below is a signal table and a way to turn it into scores.

What you will learn

  • Measure each signal's conversion against your base rate rather than copying a generic list
  • Give each visitor the points of their strongest signal, not the sum, because signals overlap
  • Record when a signal happened: a pricing view in the last 3 days converted at 14.0% against 7.6% for older ones

What are buyer intent signals on a website?

Buyer intent signals are the things a visitor does that statistically precede a purchase or enquiry: viewing pricing, coming back within days, opening a chat, reading comparison or integration pages, or starting a form. They matter because most of a buyer's research happens where you cannot see it. Gartner's 2026 survey of 646 B2B buyers (fielded August to September 2025) found 67% prefer a rep-free buying experience, and its 2025 release found 73% avoid suppliers that send irrelevant outreach. If buyers research alone and punish bad timing, the signals they leave on your own site are how you tell who is ready.

This page is about first-party signals, the ones your own analytics records. Third-party intent data (topic research elsewhere on the web) is a different product and is covered in intent data for sales. The goal here is a signal-to-score table you can build from your own numbers, not a generic list. Tracking the leads that result is in lead tracking.

Which website signals show buying intent?

SignalWhat the visitor didWhy it suggests intentCaveat
Pricing page viewOpened pricing or plansCost is a late-stage questionCompetitors and students also look
Repeat visitReturned within 7 daysStill evaluating after the first visitIncludes support users and employees
Chat openedStarted a conversationAsked a question a page did not answerPeople who chat are self-selected
Product or integration pagesViewed 3 or moreChecking fitResearchers read as much as buyers
Comparison pageRead how you differChoosing between optionsOften early, often anonymous
Form startedBegan sign-up or demo formHighest effort so farAbandoned forms are intent too, not failures
Return from a campaignArrived from an email you sentAlready known to youSource tells you who, not when
Logged-in activityUsed the product (trial or free plan)Direct evidence of valueOnly identified users

The honest limit of this table is the caveat column. Every signal is a probability, not a verdict, and the strength of each is specific to your site. That is why the next section computes the strength from your own data.

SignalPoints (rate, rounded)Suggested action
Chat opened25Reply within minutes; hand to sales on price questions
Pricing viewed10Offer help on the page; follow up if identified
Repeat visit within 7 days4Show a relevant next step; add to nurture
3+ product pages2Keep educating; no sales contact
None of these0Do not spend effort

Points are the sample conversion rates; use the strongest signal per visitor and recompute quarterly.

How strong is each signal? A worked example

Sample data: 20,000 visitors in a month, 600 of whom requested a demo, so the base rate is 600 ÷ 20,000 = 3.0%. Because signals overlap, assign each visitor to the highest-intent signal they showed (chat, then pricing, then repeat visit, then product pages), so the groups do not double-count:

Highest signalVisitorsDemo requestsRateLift vs 3.0%95% interval
Chat opened60015025.0%8.3×21.7% to 28.6%
Pricing viewed2,40024010.0%3.3×8.9% to 11.3%
Repeat visit within 7 days3,0001204.0%1.3×3.4% to 4.8%
3+ product pages4,000601.5%0.5×1.2% to 1.9%
None of these10,000300.3%0.1×0.2% to 0.4%
Total20,0006003.0%n/an/a

The groups add back to 20,000 visitors and 600 requests. The first two groups are 3,000 ÷ 20,000 = 15% of visitors but produce (150 + 240) ÷ 600 = 65% of demo requests. That is the reason to score: it tells sales which 15% to call first.

Lift is rate ÷ base rate. Chat at 8.3× is the largest, and also the one most inflated by self-selection: people who open a chat are already closer to asking for a demo. Treat it as an upper bound for what chat adds, and as a very strong signal of who is ready.

app.tapcub.com

Demo request rate by highest signal

sample, 20,000 visitors
Chat opened25.0%
Pricing viewed10.0%
Repeat visit4.0%
3+ product pages1.5%
None of these0.3%
Sample data
Sample data: 600 of 20,000 visitors requested a demo (3.0%). Visitors are assigned to their highest signal; chat and pricing are 15% of visitors but 65% of requests.

How do you turn the table into a score?

Use the observed rate as the points, rounded, and give a visitor the points of their strongest signal rather than the sum, because overlapping signals would otherwise count the same intent twice:

SignalPoints (rate, rounded)Suggested action
Chat opened25Reply within minutes; hand to sales if the question is about price or terms
Pricing viewed10Offer help on the pricing page; follow up if identified
Repeat visit within 7 days4Show a relevant next step; add to a nurture sequence
3+ product pages2Keep educating; no sales contact
None of these0Do not spend effort

A visitor who opened a chat and viewed pricing has 25 points, not 35. Adjust the thresholds with a holdout: after a month, check whether visitors scored 10 or more actually convert at about 10% or higher, and recompute the rates each quarter because they drift.

Agree a response time for each tier before you launch the score, for example a reply to an opened chat within 5 minutes during office hours, a follow-up to a pricing view within one working day, and an entry in the weekly nurture cycle for repeat visitors. A score nobody acts on within a day or two loses most of its value, as the freshness table later on this page shows, and a tier with no owner is only a report.

How many visitors do you need before trusting a rate?

Enough that the interval is narrower than the gap between neighbouring tiers. In the sample, chat's interval of 21.7% to 28.6% sits clearly above pricing's 8.9% to 11.3%, so the ranking is solid. If the chat group had only 40 visitors with 10 requests, the same 25.0% would have an interval from about 14% to 40%, wide enough to overlap pricing's range. Under about 100 visitors, treat a rate as a hint. Combine two weeks of data rather than splitting a month into days.

How quickly does a signal go stale?

A signal loses value with age, so record when it happened, not only that it did. Split the 2,400 pricing viewers in the sample by how long before the month's end they viewed pricing:

Pricing viewedVisitorsDemo requestsRate
In the last 3 days90012614.0%
4 to 14 days ago1,5001147.6%
All pricing viewers2,40024010.0%

The rows add back to 2,400 visitors and 240 requests. A fresh pricing view is worth about twice an older one (14.0% against 7.6%), so the sensible rule is to award full points inside a short window and fewer after it, for example 10 points within 3 days and 5 points within 14 days. Pick the cut-offs from your own sales cycle: a product bought in one session needs hours, one with a procurement process needs weeks.

What signals mislead?

  • Counting your own team and customers. Exclude internal traffic and existing customers, or the repeat-visit and docs signals inflate.
  • Treating one signal as a buyer. A single pricing view by one anonymous visitor is 10% in the sample, which also means 90% do not request a demo.
  • Adding overlapping signals. Use the strongest, or test the combination as its own group.
  • Reading chat as proof that chat works. Self-selection inflates the gap; compare against similar visitors.
  • Ignoring the buyer who never shows a signal. Gartner's 2025 release found 69% of buyers noticed inconsistencies between a seller's website and what the rep said, so a mismatch can end an evaluation before any signal appears.
  • Never recalculating. A scoring table from last year reflects last year's pricing page and traffic.

How do you capture these signals in TapCub?

Take the sample above. In TapCub:

  1. Measure the signals. Pricing and product page views come from Web analytics › Pages; new versus returning visitors from Web analytics › Audience; opened chats from Live chat › Inbox. Set a goal for the demo request in Web analytics › Goals so each signal's conversion can be computed.
  2. Compare chatted with not chatted. Live chat › Reports › Conversion attribution compares goal conversion for visitors who chatted and those who did not, warns when fewer than 30 chatted, and lists the pages where chats start.
  3. Know who they are. If your product has a login, call tapcub.login; the person then appears in All-platform analytics › Users › People with a day-by-day behaviour timeline, and the chat shows who it is without asking for an email.
  4. Define the groups. In All-platform analytics › Users › Segments & tags, build a rule segment such as "did Pricing viewed in the last 7 days and did not do Demo requested"; identified people only, since anonymous visitors do not enter segments.
  5. Open the visitor. In a chat, the sidebar shows source, the last 2 days of pages and past conversations, with one click to the visitor profile.

TapCub does not score leads or push them to a CRM; it records the signals and shows each one's conversion so that you can set the scores yourself.

FAQ

What are examples of buyer intent signals?

Viewing the pricing page, returning within a few days, opening a chat, reading three or more product or integration pages, starting a form, arriving from a campaign you sent, and using a free plan or trial. Strength differs by site, so measure each against your own conversions.

How do you identify buyer intent on a website?

Record the page views and events that precede a conversion, compute each signal's conversion rate against your base rate, and rank them. In the sample, chat open converted at 25.0% and pricing view at 10.0% against a 3.0% base.

What is the difference between first-party and third-party intent data?

First-party data is what visitors do on your own site and is recorded by your analytics. Third-party intent data comes from activity elsewhere on the web, collected by data providers. This page covers first-party signals; the other is discussed in the intent data article.

Do buyer intent signals work for B2C?

Yes. A shopper who views a product page three times, adds to cart and returns the next day shows intent in the same way; the signals differ (cart, wishlist, price checks) but the method of measuring each against a base rate is the same.

How many signals should I score?

Three to five. Beyond that, signals overlap and the table becomes hard to maintain. Start with chat, pricing and repeat visit, and add one only if it separates visitors who convert from those who do not.

Sources

Was this article helpful?

We use your answers to fix and expand this page.

Emma Collins

Product Analyst

Focuses on event analysis, conversion funnels, and retention, using clear metric definitions and structured analysis to explain product performance.

See which signals precede a conversion

Page, audience and goal reports, chatted-vs-not conversion, and a visitor profile one click from every chat.

See your data clearly. Find your growth.

Every click, backed by data. Install one line of code and see your first numbers in a minute.

No credit card · Free plans for analytics and chat · Cookieless analytics