A visitor who comes back to your site three times in a week is not the same lead as a visitor who showed up once. Most teams treat them identically anyway, scoring every visit the same and routing the account into the same generic queue. The fix is a visit-escalation framework: track how many times an identified account has returned within a set window, and change what you do at each threshold. Here's how to build it.

This is a companion play to our lead scoring model for identified visitors — that post weights an account's overall fit and behavior into one score. This one is narrower: it's about what a single signal, visit count over time, should trigger on its own, independent of the rest of the score.

What Actually Counts as a "Repeat" Visit

Not every reload of your homepage is a signal. Before you can escalate anything, you need a definition of a countable visit, and it should match the same denominator Knock2 uses for identification: an engaged session — a visit lasting 10 seconds or longer, or one that includes two or more pageviews. A five-second bounce off a blog post doesn't count. A visitor who lands on your pricing page, scrolls, and clicks into your integrations page does.

Once you're only counting engaged sessions, define the window. Two engaged sessions from the same identified account 90 days apart are a different signal than two sessions four days apart. Most teams get this right instinctively for obvious cases (a visitor back the next day is clearly warm) and get it wrong for the in-between cases, where a visit count resets silently because nobody defined a decay window. Set one explicitly — 14 days is a reasonable default for mid-market B2B sales cycles, tighter (7 days) if your sales cycle is short, looser (30 days) if it's long and considered.

Why One Visit Means Almost Nothing, and Three Means a Lot

A single engaged session tells you an account is aware of you. It doesn't tell you they're evaluating you. Multiple engaged sessions within your decay window, especially when later visits touch higher-intent pages (pricing, demo, integrations, case studies) than the first, are a materially stronger signal of active evaluation than any one visit could be on its own. That's the whole argument for tracking frequency as its own trigger instead of burying it inside a single composite lead score: a repeat visit deserves a different action, not just a few extra points.

The teams we talk to who get real pipeline out of visitor identification aren't the ones with the most sophisticated scoring model. They're the ones with the shortest path from "this account came back" to a rep doing something about it. A 40-factor score that takes a week to review loses to a 3-tier rule that fires in real time.

The 3-Tier Visit Escalation Framework

This is the framework itself, ranked by urgency. Use it as a standing rule on top of whatever base lead score you're already running, not as a replacement for it.

• 🔴 Tier 3 — Third-plus engaged session inside the window, or any return visit to pricing/demo: Escalate straight to a named SDR or AE, same day. Use the visit history as the opener ("noticed you were back checking out pricing") instead of a generic cold intro. This is a phone-and-LinkedIn tier, not an email-only tier — see our cold calling scripts for identified visitors for the talk track.
• 🟠 Tier 2 — Second engaged session inside the window: Trigger a Slack alert to the account owner and start light-touch outreach — a LinkedIn view or connect request, not a pitch. This is the tier where you're confirming interest, not closing it.
• 🟡 Tier 1 — First engaged session, ICP-qualified account: Log it, tag the account as "watching," and hold. No outreach yet. Most false starts in this motion come from reps who skip straight from Tier 1 to a pitch on the strength of a single visit.
• 🟢 Below threshold — Single visit, non-ICP or low-fit account: Log for pattern data only. Don't route to a human.
• ⚪ Stale — No return visit before the decay window closes: Visit count resets to zero. If the account comes back after the window, it re-enters at Tier 1, not wherever it left off.

The reset rule matters more than teams expect. Without it, an account that visited once in March and once in September gets scored as a "two-time visitor," which overstates the signal and trains reps to stop trusting the alert.

Where Most Teams Get This Wrong

Three mistakes show up constantly when we look at how companies actually implement visit tracking:

• Counting sessions instead of engaged sessions. If your tracking counts a two-second bounce the same as a five-minute pricing-page session, your "repeat visitor" list fills with noise and reps stop trusting it within a month.
• No decay window. Visit count becomes a lifetime counter instead of a momentum signal. An account that visited twice a year ago and once yesterday isn't accelerating — but a raw count treats it like it is.
• Same treatment for account-level and person-level repeats. A different person at the same company visiting twice is a buying committee forming, which is a distinct signal from the same named person coming back. Both matter, but they should trigger different plays — the first is a multi-threading opportunity, the second is a one-to-one follow-up.

How to Set This Up This Week

You don't need a data team for this, just a clean definition and a place to enforce it:

• Pick your window. Start with 14 days. Adjust after a month once you can see how it maps to your actual sales cycle.
• Dedupe by identified account, not by session. Waterfall enrichment and identity matching need to resolve to one account record so visit count doesn't fragment across near-duplicate company entries — see our waterfall enrichment guide if you're seeing this problem.
• Wire the tiers into your existing alert channel. If you already run tiered Slack alerts, add visit count as a second axis alongside page-intent tier — see our Slack alerts playbook for the no-noise version of this.
• Give reps the visit history, not just the alert. "Second visit in 6 days, pricing then integrations" is a usable opener. "New identified visitor" is not.
• Exclude bot and agent traffic before you count anything. Automated crawlers and AI agents can generate multiple "sessions" against the same account inside a day, which will fire your Tier 3 rule on noise. Filter this upstream — see our guide on AI agent traffic corrupting visitor ID data.

We've seen this play work across very different ICPs, from a three-person vertical SaaS team to a regional law firm to a real estate brokerage — the accounts and sales cycles look nothing alike, but in every case the same pattern held: a second engaged session inside two weeks was a far better predictor of a real conversation than any single-visit score, no matter how the first visit was weighted.

Where This Fits With Identification Accuracy

None of this works if the underlying identification is unreliable — a visit-count trigger built on shaky matches just escalates false positives faster. Knock2's identification benchmarks, measured against engaged sessions, are 93%* at the account level and 62%* at the person level for US traffic. Identification itself runs on an identity graph built from a consent-based publisher network, matched against your site's engaged sessions — we don't publish the underlying matching mechanics beyond that, but the accuracy numbers are what determine whether a "second visit" alert is trustworthy enough to hand to a rep.

*Identification rates measured against engaged sessions. Results may vary by traffic profile, geography, and industry.

FAQ

How many repeat visits before a rep should reach out?
Two engaged sessions inside your decay window is enough for light-touch outreach (a LinkedIn view or connect). Save direct SDR/AE outreach for a third visit or any return trip to a pricing or demo page.

Does a different person at the same company count as a repeat visit?
Not in this framework. A second engaged session from a different named contact at the same account is a buying-committee signal, not a visit-frequency signal — route it differently.

What if the visits are months apart?
They shouldn't count as consecutive. Set a decay window (14 days is a reasonable default) and reset the count to zero once an account goes quiet past it.

How is this different from a full lead scoring model?
A lead scoring model weighs firmographic fit, page intent, and behavior into one number. Visit-frequency escalation is one input inside that model, promoted to its own trigger because it's fast to compute, hard to game, and a strong enough signal on its own to justify an action.

Will bot traffic break this?
Yes, if you don't filter it first. AI agents and crawlers can rack up multiple sessions against one account in a single day. Filter known agent and automated traffic before visit counts feed into the escalation tiers.

Want to see this running against your own traffic instead of building the tracking by hand? Book a demo and we'll show you what a visit-escalation view looks like on your actual identified accounts, or see how Knock2's identification and alerting handles the tiering automatically.