Every website has two kinds of visitors: those who are just browsing, and those who are about to do something. The hard part has always been telling them apart before the moment passes.
Traditional analytics can tell you what a visitor did after the fact. Intent scoring works the other way around. It watches behavior in real time and assigns a probability, a score between 0 and 1, for how likely that visitor is to convert in the near term.
"Intent scoring doesn't tell you who visited. It tells you who is about to act."
That sounds like a small difference, but it matters commercially. A high-intent visitor is worth far more attention than a casual browser, and now you can tell them apart automatically, without asking anyone to fill out a form.
The Problem with "Tell Us Who You Are"
The dominant model of web lead capture has been around for thirty years: put a form on the page and ask visitors to identify themselves. It works well enough until you notice all the people who don't fill it out.
Forms Capture a Small Fraction of Intent
Industry benchmarks suggest average B2B lead form conversion rates sit in the low single digits. That means most high-intent visitors to your pricing page leave without a trace. Many of them wanted to buy but weren't ready to fill out a form, so you never knew they existed.
Intent scoring gives you a read on those visitors too: the few who converted, and the many who didn't but whose behavior still gave you something to act on.
Forms Introduce Friction at the Worst Moment
A visitor who has read three blog posts, scrolled your pricing page twice, and spent four minutes on your features list is primed. Interrupting that moment with "Enter your work email to continue" breaks the flow. Intent scoring lets you know they're ready without asking them to announce it.
A form is a proxy for intent. Before behavioral ML it was the best proxy available, but that is no longer true.
The Signals: What the Model Actually Reads
Intent scoring is built on behavioral signals: things a visitor does, not things they tell you. Here's the signal taxonomy that a well-built intent model draws from:
Engagement Depth Signals
- Page depth: How many pages in one session. A visitor who reads 7 pages signals stronger intent than one who reads 1.
- Scroll depth: Did they read to the bottom of your pricing page, or bail at the first paragraph? 75%+ scroll on a conversion-focused page is a strong signal.
- Time on high-value pages: Time spent on /pricing, /demo, or /features correlates strongly with conversion. General browsing time doesn't.
- Click patterns: Clicks on CTAs, feature comparison tables, and testimonials indicate evaluation behavior. Clicks on nav menus and search suggest exploration.
Recency & Frequency Signals
- Return visits: A visitor who returns three times in a week is researching, not just browsing. Return frequency is one of the strongest predictors of near-term conversion.
- Recency: A visitor who was on your pricing page 2 hours ago is more likely to convert today than someone who visited 3 weeks ago.
- Session acceleration: Visits that get longer and more focused over time. First visit: homepage. Second: blog. Third: pricing + demo page. The trajectory tells the story.
- Time of day: B2B visitors during business hours on weekdays convert at higher rates than weekend browsers. The model learns your site's specific pattern.
Context Signals
- Traffic source: A visitor from a branded search ("WysLeap pricing") converts at a much higher rate than a visitor from a generic blog post referral. Source carries prior intent.
- Device type: For B2B SaaS, desktop visitors convert significantly more than mobile visitors. The model learns your specific mix.
- Geography: Visitors from your strongest markets, or from cities with high concentrations of your ideal customer profile, carry a context bonus.
- Page sequence: Certain page paths are highly predictive. Homepage → Features → Pricing → Demo request is a classic high-intent funnel. The model identifies yours.
WysLeap's own model deliberately starts smaller: it scores visitors on five signals (sessions, pages viewed, engaged time, scroll depth and clicks) so it can learn reliably from the conversion volume a typical site actually has.
What Intent Scoring Does NOT Use
A properly built intent model works entirely on behavioral signals. It doesn't require names, emails, company names, job titles, or any personally identifiable information. This is what makes modern intent scoring privacy-first by design, and it's why it doesn't depend on cookie consent.
The model reads what people do, not who they are.
How the Model Works: From Signals to Score
The signals above are inputs. The model turns them into a single number: an intent score between 0 and 1. Here's how that transformation happens, without the jargon.
Step 1: Feature Extraction
Every visitor action is converted into a numerical feature. "Visited /pricing" becomes 1. "Visited /pricing 3 times in 5 days" becomes 3 with a recency weight applied. "Scrolled 80% of /pricing" becomes a weighted decimal. Each interaction enriches the visitor's feature vector.
Step 2: Weight Assignment (Where the Learning Happens)
This is the core of the model. Each feature is assigned a weight that reflects how strongly it predicts conversion on your specific site. The weights come from your site's actual historical conversions, not a generic industry model. Generic models might assume "visited pricing page" is worth 20 points. But on your site, maybe visitors who come from LinkedIn are 4x more likely to convert than those from Twitter. Your model learns these site-specific weights from your actual conversion history.
Step 3: Continuous Adaptation
Static models decay. Visitor behavior changes with seasons, marketing campaigns, product launches, and market conditions. An adaptive intent model retrains as new conversions arrive, adjusting weights to reflect the current reality. WysLeap's model starts from default weights, trains on your own data once a site has at least 15 conversions and 30 non-converting visitors in the last 90 days, and then retrains every 10 conversions. Each new model is only used if it beats the default on held-out visitors. We don't publish accuracy figures.
The Output: A Score and a Label
In WysLeap, every visitor gets a score between 0 and 1 and a tier:
- High intent (0.6 and above): The strongest buying signals. Prioritize these visitors: trigger live chat, personalize the page, alert sales.
- Medium intent (0.3 to 0.6): Actively evaluating. Worth nurturing with retargeting, educational content, and case studies.
- Low intent (below 0.3): Early stage or casual. Don't over-invest. Let them explore and wait for signals to develop.
What You Can Actually Do With an Intent Score
An intent score is only useful if you act on it. Here's where it pays off:
For Sales Teams: Prioritize Outreach
If your site has any sales component, such as demos, trials, or direct sales, intent scores give your team a daily list of who to contact first. Some leads are much warmer than others. A visitor who has been on your pricing page three times this week is a warmer lead than one who signed up to a webinar six weeks ago. Treat them differently.
For Product Teams: Personalize the Experience
A high-intent visitor seeing your homepage for the third time shouldn't see the same hero as a first-time visitor. Show them the demo CTA, a testimonial from their industry, or a limited-time offer. Personalization based on intent beats personalization based on demographics, because it reflects what the person is actually doing right now.
For Marketing Teams: Smarter Retargeting
Retargeting everyone who visited any page wastes budget. Retargeting only high-intent visitors concentrates spend on the people most likely to convert.
For Live Chat: Trigger at the Right Moment
Proactive chat prompts that fire on every visitor after 30 seconds are annoying and convert poorly. Prompts that fire only for a high-intent visitor on the pricing page for the second time in three days arrive when help is actually wanted. Intent makes your chat tool intelligent instead of aggressive.
Intent Scoring in Practice: Industry Examples
E-Commerce
High-intent signals: visited a product page 3+ times, added to cart without purchasing, filtered by size/color (indicating they're past browsing and into evaluating), returned within 24 hours.
Action: Trigger a "limited stock" nudge or free shipping offer only for high-intent visitors on product pages.
B2B SaaS
High-intent signals: visited /pricing, viewed the Enterprise plan specifically, read 2+ case studies, returned from a branded search, visited during business hours on a weekday. SaaS evaluation cycles are longer, so medium-intent visitors also matter and are worth nurturing.
Action: Surface high-intent visitors as PQLs (product-qualified leads) for sales teams to follow up. Enroll medium-intent visitors in an automated nurture sequence.
Content & Media
High-intent signals: read 5+ articles in one session, scrolled to the paywall multiple times, visited subscription pricing page, came from email newsletter (indicating existing relationship). For media, "conversion" means a subscription, and intent scoring identifies who is one nudge away.
Action: Show a personalized subscription offer only to high-intent visitors, without a generic popup that annoys everyone else.
3 Common Misconceptions About Intent Scoring
"You need a huge dataset to start"
Not with WysLeap. Every site starts on sensible default weights from day one, and the model begins training on your own data once there are at least 15 conversions and 30 non-converting visitors in the last 90 days. After that it retrains every 10 conversions. You don't have to wait until you have enough data. You start, and it learns.
"It requires personal data or cookies"
Intent scoring built on behavioral signals requires zero personal data. No email, no name, no cookies. All signals come from anonymous behavioral patterns, so it is privacy-first by design and cookieless by default.
"High score always means they'll convert"
Intent scoring is probabilistic, not deterministic. A high score means a high probability of conversion, not certainty. The goal is resource allocation: spending more energy on high-intent visitors, not guaranteeing they'll buy.
See which of your visitors are high-intent right now
WysLeap's intent scoring starts from default weights on day one, with no cookies and no personal data required. It learns from your own conversions once there are enough of them, and only switches to a new model when it beats the default on held-out visitors. Start free, no credit card needed.
