Buyer Intent Score for Real Estate: How to Build One That Actually Predicts Site Visits

Most lead scores are useless. They are built from CRM field completeness and rep optimism, not from what the buyer actually did. A real buyer intent score weighs behavior, recency, and conversation signals — and it should correlate directly with who shows up for a site visit. Here is how to build one.

Every real estate sales team has some version of a lead score. Usually it is a 1-to-5 rating filled in by reps after a qualification call. Occasionally it is a field in the CRM with a traffic-light color. In the worst cases, it is a label like "hot" or "warm" that was applied four weeks ago and never updated. These scores rarely predict anything useful because they were not built to.

A buyer intent score is different. It is a quantitative measure that combines what the buyer has actually done — microsite engagement, conversation depth, recency of activity — into a single number that tells a rep which lead to call next. When it is built correctly, the correlation between the score and actual site-visit attendance is strong enough that the sales manager stops arguing about which leads to prioritize.

Why Most Lead Scores Fail

Before building something better, it is useful to understand why the existing score is not working. Almost every failed lead score shares the same weaknesses.

1. It is filled out by humans, not captured from behavior

If the score is a rep-entered rating, it reflects rep optimism, not buyer interest. Reps systematically over-rate leads they connected with recently, and under-rate leads with difficult conversations that happen to be genuinely serious.

2. It does not decay with time

A hot lead from three weeks ago is not a hot lead today. If the score does not decay when there is no new activity, every old lead looks equally relevant to every new lead, and the prioritization becomes useless.

3. It weights profile data over behavior data

A lead with a complete profile — budget field filled, job title filled, city captured — gets a high score. A lead with half the fields empty gets a low score. But profile completeness is not the same as buying readiness. A buyer who opened the payment plan section of your microsite three times on a Sunday evening is more serious than one who filled every field on a form and never returned.

4. It is never validated

A score that is never tested against actual outcomes — site visits attended, bookings closed — cannot improve. Most teams never run this validation, so they do not know which parts of their score are predictive and which are noise.

The Three Dimensions of a Good Intent Score

A buyer intent score that actually predicts site visits is built from three dimensions. Each one captures a different kind of signal. The final score is a weighted combination.

Dimension 1: Behavior signals

What the buyer has done on their own — outside of rep conversations. These are the highest-value signals because they reflect the buyer choosing to engage when nobody was prompting them.

• Microsite sessions: number of distinct visits to the property microsite.
• Time on site: total time spent in active engagement (idle time excluded).
• Depth of engagement: did they view floor plans, pricing, payment plans, and amenities?
• Commercial section visits: specifically visits to pricing and payment plan sections — these weight highest.
• Repeat pricing views: a buyer who returns to the pricing page multiple times is signaling seriousness.
• Brochure or payment plan downloads.
• Share events: the buyer forwarded the microsite or brochure to another person.
• Off-hours engagement: evening and weekend sessions correlate with family decision-making.

Dimension 2: Conversation signals

What the buyer has done in interactions with reps or AI. These are still valuable but slightly more noisy — they depend on how the conversation was run.

• Connect success: did the rep actually reach them?
• Call duration: longer qualification calls correlate with higher intent once past a threshold (30+ seconds of real conversation).
• Qualification completeness: did the buyer answer budget, timeline, and buyer-type questions?
• Specific objections raised: objections like "need to check with family" are positive signals — they indicate the buyer is actually considering.
• Site visit scheduling: the strongest conversation signal is a confirmed date on the calendar.
• Response rate to WhatsApp: quick replies are a lightweight but meaningful signal.

Dimension 3: Recency

When the signals happened. A perfect signal from two weeks ago is not the same as an average signal from two hours ago. Recency decay is not optional — it is what makes the score represent the current moment instead of a historical high-water mark.

• Signals from the last 24 hours: weighted at 1.0.
• Signals from 1 to 7 days ago: weighted at 0.6.
• Signals from 7 to 30 days ago: weighted at 0.3.
• Signals older than 30 days: weighted at 0.1 or excluded entirely.

A Reference Scoring Model

Here is a concrete scoring model that has worked across multiple real estate deployments we have instrumented. Treat it as a starting point — the weights should be tuned to your data after you have enough outcomes to validate against.

Behavior scoring (60% of total)

• Microsite visit: +5 points per visit, capped at 20 points.
• Pricing page view: +8 points per view, capped at 24 points.
• Payment plan download: +12 points per event, capped at 24 points.
• Floor plan view: +4 points per view, capped at 12 points.
• Session exceeding 3 minutes: +6 points.
• Off-hours engagement (evenings, weekends): +4 points on top of the base signal.
• Share event (forwarded microsite link): +10 points — this is one of the strongest predictors.

Conversation scoring (30% of total)

• Successful first connect: +8 points.
• Qualification call exceeding 60 seconds: +10 points.
• Complete qualification (budget, timeline, buyer type captured): +12 points.
• Site visit scheduled: +20 points.
• Site visit attended: +30 points.
• WhatsApp response within 24 hours: +5 points.

Recency decay (applied to all signals above)

Multiply each signal by its recency weight. A pricing page view from yesterday contributes the full 8 points. The same view from 10 days ago contributes 8 × 0.3 = 2.4 points. This ensures the score reflects current intent, not historical interest.

Thresholds and Routing

The score becomes operationally useful when it drives routing. Define three or four bands and tie each band to an action.

• Score 80+: surface immediately to a senior rep with context of the signal that triggered the band change.
• Score 50–79: add to today's call queue for the assigned rep, with priority.
• Score 25–49: include in the weekly nurture touch with personalized content based on what the buyer engaged with.
• Score under 25: keep in dormant drip until a new signal raises the score.

How to Validate the Score

A score you do not validate is a score you cannot trust. The validation is simple but non-negotiable.

Step 1: Freeze the score at a point in time

At the end of each week, snapshot the score for every active lead. Store it. This prevents the score from moving retroactively when you look back.

Step 2: Track the outcome 14 days later

For each lead in the snapshot, record whether they attended a site visit within the next 14 days. That is your outcome variable.

Step 3: Plot site-visit rate by score decile

Sort leads by score, split into 10 equal buckets, and plot the visit rate per bucket. A good score produces a steadily rising line. The top decile should have a visit rate at least 5x the bottom decile. If the line is flat, the score is not predictive — reweight and try again.

Step 4: Retune every quarter

Buyer behavior shifts — launch cycles, seasonality, ad creative changes. The weights that worked in January may not work in June. Retune quarterly using the latest outcome data.

Common Mistakes to Avoid

• Capping at the wrong number: if every serious lead hits the max score, the score stops distinguishing between them. Test the distribution and recalibrate.
• Ignoring negative signals: a lead that explicitly says "not interested" should have a meaningful score penalty, not just a lack of positive signal.
• Letting reps override the score silently: if reps can change the score to reflect their gut feel, you lose the dataset. Make overrides visible and audited.
• Treating the score as a final answer: the score is a prioritization signal, not a decision. A rep's context still matters when choosing what to say on the call.
• Not capturing enough behavior data: if your microsite is a static PDF, the score will always be thin. Invest in the measurement infrastructure first.

When the Score Changes How the Team Operates

The deeper impact of a real buyer intent score is cultural. Weekly pipeline reviews change. Instead of reps narrating which leads they feel good about, managers look at the score distribution and ask specific questions: why has this lead been stuck at 75 for two weeks, why did this lead's score jump overnight, why are we calling bucket 5 more than bucket 1. The conversation moves from opinion to data.

Reps also stop defending dead leads. When the score has decayed from 80 to 15 because there has been no activity for three weeks, nobody is defending that lead in a pipeline review. The score does the honest accounting so the manager does not have to.