How AI Risk Scoring Predicts No-Shows Before They Happen

Most agents react to no-shows after they happen. They arrive at the property, wait 15 minutes, then drive home. The viewing is wasted. The owner is frustrated. The agent adjusts nothing for the next lead.

A better approach is prediction. If you know which leads are likely to ghost before the viewing, you can intervene — send an extra reconfirmation, call instead of text, or deprioritise the viewing relative to higher-probability ones.

Fox's risk scoring model predicts no-show probability with 78% accuracy using behavioural signals available at the time of booking and during the pre-viewing window. This article explains what those signals are, how they combine, and how agents use the scores to change outcomes.

Why Prediction Beats Reaction

The reactive approach to no-shows has a fundamental problem: by the time you know the lead is not coming, the cost is already incurred. You have already driven to the property. The owner has already prepared. The slot is already lost.

Prevention — through reminders, reconfirmations, and logistics messages — reduces no-shows significantly. But prevention treats every lead the same. The lead who replied "see you tomorrow!" at 9pm last night gets the same reminder sequence as the lead who has not engaged since booking three days ago.

Prediction adds a layer of intelligence. It tells you which leads need more attention and which are fine with the standard sequence. This allows agents to allocate their limited pre-viewing time where it makes the most difference.

The gap between high-risk and low-risk leads is stark. In our data, leads scored as high-risk no-show at 45%. Leads scored as low-risk no-show at 14%. That 3.2x difference means the high-risk leads are where intervention effort should concentrate.

The Signals

Fox's risk model evaluates six categories of behavioural signals. Each contributes a weighted score, and the combined score maps to a risk label: low, medium, or high.

Signal 1: Booking Lead Time

The interval between when a lead books and when the viewing is scheduled is one of the strongest predictors.

Short lead time (0-24 hours): 19% no-show rate. Leads booking same-day or next-day viewings have high intent. They are actively searching and want to see the property soon. The short window also limits the number of competing commitments that could displace the viewing.

Medium lead time (2-4 days): 27% no-show rate. The standard window. Leads have time to book other viewings, reconsider their priorities, and forget about the commitment.

Long lead time (5+ days): 38% no-show rate. The longer the gap, the more likely the lead's circumstances or interest will change. A viewing booked 7 days out has survived a full week of competing priorities, schedule changes, and new listings. Many do not survive.

The model weights booking lead time at approximately 20% of the total score. It is the single most predictive at-booking signal.

Signal 2: Message Engagement

How the lead interacts with pre-viewing messages is the most dynamic signal — it updates in real time as the viewing approaches.

The model tracks three engagement metrics:

Reply velocity: How quickly does the lead respond to messages? Leads who reply within 30 minutes have a 12% no-show rate. Leads who take 6+ hours have a 31% rate. Leads who never reply have a 44% rate.

Reply depth: Does the lead send substantive replies ("Yes, I'll be driving — is there parking nearby?") or minimal ones ("ok")? Substantive replies indicate active engagement and planning. Minimal replies indicate passive acknowledgment.

Message read receipts: On WhatsApp, blue checkmarks confirm the message was read. A message that is delivered but unread for 4+ hours during waking hours is a negative signal. A message that is read but not replied to is a weaker negative signal — the lead is aware of the viewing but not engaged enough to respond.

Message engagement carries approximately 30% of the total risk score. It is the highest-weighted category because it is both highly predictive and actionable — a lead who is not engaging can be targeted with an extra touchpoint or a phone call.

Signal 3: Time-of-Day Patterns

When the lead booked and when the viewing is scheduled both influence no-show probability.

Late-night bookings (10pm-2am): 34% no-show rate. Leads browsing property listings late at night are often in a different mental state than they will be at viewing time. The booking is aspirational, not operational.

Working-hours bookings (9am-6pm): 24% no-show rate. Standard baseline.

Morning viewings (before 10am): 31% no-show rate. Higher than afternoon viewings (25%) because morning viewings require advance planning (setting alarms, adjusting morning routines) that afternoon viewings do not.

End-of-day viewings (after 6pm): 29% no-show rate. Slightly elevated because work-day fatigue and last-minute schedule overruns create friction.

Time-of-day contributes approximately 10% of the risk score. It is a secondary signal — useful in combination with others but not strongly predictive on its own.

Signal 4: Reschedule History

Leads who have already rescheduled the viewing once are, counterintuitively, slightly less likely to no-show than first-time bookings. Their no-show rate is 21% vs 28% for first-time bookings.

The reason: rescheduling requires active engagement. A lead who cancels a Tuesday viewing and rebooks for Thursday has demonstrated both willingness to communicate and continued interest in the property. They did not ghost — they adjusted.

However, leads who have rescheduled twice or more have a 39% no-show rate. Multiple reschedules signal chronic indecision or low commitment. The lead may continue rescheduling indefinitely without ever attending.

Reschedule history contributes approximately 10% of the risk score.

Signal 5: Source and Channel

Where the lead came from and through which channel they booked affects their baseline commitment level.

Agent-referred leads: 18% no-show rate. A personal referral from the agent carries social weight that portal leads lack.

Portal leads (PropertyGuru, 99acres, Zillow, etc.): 32% no-show rate. High volume, lower intent. Portal browsing generates many casual enquiries.

Social media leads: 35% no-show rate. Leads from Instagram or Facebook property ads have the lowest baseline commitment. The booking was one click away from scrolling past.

Repeat leads (previously viewed another property with same agent): 15% no-show rate. Prior relationship and established communication patterns significantly reduce ghosting.

Source contributes approximately 15% of the risk score.

Signal 6: Property and Viewing Context

Certain viewing characteristics correlate with higher or lower no-show probability.

Exclusive viewings (one-on-one): 23% no-show rate. The lead knows the agent is coming specifically for them, which increases social obligation.

Open house viewings: 33% no-show rate. The perceived social cost of not attending is lower because the lead is one of many expected visitors.

High-value properties: 19% no-show rate. Similar to the transaction-stakes factor in market benchmarks — leads viewing expensive properties are typically more serious.

Rental viewings: 29% no-show rate vs 22% for purchase viewings. Rental leads tend to book more viewings and have lower switching costs.

Property context contributes approximately 15% of the risk score.

How Fox Uses the Score

The risk score is not just informational. It drives three automated behaviours in Fox's coordination system.

Adaptive Reminder Intensity

Low-risk leads receive the standard three-touch reminder sequence (24h, 2-3h, 90min). Medium-risk leads receive the same three touches plus an additional midpoint check for bookings with 3+ day lead times. High-risk leads receive the full sequence plus a flag to the agent recommending a phone call 4-6 hours before the viewing.

This graduated approach ensures that low-risk leads are not over-messaged (which can feel intrusive and actually increase no-shows) while high-risk leads get the extra attention they need.

Smart Followup Scheduling

Fox's assurance engine uses risk scores to determine the urgency of post-booking followups. High-risk bookings trigger an earlier confirmation check — 36 hours before instead of 24 — giving the agent more time to react if the lead does not confirm.

If a high-risk lead fails to respond to the 36-hour check, Fox escalates to the agent with a recommendation: call this lead or consider rescheduling the viewing. This early warning has prevented thousands of wasted trips in our data.

Dashboard Prioritisation

On the Fox dashboard, upcoming viewings are sorted by risk score, with high-risk viewings flagged visually. This lets agents scan their schedule and immediately see which viewings need attention and which are likely to proceed smoothly.

The visual risk indicator is deliberately simple — a red/amber/green badge next to each viewing. Agents do not need to understand the scoring model. They need to know which leads to call.

What Agents Do Differently With Early Flags

The risk score changes agent behaviour in four measurable ways:

1. Selective Pre-Viewing Calls

Agents who see a high-risk flag are 4x more likely to call the lead before the viewing. A 2-minute phone call converts 35% of high-risk leads into confirmed attendees or explicit cancellations. Without the flag, agents rarely make pre-viewing calls — the overhead does not seem justified when you assume every lead will show.

2. Schedule Restructuring

Agents with multiple viewings in a day use risk scores to sequence them. High-risk viewings are scheduled between low-risk ones, so a no-show creates a gap that can be absorbed without derailing the entire day. Some agents double-book high-risk slots — scheduling two leads for the same time in the same area — knowing that one is likely to cancel.

3. Owner Communication

When a viewing is flagged as high-risk, agents proactively tell the owner: "I have a viewing confirmed for 3pm but the lead has been unresponsive — I'll confirm by noon whether it's going ahead." This manages the owner's expectations and protects the agent's credibility. It is far better than showing up with no lead and explaining after the fact.

4. Pipeline Prioritisation

Over time, agents who use risk scores start filtering their pipeline earlier. Instead of booking every lead who requests a viewing, they invest in a brief qualifying conversation first — especially for leads from high-volume portal sources. A 3-message WhatsApp exchange before booking separates serious leads from casual browsers and prevents the no-show before the viewing is even scheduled.

What Risk Scoring Does Not Do

The model is a heuristic, not a crystal ball. Important limitations:

It does not predict individual outcomes with certainty. A 78% accuracy rate means 22% of predictions are wrong. Some low-risk leads will ghost. Some high-risk leads will show up enthusiastically. The model improves decision-making on average, not in every case.

It does not replace good practices. Risk scoring without a solid reconfirmation process is a dashboard feature, not a no-show solution. The score tells you where to focus. The reconfirmation scripts and reminder timing do the actual work.

It does not penalise leads. The score is internal to the agent. Leads never see their risk label. A high-risk lead receives more attention, not less — extra messages, a phone call, a more personal touch. The goal is to help the lead show up, not to gatekeep access to viewings.

It improves over time. As Fox accumulates more viewing data, the model's signal weights are recalibrated. Early versions relied heavily on booking lead time. Current versions weight message engagement more heavily because it has proven more predictive in production. The model you see today is more accurate than the one from six months ago.

Compared to generic scheduling tools that treat every booking identically, risk-aware coordination fundamentally changes how agents allocate their most scarce resource: time. Instead of spreading attention evenly across all viewings and hoping for the best, agents concentrate effort where it produces the greatest reduction in no-shows.

The result is not just fewer no-shows. It is a calmer, more predictable practice. Agents who use risk scoring report less stress about their schedules because they know, before leaving the house, which viewings are solid and which ones need a phone call first.

That confidence — knowing what your day actually looks like before you start driving — is what prediction gives you that reaction never can.