LOYI: Intervening Before Customers Are Lost with Churn Prediction

Winning back a customer after losing them is always far more expensive and far more difficult than retaining them before they are lost. A customer who has churned usually doesn’t just stop visiting — they drift to a competing brand. Once the new habit the competitor creates has taken hold, bringing them back to you requires a large campaign investment and a significant time commitment. Reaching out to a customer who hasn’t fully churned yet but has started showing churn signals, on the other hand, is much easier. Their habit hasn’t broken yet; a small reminder or a modest gesture is enough. Loyi’s churn prediction engine kicks in at exactly this point, alerting you to the risk 60 to 90 days before a loss occurs.

What is churn and how is it defined?

The word “churn” refers to a customer ending their relationship with a brand. In the restaurant industry, this definition needs a bit more nuance because the customer has no official “cancel subscription” button. Churn is determined by the customer simply not showing up for a defined period of time. That period varies by category — around 60 days for a cafe, 90 days for a fast casual restaurant, 180 days for fine dining. However, fixed thresholds are often misleading because they don’t account for each customer’s individual rhythm.

Instead of fixed thresholds, Loyi’s AI looks at each customer’s own behavioral pattern. For a guest who comes twice a month, a 30-day gap is a serious alarm signal — yet for a guest who comes once a month, the same gap is completely normal. The system learns each customer’s visit intervals over the past six months, predicts their expected next visit date based on those intervals, and the further their actual behavior deviates from that prediction, the higher their risk score.

What signals are used for prediction?

The signals used by the AI model go well beyond visit intervals. Dozens of different cues combine to form the risk score. The most important signals are: time elapsed since the last visit, average gap between previous visits, whether the average basket size has declined over the last three visits, a drop in campaign message open rates, an increasing gap since the last app login, a decrease in order frequency in a favorite category, the possibility of having switched to a different location, and behavioral shifts such as previously coming with a friend but now coming alone.

Each of these signals is not meaningful on its own; but when they all point in the same direction, the risk score rises quickly. For example, a customer who normally came twice a week, hasn’t visited in nine days, hasn’t opened campaign messages in the last two weeks, and whose average basket has dropped 20% over their last three visits — even if they haven’t yet crossed the churn threshold — carries red-alert-level risk. You’ll see this customer in the “high risk” section of the Loyi dashboard.

The power of early intervention

The biggest advantage of early intervention is that you can act before the customer’s decision becomes final. As long as a customer hasn’t completely lost their habit of coming to you, even a small touchpoint is enough to bring them back. In the opposite scenario — when you try to reach a customer who hasn’t visited in six months — even a high-value incentive often isn’t enough, because the customer has established a new routine with another brand. Internal analyses from the Loyi platform show that customers intervened with 90 days in advance have a return rate two to three times higher than those reached late.

Launching flows based on risk score

The moment a risk is detected, an automated flow is triggered — one you’ve configured in advance to run on its own. A customer at a low risk level receives only a gentle reminder, such as a new menu announcement or a weekend event notification. A customer at a medium risk level gets a personalized offer: a short-duration campaign built around their favorite product. A customer at a high risk level triggers an aggressive win-back flow. This flow runs in several steps — a soft reminder in the first message, a concrete incentive three days later, a more generous offer seven days after that. At each step, if the customer has visited, the flow stops; if not, the next step activates.

The risk score is explainable

Loyi’s churn prediction is not a black box. For every customer flagged as high risk in the dashboard, there are reason cards explaining why they carry that risk. Explanations like “no visits in the last 21 days, yet they normally came every 8 days,” “last three push notifications not opened,” and “average basket dropped 25% over the last three visits” keep you informed. These explanations both increase your confidence in the system and help you fine-tune your intervention strategy on an individual basis. For example, for a customer showing a basket decline, sending a product recommendation campaign rather than a classic win-back may be more effective.

Why this feature is superior to traditional reporting

Traditional loyalty platforms typically provide static lists such as “customers who haven’t come in 60 days.” These lists are useful, but they have two important weaknesses. First, the list informs you after the fact — the 60-day silence has already passed and the customer’s habit has largely been lost. Second, the list views every customer through the same lens, without any truly individualized risk assessment. Loyi’s prediction engine addresses both weaknesses. It alerts you before the loss occurs, and it ensures each customer is evaluated against their own rhythm.

Conclusion

In lost-customer management, the most valuable moment is when the customer hasn’t fully given up yet but the signals are starting to turn red. Loyi’s churn prediction engine alerts you to that exact moment. The automated win-back flow that kicks in afterward re-engages the customer who is about to be lost. This approach significantly reduces your customer acquisition costs, because retaining the guests you already have is always more economical than finding new ones to replace them. The restaurants that survive in 2026’s competitive landscape will be those who continue acquiring new customers while not losing existing ones — Loyi’s churn prediction provides the infrastructure for exactly that two-sided strategy.