Between 35 and 55 percent of drivers who join a regional ride-hailing platform don't reach their third month of activity. A significant share of that attrition — especially in the first 30 days — is not random: it is predictable from the behavioral data the platform generates in the first 7 days. The number of sessions in week one, income per active hour in those first sessions, and completion rate in the first 10 to 15 trips contain enough signal to identify, before day 10, which newly onboarded drivers have a high probability of staying active at 90 days and which are on a behavioral trajectory that leads to departure before month two. The operator who reads those signals in the first week can intervene at the moment when the cost of intervention is lowest and the probability of recovery is highest.

This article is for the operator with 20 to 80 active drivers who sees attrition among new fleet drivers — 30 to 50 percent don't reach month two — without being able to identify which of those departures were avoidable or when the intervention window was open. It covers why new driver churn follows a different dynamic from established driver churn and responds to different interventions, what the three first-week behavioral indicators are that distinguish consolidating drivers from departing ones, what operational ranges signal a new driver at risk of early departure, why the second session is the single most predictive event in the first 14 days, what a specific intervention at days 1, 3, and 7 should include, and how the agent produces weekly monitoring of new driver cohorts. The thesis is practical: first-week data doesn't predict the future — it describes a behavioral state with a measurable probability of leading to departure if nothing changes. The intervention requires neither more money nor more drivers: it requires the new driver to receive, in the first week, the information they need to have a productive session instead of one that confirms the platform isn't worth their time.

Why new drivers leave for different reasons than drivers with months on the platform

The driver who has been on the platform for four months and leaves does so for reasons tied to income per active hour: the platform stopped justifying the time they invested in building their work schedule around it. That driver has prior investment — time spent learning zones, established habits, accumulated ratings — that raises the cost of leaving. Their departure is gradual and detectable three to four weeks in advance, as the previous article in this series describes. The new driver who abandons in the first three weeks has no prior investment. They haven't built habits, have no accumulated history, and carry no established routine. For them, the platform is a hypothesis: can I earn here during my available hours? If the first two or three sessions don't produce an income per active hour that confirms that hypothesis, the cost of not returning is zero. The driver simply doesn't open the app the next day, and the operator receives no signal that anything specific happened.

The practical implication is that new driver attrition does not respond to the same interventions as established driver retention. A repositioning message based on historical zones works for the driver who knows the platform: they understand the reference, know which zones have higher demand, trust the data. A new driver who receives the same message in their first week has no reference frame to evaluate it. What the new driver needs in week one is not retention: it is productive onboarding — a first week that confirms the economic hypothesis before the driver decides the platform isn't worth their time. The difference between intervening in week one with specific positioning information and not intervening isn't in the outcome for the driver who was going to stay anyway: it is in the driver who was on the margin, who had one or two sessions with low hourly income and interpreted that result as a structural characteristic of the platform rather than a positioning problem solvable with concrete instructions.

The three first-week indicators that predict 90-day retention

The data available in a new driver's first 7 days is limited: at most 3 to 5 sessions, 15 to 30 trips, and enough behavioral history to identify initial patterns. Three of those indicators have the highest correlation with 90-day retention in operations of 25 to 80 active drivers in cities of 150,000 to 500,000 residents:

Operational ranges: when first-week numbers signal a driver in the risk zone

In operations of 20 to 80 active drivers in cities of 150,000 to 500,000 residents in Mexico and Central America, first-week behavior for new drivers distributes across three ranges with distinct 90-day retention probabilities. The healthy range — probability of remaining active at 90 days above 60% — corresponds to drivers with 3 or more sessions in week one, average session duration of 3.5 hours or more, income per active hour between 80 and 110 MXN, and completion rate above 80%. The attention range — 90-day retention probability between 35 and 60% — corresponds to drivers with 2 sessions or 3 sessions with average duration below 3 hours, income per active hour between 60 and 79 MXN, or completion rate between 70 and 79%. Drivers in this range have no clear signal in either direction: intervention at days 3 to 7 can shift the trajectory toward the healthy range if the below-range performance is caused by positioning — the most common cause in week one.

The risk range — probability of remaining active at 90 days below 35% — corresponds to drivers with 1 session or 0 sessions after day 5, income per active hour below 55 MXN in all completed sessions, or completion rate below 70% in the first 10 trips. A driver who enters the risk range in week one without intervention has the same probability of remaining active at 90 days as a brand-new recruit: the expected loss from not intervening is equivalent to a full driver replacement, with the additional cost of document verification time and an initial low-productivity period. Variation across markets is 10 to 15 MXN in hourly income thresholds depending on local cost of living and fuel costs, but the behavioral thresholds — sessions, duration, completion — are consistent across regional markets.

The second session: the most predictive event in the first fourteen days

The single most predictive event for a new driver's 90-day retention probability is not the first session: it is the second. A driver who completes a first session and doesn't have a second session within the following 72 hours has a probability of remaining active at day 30 that falls below 40%. A driver who has a second session within 24 to 48 hours of the first, especially if that second session is longer than the first, has a 90-day retention probability above 55%. The mechanism is behavioral: the first session is exploratory. The driver is learning the app, experiencing the process, calibrating how many trips they can make in a session. The second session is the first indication that the driver intends to make the platform part of their work routine. If the second session doesn't occur, the most common cause is one of three: the income from the first session was below the threshold that justifies a second attempt; the first session had a high proportion of idle time that the driver interpreted as insufficient demand for them; or the driver encountered an operational friction — technical, documentary, or in earnings payment — that created an obstacle to the second connection.

The first two causes are measurable from the data. The third requires a direct communication channel: the operator who contacts the new driver after the first 36 hours without a second session and openly asks what happened in the first session has a 2 to 3 times higher probability of recovering that driver than the one who waits for the driver not to show up for 7 days. That difference doesn't arise because the question is especially sophisticated: it arises because the driver who receives direct operator contact within 48 hours of a below-average first session updates their hypothesis about the platform — the operator is paying attention, there is someone who can guide them — before they have made the decision not to return. The driver nobody contacts has no additional information to revise that hypothesis: their only reference is the income from the first session, and if that income wasn't sufficient, the hypothesis closes negative.

The week-one intervention: what to communicate on days 1, 3, and 7

The week-one intervention is not a generic welcome protocol: it is a three-contact cycle with information specific to the driver's performance and the operational context. Each contact answers a different question the new driver has at that point in their experience with the platform:

How the agent produces weekly monitoring of new driver cohorts

The agent query that produces new driver cohort monitoring: 'For drivers who completed their first session in the last 14 days, show me: number of sessions in week one, average session duration, income per active hour in week one, completion rate in the first 15 trips, and whether they had a second session within 72 hours of the first. Group the result by incorporation week and flag in red drivers who meet at least two of the following conditions: fewer than 2 sessions in week one, income per active hour below 60 MXN, completion rate below 75%, or no second session within 72 hours of the first.' The result produces a new driver cohort map without additional metrics: drivers flagged in red are those on the behavioral trajectory leading to departure before month two if there is no intervention in the next 24 to 72 hours. The operator who runs this query every Monday morning has a complete list of new drivers who need a direct contact before the end of that same week.

The 30-day follow-up query that produces the empirical reference frame: 'For each cohort of new drivers — grouped by first-session week — show me the percentage who had at least 3 active sessions in weeks 3 and 4. For cohorts where that percentage fell below 50%, show me the week-one income per active hour and session count for drivers who consolidated versus those who didn't.' That comparison calibrates which first-week thresholds are predictive in the specific operation. Different markets have different residual demand levels and fuel costs that affect income thresholds, but the structure of the pattern — driver with a healthy first week consolidating by weeks 3 and 4; driver with a risk-range first week departing before week 5 — is consistent across regional operations in Mexico and Central America. That query also answers the question most operators can never answer: of the drivers who left in month one, how many gave measurable signals in the first week?

The series of articles that preceded this one — completion rate, assignment radius, proactive positioning, income per active hour — describes an operational improvement cycle that has a structural prerequisite: the driver fleet must stay active long enough for those improvements to produce consistent results. When a platform loses 40 to 50 percent of its new drivers before month two, the operational improvement cycle restarts with new drivers every six weeks. Each restart has a cost: document verification time, an initial low-productivity period, positioning instructions that the veteran fleet no longer needs. First-week monitoring doesn't just reduce new driver attrition: it protects the investment in all prior operational improvements by ensuring the fleet that applies those improvements remains stable.

Drivers who leave in month one generally had a first week that told them the platform wasn't going to work for them — not because the platform actually didn't work, but because they didn't have the information to use it productively. A new driver who starts the first session without a specific zone, without an income reference, and without knowing where demand concentration is going to be in the first hour of their connection is going to experience exactly the idle time and low income that produces the 'this isn't worth it' comparison. The week-one intervention is the same one that benefits the veteran fleet: making the data already in the platform leave the reports and reach the driver at the moment when they can still use it. The only difference is that for the new driver, that moment is the first session, not the fourth week.