Why average tenant response time doesn't tell you where things go wrong

Friday afternoon, half past three. You call back about a damp report from last Monday and hear on the other end: "Oh, you're the first person to get in touch." Not angry, not aggrieved, just tired. She hasn't been sitting there for four days with a stopwatch. She's simply switched off.

Two days later, your weekly report shows that your response time on maintenance requests that week was 1.8 days. Below target. Green figure. You treat it as accurate. And in a sense it is, because you did pick up 80 per cent of your requests within 24 hours. The problem sits in the other 20 per cent. And that other 20 per cent is almost always the same group of people.

The average is a shameless liar

An average response time under two days looks fine on a dashboard. But averages are a disastrous thing to steer tenant communication by, because the satisfaction losses concentrate in the tail. You don't lose a tenant because you helped 80 per cent of people quickly. You lose her because, on the one request that went wrong, she heard nothing for four days.

This pattern shows up in the complaints Woonbond received in the first half of 2025: one in five people who file a complaint say their landlord doesn't respond to maintenance issues, and 27 per cent of all complaints concern a landlord who simply never comes back. That isn't 20 per cent of cases scattered at random, it's a fixed group of reports that structurally slips past the system.

The Aedes benchmark 2024 makes the pattern even clearer for repairs. When a request is resolved in a single visit, tenants rate the process 8.5 on average. When four or more visits are needed, that score drops to 6.6. A drop of almost two full points on a ten-point scale, carried by a fraction of your requests.

The tail isn't random, it's a cohort

The second problem with steering by averages is that it makes you think the tail is random. As if, every week, a random handful of requests just happen to take longer. That's almost never the case. When you break down your incoming requests, the same profile tends to come back.

Three cohorts where most organisations lose track of the tail:

Non-urgent requests through the wrong channel. A tenant emails on Tuesday about a damp problem that isn't urgent. No automatic confirmation, no ticket number, no status. On Thursday she calls because she's heard nothing. The front desk logs it as new, because they can't find the email. By Friday it exists somewhere, but now in two systems.

Repeat reports on an open ticket. The first report moves fast: an engineer visits, the ticket is set to "waiting for materials", closed in the count. The tenant hears nothing for three weeks and calls again. That second contact often gets logged as a new request, with a new response-time clock. On paper your service level looks fine, in reality she's waited nine weeks.

Requests that depend on a third party. The contractor schedules the first appointment fine, but the second appointment shifts without anyone telling the tenant. Between you and the contractor, it goes quiet. The tenant only sees the silence.

In practice, these three cohorts account for a large part of your tail. And they share one thing: they arise at the handover points in your process, not at the first response.

What you should measure instead

The question isn't "what's my average response time". The question is "how many requests are currently waiting longer than five working days for a first or follow-up response". That's a different measurement, with a different lever to pull.

If you build the figure as a percentile instead (P95 rather than an average), it forces you to look at the tail. The Aedes benchmark 2025 shows the sector average sitting stable at around 7.7. Stable, not rising. The gain isn't in adding another tenth of a point to the average, it's in the tenants who currently give a 4 or a 5 and drag the whole score down.

In practice, that means a daily list of requests older than five days, split by the three cohorts above. Not to hold someone to account against a KPI, but to see where the handover stalls. It's often the same handover, week after week.

The conversation that changes this

The hardest part of this shift isn't the measurement, it's the conversation with the organisation. Someone has set your average response time to green, and you come along to say that the figure they hit isn't what tenants actually experience. That rarely goes down well.

What helps is placing the cohort data alongside it. Not "our average is lying", but "these are the six requests from last week that have been open for more than five days, and in four of them the hold-up sits at the same step". Then the conversation stops being about whether the figure is correct and starts being about where the organisation is stalling. And that's the conversation that actually helps the tenant.