Callback rate is the maintenance metric that tells you if the work was actually done

A closed work order and a solved problem are different events, and most maintenance reporting only counts the first one.

Callback rate is the metric that separates them. It is the percentage of work orders that generate a second work order for the same issue in the same unit within a defined window, and it is the closest thing maintenance has to a quality measure.

Closure rate rewards speed and punishes nothing

Consider what happens when closure rate and time-to-close are the only numbers on the board.

A tech with a slow toilet at unit 302 has two options. Jiggle the flapper, close the ticket, ten minutes. Or pull the tank, find the failed fill valve, replace it, forty minutes, plus a trip to the shop for the part.

Both close the ticket. One closes it four times faster. The first one produces a new work order in nine days, which also closes, and that second closure improves the numbers again.

So a portfolio measuring only closure is quietly paying for the flapper jiggle twice and calling it productivity. Callback rate is the correction, because it makes the second visit visible as a failure of the first rather than as additional throughput.

Define the window before you measure anything

The definition decides the number, so this is where the work is.

Window. Thirty days is the common choice and a reasonable default. Long enough to catch a bad repair, short enough that unrelated failures do not pollute it. Some categories deserve longer: an HVAC repair in April that fails in July was probably a bad repair, and a 30-day window will miss it.

Same issue. The hardest part. A follow-up work order in the same unit for a different problem is not a callback. Matching on category rather than on free text is the practical approach, imperfect in both directions.

Same unit. Necessary but insufficient on its own, which is why category matching matters.

Exclusions. Multi-visit work that was planned as multi-visit is not a callback. Neither is a repair pending a part, or a repair the resident refused access for. If you do not exclude these, techs will learn the metric is unfair and stop respecting it, which costs you more than the metric was worth.

What the number should be, roughly

Published benchmarks for multifamily callback rate are thin, and anyone quoting a precise industry figure should be asked for the source. What is defensible:

Under 5% is good. Between 5% and 10% is normal for a portfolio not tracking it. Above 15% means something structural, and it is usually parts availability or scoping rather than skill.

The more useful comparison is internal. Your own rate this quarter against last quarter, and your properties against each other, controlling for age. A single property three times the portfolio average is a finding you can act on this week. An industry benchmark is not.

The causes are mostly not skill

When operators dig into callbacks, the pattern that shows up repeatedly is that the tech knew the right fix and did not do it.

No part on the truck. The most common cause. The tech patches to close the resident's immediate problem, intends to come back, and the intent evaporates. This is an inventory problem wearing a quality problem's clothes.

Time pressure from the queue. Fourteen open orders makes forty minutes on one toilet feel indefensible.

Wrong scope on intake. "Toilet running" dispatched as a five-minute job when it was a valve replacement. The tech arrives with the wrong expectation and the wrong parts.

Genuine training gaps. Real, and the smallest category. Usually concentrated in HVAC and in appliance diagnosis rather than spread evenly.

Reading callback rate by cause rather than by tech is what makes it useful. Three of those four causes are fixed by the operation rather than by the technician.

Parts availability is the intervention with the best return

If you fix one thing, fix the truck.

A standard stocked parts list per tech, checked weekly, built from your own top failure categories, removes the largest single cause of callbacks. It is unglamorous and it is cheaper than any training program.

Our own honest limitation here: GetDone does not do parts inventory. No stock levels, no reorder points, no truck stock. Categories and history will tell you which parts you need, and the stocking itself is a separate system or a spreadsheet. Anyone selling maintenance software as an inventory solution should be asked what happens when the count is wrong.

What we do track, and what we do not

Repeat work orders in the same unit are visible in history, and category-based reporting will show you where they cluster. Automatic callback classification against a configurable window is not something we ship today. The report exists at the category and property level, and the per-tech attribution with exclusions has to be a review conversation rather than a dashboard number.

That is a real gap, stated because a vendor that pretends its reporting is complete is a vendor you will discover the gaps in later.

Pull last quarter and count by hand

Take one property, one quarter, and one category. HVAC is the best place to start.

Count units with two or more work orders in that category within 30 days of each other. Divide by total orders in that category.

That number took twenty minutes to get and it is more useful than any benchmark. What was it?