Guides
Getting window cleaning KPIs right the first time
Window cleaning KPIs: eight monthly measures with formulas, sources and the action each triggers, plus the three things these numbers cannot tell you.
What to take away
- Track eight numbers monthly. Each one has a formula, a source, and an action it triggers.
- A measure with no threshold is a hobby. Set your own baseline from three months of your own data and act on movement against it.
- No benchmark is given here, and any article that gives you one is describing a different route in a different city.
- Two of these numbers are about cash rather than profit. A profitable company can still run out of money.
- Look at all eight once a month, together. Watching one weekly produces reactions rather than management.
This is general business guidance rather than accounting advice. Take qualified advice on how any of this applies to your own accounts.
How to use the list
For each measure: write the formula, name the source of each input, record it monthly in the same place, and write down in advance what you will do when it moves. That last part is what turns a number into management.
Three months of data gives you a baseline. Use your own median rather than anybody's benchmark, because your geography and service mix decide the level.
The eight measures
1. Revenue per working day. Revenue for the month divided by days actually worked. Source: invoices and the schedule. Watch it against the same month last year rather than against last month, because seasonality dominates. When it falls with the same number of jobs, the mix has shifted toward smaller tickets.
2. On-site hours as a share of paid hours. On-site hours divided by total paid hours. Source: job capture. This is the route efficiency measure, and it is the one that quietly explains most bad months. A fall means driving, waiting or setup has grown.
3. Contribution per productive hour. Revenue minus materials, divided by all crew hours including travel and setup, by job. Source: job capture and purchase records. Group the results by service type and by area. This is the measure that tells you what to sell more of.
4. Recurring share of revenue. Revenue from signed frequencies divided by total revenue. Source: agreements and invoices. The single best indicator of whether the business has a future beyond your own effort. A falling share means first visits are not being converted.
5. Rework rate. Return visits divided by jobs completed. Source: job records. A return visit is a full cost with no revenue, so this measure is worth more attention than its size suggests. Concentrated on one crew it is a training signal; concentrated on one day it is a scheduling signal.
6. Wasted visits. Visits where access failed, divided by scheduled visits. Source: job records. Almost always fixable through the agreement rather than through effort, and almost always ignored because each one feels like bad luck. Access failures are one of the five causes of delay that scheduling window cleaning work has to hold against.
7. Days to collect. Average days from invoice to payment received. Source: invoicing and bank records. Profit and cash are different things, and this measure is why a good month can still be tight.
8. Retention. Accounts still active at the end of the period as a share of those active at the start. Source: the customer list. Losing accounts quietly is the most expensive thing that can happen to a route, because replacing them costs acquisition and density at the same time.
What each one should trigger
| Measure | When it moves against you | The first thing to check |
|---|---|---|
| Revenue per working day | Falls with jobs unchanged | Service mix and pricing |
| On-site share | Falls | Route density and schedule build |
| Contribution per hour | Falls on one service | Time on site, travel, scope creep |
| Recurring share | Falls | Whether the rebooking question is being asked |
| Rework rate | Rises | Training, or a schedule with no slack |
| Wasted visits | Rises | Access arrangements out of date |
| Days to collect | Rises | Invoicing timing and follow-up routine |
| Retention | Falls | Missed visits, price changes, quality drift |
Check the named cause before doing anything else. Most measures have one dominant explanation in this trade, and hunting for a complicated one wastes a month.
The eight measures
- 1Revenue per working day
- 2On-site hours share of paid hours
- 3Contribution per productive hour
- 4Recurring share of revenue
What these measures cannot tell you
They are all lagging. Every one of them describes what already happened, and none of them warns you.
Three things they miss entirely. Customer sentiment, which shows up in retention months after the goodwill actually went. Safety, which is not a measure at all until something goes wrong. And crew morale, which precedes both rework and turnover and is visible only by talking to people.
So treat the eight as instruments rather than as the whole dashboard. A month where every number looks fine and the crew is unhappy is not a good month.
Where the inputs come from
None of this exists without per-job capture: arrival, departure, travel, materials, revenue, and whether a return visit happened.
The Internal Revenue Service recordkeeping guidance describes the standard a system has to reach before it shows income and expense clearly, which is the same standard these measures need to be trustworthy over a year.
The system holds customer addresses, access codes and photographs of people's homes.
small business cybersecurity quick start guides from the National Institute of Standards and Technology and small and medium business material from the Cybersecurity and Infrastructure Security Agency set out basics. A small company can act on them. They include individual accounts, multi-factor authentication, tested backups, removing access the day somebody leaves.
Reviewing them properly
Once a month, same day, all eight together, in one sitting.
Write one sentence per measure: what it did and why you think so. Then choose one action. Companies that decide on three actions from one review usually complete none of them and cannot attribute the result of any.
What the numbers connect to
Each measure points at a decision made somewhere else.
Contribution per hour and revenue per working day are inputs to pricing and to what you choose to sell. The definitions, the thresholds and the system that produces them are in the software and KPI guide.
Rework and wasted visits are operational. Fixes are people and equipment, not analysis: training and sign-offs, covered by hiring and training window cleaners, and gear that works, from equipment and setup choices.
Retention that falls after a crew change is usually a hiring question. The process is in the guide to hiring reliable window cleaning staff.
Common questions
What is a good rework rate?
There is no figure worth quoting. Measure your own for three months, take the median, and treat sustained movement above it as a signal. Comparing yourself to a published number tells you nothing about your buildings.
Should I share these numbers with the crew?
Share the ones they influence, defined clearly and without blame: rework, wasted visits, on-site share. Sharing revenue figures with no context tends to produce anxiety rather than better work.
How long before the numbers mean anything?
Three months for a baseline, a year before seasonality stops confusing you. Keep recording from the beginning anyway, because you cannot backfill job times.
Which one should I fix first if several are bad?
Wasted visits, then rework. Both are pure cost with no revenue attached, both are usually fixable inside a month, and fixing them frees the capacity you need to work on everything else.
