What close to a million service visits say about running a route business.
Some businesses run on a route: the same technician, the same properties, on a repeating schedule, with a standard that has to hold between visits. We studied the operating records of more than twenty companies that work this way, to find out where the money actually goes.
The companies are not named here, and will not be. Every figure below is reported at a level that cannot identify a business — rounded totals, no locations, no dates, no company codes. That is the same standard we would apply to your records.
A missed standard has a price. Now we know what it is.
When a visit doesn’t hold the standard, the customer notices. Most of them don’t complain. They remember, and the memory fades slowly. Enough of those and they leave — and they rarely tell you that was the reason.
We measured it against about 2,200 customers who actually cancelled. Each remembered letdown raises the chance a customer quits by roughly 6%.
In money: on a customer who has had no trouble with you, one missed standard destroys $150 to $356 of future revenue. That is three to seven weeks of what they pay you. Preventing it costs about five dollars of materials.
Every square is $5. The single filled square is what the materials cost to hold the standard on one visit. The whole field is what losing that customer costs you instead — thirty squares at the low end, seventy-five at the top.
The part that surprises people: the damage is worst on your best customers. Once somebody has been let down repeatedly, the next one costs you less — they were already halfway out the door.
If you want the technical version. A churn hazard fitted to ordinary visit records, then converted into a lifetime-value penalty: the future revenue destroyed by one more remembered failure, indexed by how much the customer already remembers. Under review at a top-tier journal in operations research.
The fixed weekly round is quietly expensive.
Most route businesses drive the same round every week. It is simple, the crew knows it, and it looks efficient.
We compared it against deciding each week which properties actually need a visit. That decision uses two things: how fast each one is drifting, and how far it sits from the rest of the day’s work.
At the same driving time, the fixed round let 38% more properties fall below standard.
Then we cut the crew’s hours by 40% and ran it again. The smaller crew still beat the full-strength fixed round on all three measures at once.
Both crews were measured over the same stretch of work, on the same properties. The shorter emerald bar is the better result in every row.
The gain here is not working harder. It is sending the crew where the drift actually is, in an order that keeps the driving down.
If you want the technical version. The schedule is modelled as a restless bandit whose activation cost is a routing cost, then triaged by a travel-adjusted index. We prove the policy is asymptotically optimal once client density makes the marginal travel of a visit predictable. Under review at a top-tier journal in operations management.
Planning on averages breaks, and you find out late.
Every schedule rests on an assumption about how fast things drift between visits. Almost all planning uses the average.
The average hides the bad weeks. We built plans both ways, then audited each one against the level of risk the business said it was willing to accept.
Plans built on the average broke that limit on every single test — some by six times over. Plans that carried the uncertainty through held every time. They cost 10% to 15% more, and that premium shrinks the longer your records go back.
Each dot is one planning test. Above the dashed line means the plan broke the risk limit the business had set for itself.
It is also why we quote a range instead of a single number, and say on the page what a plan is assuming.
If you want the technical version. Visit frequencies computed from statistical confidence sets rather than point estimates, with a proved bound on the extra cost of not knowing the true drift rate in advance.
Then we tested it on companies it had never seen.
A model that explains the data it was built on has proved nothing. So we fitted on part of each company’s history and tested against the part we held back — 27 tests across 14 companies.
Predictions landed within about one percentage point of what actually happened. Across all the companies, the model said 9.4% of properties would slip below standard in a given week. The real figure was 9.5%.
Each dot is one company. The model was fitted on part of that company’s history, then asked to predict the part it had never seen.
If you want the technical version. Out-of-sample back-test with both level and shape gates. All 27 splits passed both.
This is the work we would do for you.
Everything above came out of records these companies already kept — visit logs, schedules, and public weather. Nobody installed a sensor. Nobody changed how the crews work. Nobody bought new software to make it possible.
That is where we start with you as well. The questions change with the trade — which jobs run over, which customers have gone quiet, which week next quarter is short-handed — but the method is the same one. Read what your records already know. Put a number on it. Say plainly what it points to.
Three of the traps above turn up almost everywhere, and they are the first things we go looking for:
- A schedule nobody has revisited since the business was half the size.
- A standard that slips quietly, because nobody has ever priced what slipping costs.
- A plan built on an average, which holds right up until the week it doesn’t.
Whether any of that is your situation is your call — you know your operation better than we ever will. Our half is putting a number on it.
Bring us one question you can’t answer.
Point at the thing you’d want to know — which jobs make money, which customers are drifting, whether next month is covered. We’ll tell you straight whether it’s answerable from what you already record.
Book a free call →