Arrival Time Windows: How Wide to Promise and How to Keep Them in 2026

TL;DR

The arrival window is the first promise you make to a customer and often the only one they can judge before work begins. Quote all day and you have told them their time is worth nothing. Quote an hour you cannot hold and you have made the same point in a more expensive way.

As of September 2026 the useful question is not "how narrow can we go" but "what width can we resource, and does it change through the day". This guide covers what an all-day window actually costs, how to choose a width from your own arrival timestamps, why afternoon slots need more room than morning ones, how to move from a fixed promise to a live estimate, what to say when you are running late, and how to measure compliance so the number stops being a matter of opinion.

Why the all-day window survives, and what it costs

All-day windows are still common, and the reasons are not stupid ones.

They are operationally easy. A dispatcher with an eight-hour promise can reorder the day freely, absorb an emergency, and never make a call to explain anything. And in trades where jobs genuinely run from twenty minutes to four hours, a narrow window looks like a promise nobody could keep.

The cost is just paid somewhere else, by someone who is not in the room. The customer takes a day off work, or arranges childcare, or cancels something. They wait. Somewhere around hour five they start checking the phone, and by hour six they have decided you do not respect their time. When the technician finally arrives, that opinion is already formed and the first ten minutes on site go into absorbing it.

Then the second-order cost, which is the expensive one. Chronic uncertainty manufactures no-shows: a customer who has waited all day twice will run an errand during the third window, and the missed appointment gets recorded as their failure when the schedule caused it. That loop is the subject of reducing no-shows, and the arrival window is the input most owners overlook when they go looking for the cause.

There is a competitive cost as well: given two similar quotes, a customer takes the two-hour window nearly every time.

How wide should you promise

One honest rule, and everything else follows from it: a narrow window you miss is worse than a wide one you keep.

That is not a licence to quote all day. It means the width has to come from what you can actually do rather than from what sounds good on the phone. A shop hitting a four-hour window ninety-five percent of the time delivers a better experience than one hitting two hours sixty percent of the time. So pick the width from your own data, then narrow it as operations improve — not the other way round.

Two or three hours suits most small service businesses, with the first slot sold narrower and the last sold wider. Which brings us to the part most schedules get wrong.

Window width should follow position in the day

Here is the arithmetic nobody runs, and it is why afternoon customers are disproportionately the ones who complain.

Variance accumulates, and it does not cancel out, because the failures are asymmetric: jobs run long far more often than short, and traffic delays you far more often than it speeds you up. Suppose four consecutive jobs each run twenty minutes longer than scheduled — no catastrophe, just four ordinary overruns. The technician arrives at the fifth appointment eighty minutes late, and there is no single event you can point at as the cause.

What that means for the promise is straightforward. The first job has one unknown in front of it, the morning drive, so it can carry a one-hour window or even a named time and hold it. The second and third have one or two jobs of accumulated variance ahead of them, where two hours is realistic. The last job has an entire day of it, so quoting two hours is a coin toss and quoting three or four is simply honest.

Most booking systems and most dispatchers apply one uniform width across the whole day. That single decision is why the 4pm customer is systematically the one you disappoint, and it is fixable at zero cost by widening the late slots and narrowing the early ones.

The related discipline is not overfilling the day in the first place, because a schedule built on optimistic durations is late before it starts — the argument in technician capacity planning, and upstream of everything here. Multi-technician work adds a layer, because a job needing two people has to align two schedules, each carrying its own variance, so the window must reflect the later of the two arrivals rather than the average — the mechanics are in scheduling two-technician crew jobs.

Stop promising a time, start reporting an ETA

The deepest fix is a change of category. A window is a promise made yesterday. An ETA is a fact reported now.

Customers accept a moving estimate far more readily than owners expect, provided it moves in front of them rather than behind their back. What they will not accept is a fixed promise that silently becomes false while they sit at home believing it.

This is where live location and traffic-aware routing stop being a fleet-management feature and become a customer-experience one. A dispatcher looking at a live map is not guessing whether the 2pm slot still holds; they can see the technician is still on site three suburbs away and act on it forty minutes before the customer starts wondering. That is what GPS tracking is for on the customer-facing side, with the hardware detail on the GPS device page and the wider operational case in GPS fleet tracking for service businesses.

One caution. An automated ETA is only as honest as its data: if the technician has not marked themselves en route, the system is estimating from the schedule rather than from reality, and a confidently wrong ETA is worse than none.

The on-my-way message is the product

For a large share of your customers, the on-my-way text is the entire pre-arrival experience. It is worth writing properly rather than accepting the default template.

A good one does three things. It names the person, because a customer given a name is a customer who opens the door. It gives an estimate framed as an estimate, because "about 25 minutes" survives a traffic light and "arriving at 2:15pm" does not — and a live tracking link is better again, since it moves the customer from waiting to watching. And it says what to have ready: the driveway clear, the dog put away, the vehicle unlocked, the meter reachable, each of which is a minute not lost on site.

Timing matters more than wording. Send it at departure, not at dispatch: at departure the drive is the only remaining unknown, so the estimate is defensible, while a message sent when the job is merely assigned is a guess wearing the clothes of a fact.

Then leave the channel open. A customer who can reply will ask "can you make it after 3" instead of not being home, and that reply has to land somewhere a human sees it. A two-way SMS inbox turns the notification into a conversation, and when nobody is free to watch it, an AI receptionist versus a human receptionist is the trade-off you are choosing between.

When you are going to be late, say it first

You will miss windows. The only question is who speaks first.

The rule is simple and almost nobody follows it: contact the customer before they start wondering. In practice that means as soon as the dispatcher knows the window is at risk, not once it has failed. If the 2pm to 4pm window is in trouble at 1:40pm, the call happens at 1:45pm.

Three things belong in that contact. A new estimate you actually believe — the temptation is to shave it because a smaller delay sounds better, but a customer told "about an hour late" who then waits two hours has been misled twice. The reason, briefly: "the job before yours needed more work than expected" is enough, and customers are far more tolerant of honest overruns than of vagueness. And the reschedule, offered by you, which is the part owners resist and the part that saves the relationship, because offering to move the appointment rather than making the customer ask signals that you know you have cost them something.

There is a hard edge to this. If the day has collapsed and the last appointment cannot be served, cancelling it at 2pm with an apology and a priority slot tomorrow beats arriving at 6:30pm to a customer who has been angry for two hours. What you owe the customer when the cancellation is yours belongs in a cancellation and rescheduling policy.

Measuring window compliance

Everything above is opinion until you have the number, and the number is easy to get.

If technicians mark themselves arrived on site, every completed job has an arrival timestamp. Compare it against the window promised at booking and you have window compliance: the percentage of jobs where the technician arrived inside the window you sold.

Report it three ways. Overall, as the headline that tells you whether you have earned the right to narrow your windows. By slot position, which is where accumulated variance shows up: if the first two slots hit ninety-five percent and the last two hit sixty, you do not have a lateness problem, you have a window-width problem in the afternoon. By technician, carefully and on like work.

Use compliance rather than average lateness. An average of twelve minutes sounds tolerable and can contain a pattern where four jobs in five are early and the fifth is ninety minutes late; the customer in the fifth slot does not experience the average. Grouping promised windows against actual arrivals is standard reporting, and it is what lets you tighten windows on evidence rather than optimism.

Paying for certainty

Once you can hold a window reliably, it becomes something you can sell.

The logic is sound. A first-of-the-day slot is genuinely scarce: there is one per technician per day, and giving it to one customer means nobody else can have it. Holding a route open around a fixed one-hour promise costs real capacity. Charging for that is not gouging, it is pricing a constrained resource.

Three conditions if you do it. State it at booking in plain language, so the standard window is free and the guaranteed slot carries a set fee. Keep the standard window genuinely serviceable, because a premium that exists only because the free option is unreliable is a tax on your own poor scheduling. And resource it before you sell it: a guaranteed window you miss after charging for it is the worst outcome in this entire post.

What this costs

GetTimePad is $79/mo Starter for one staff member, with the smart calendar, online booking portal, automated reminders, customer database, cancellation tracking and basic reporting. $199/mo Pro covers up to five staff and adds live GPS with traffic-based ETAs, Tech Mode, smart review routing, Stripe payments and deposits, automations, Job Readiness scoring, a two-way SMS inbox and staff roles. $499/mo Agency is unlimited staff with multi-location, REST API and webhooks, device management, priority support and custom statuses. Extra seats are $25/mo on Starter and Pro, the SMS bundle is $19/mo, device connectivity is $15 per device per month, and annual billing gives you two months free. There is a 14-day free trial and a live demo with no signup required. Full detail on pricing.

For arrival windows, Pro is the relevant tier: live GPS with traffic-based ETAs makes a day-of estimate factual rather than hopeful, and the two-way SMS inbox catches the reply to your on-my-way message. If you are comparing options, see best dispatch software for service companies.

Where to start

Pull the last three months of completed jobs, compare arrival timestamps against the windows you promised, and read the result by slot position. That one report usually tells the whole story: the mornings are fine and the afternoons are not.

Then widen the last slot of the day by an hour, narrow the first slot, and make the on-my-way message standard at departure. Give it a month and re-run compliance. Once it is above ninety percent you have earned the right to advertise the window as a feature, which is where all of this stops being an operational chore and becomes a reason customers pick you.