When my team shipped a major change to a field-services app for a client, a family-run plumbing business, about half the trucks were still running the previous version. Nobody had done anything wrong. A plumber opens that app to log the materials used on a job, then puts the phone away. Update prompts are easy to ignore.
This isn’t limited to one plumbing crew. Apple reported in June that iOS 26 was running on 79% of all iPhones, nine months after it launched. If about one in five iPhone users skip a free system update, plenty of people will skip yours.
For consumer apps, that’s an annoyance. For apps that run payroll, invoicing, or dispatch, old versions in the field can slowly corrupt the numbers a business runs on.
WHY DON’T FIELD CREWS UPDATE THEIR APPS?
Field crews likely don’t update their apps because they see them as a tool, and nobody updates a tool that still seems to work.
The plumbers on this project weren’t especially tech-savvy, and they didn’t need to be. Our project manager found that small design choices decided whether the app got used at all.
Forcing updates wasn’t a realistic answer either. There’s no reliable way to push an update to someone who only opens the app to log materials.
So the question changes. Instead of asking how to get everyone updated, ask what the old version will see when the new one ships.
CHANGE THE DATA WITHOUT BREAKING OLD VERSIONS
The safest way to change an app’s data is to keep answering old versions in the format they expect. Our change was simple on paper. The original app assumed one supplier visit per job. Real jobs don’t work that way, so we rebuilt jobs to support multiple purchase orders and reimbursements.
Old versions of the app only understood one purchase order. Sending them the new format would have broken them on the truck.
Our engineers built a translation layer on the server. Each app tells the server which version it is. New versions get the full data. Older versions get the same job flattened back into the single purchase order shape they understand.
The rule we took from it is that every change to a field app needs an answer to one question before it ships. What does the old version see?
STORE THE INPUTS, THEN RECALCULATE THE TOTALS
The second decision on this project protected the payroll numbers.
The common approach is to store each plumber’s weekly pay total and update it as things change. In our engineer’s experience, every payroll system built that way eventually drifts. A corrected invoice price stops matching the commission on screen. An archived material still counts toward someone’s pay.
We treat payroll as a recalculation instead. When anything that affects pay changes, the system rebuilds that week’s payroll from the underlying jobs, materials, and commission rates.
Completed invoices are locked, so a finished week can’t shift underneath anyone. Any week’s payroll can be regenerated from its source records, which gives the business an audit trail. Manual adjustments sit on top of the computed numbers instead of fighting them.
This matters more when several app versions are writing data at the same time. Totals built from source records stay consistent even when the apps sending those records don’t match.
SLOW APPS GET SKIPPED TOO
A slow app gets used about as little as an outdated one.
For our plumbing business client, plumbers pick parts from a 781-item materials list, and each job uses 10 to 30 of them. The complaint we heard was that the app got slower the longer they used it on big jobs.
The cause was a memory leak in the search box. Every keystroke redrew every visible row in the list, and the small leaks added up over a big job. Our engineers stopped the redraws and stabilized the code behind the list.
QUESTIONS TO ASK BEFORE SHIPPING A FIELD APP CHANGE
These come straight from the decisions on this project.
1. What does the oldest version still in use see after this change? Name that version before the change ships. On the plumbing project, half the trucks were on the old single purchase order version, so the server kept sending them that format.
2. Can the server answer old and new versions at the same time? That depends on each app telling the server which version it is. If yours doesn’t do that yet, add it before the next change, because every later fix depends on it.
3. Are totals like pay, invoices, and inventory recalculated from source records or stored and patched? Stored totals look fine until someone corrects an old record. Then the total and the records disagree, and nobody can say which one is right.
4. Which records get locked once they’re final? Locking is what makes recalculation safe. Once an invoice on our project is marked completed, it can’t be edited, so that week’s pay can’t change later.
5. How does the app perform on the biggest job, not the average one? Test with the longest jobs your crews run. The plumbing app’s slowdown showed up on big jobs with up to 30 parts, which a quick demo never reaches.
BUILD FOR THE CREW THAT NEVER UPDATES
Most teams plan an app change around the people who will install it. In my experience, the harder planning is for the people who won’t. On a field app, those are often the crews logging the hours and parts that payroll and invoices depend on. Supporting both versions took extra server work on the plumbing project. It also meant the half of the fleet still on the old version could keep logging jobs without changing anything.
Daniel Haiem is CEO of AppMakers USA.