Offline-First Apps for Field Teams: Sync, Conflicts and What to Decide Before Building
November 11, 2026 · 8 min read · Etzal Group Team
A plantation company equips its field assistants with a harvest recording app. In the office demonstration it is excellent. In the blocks, eight kilometres from the nearest tower, it shows a spinner and then an error. Assistants go back to paper and type everything in at the mess in the evening, when the signal is good. The app works. It was simply built for a place where the work does not happen. Offline-first apps are designed the other way round: from the field inward.
Online-first and offline-first
An online app reads from and writes to a server on every action. Without a connection, it stops.
An online app with a cache remembers what was recently viewed, so some screens open offline. Writing usually still needs a connection.
An offline-first app keeps a working copy of the needed data on the device. Every read and write happens locally and at once. A background process exchanges changes with the server whenever it can.
The difference is in the assumption. Online-first treats disconnection as a fault. Offline-first treats it as an ordinary condition of work.
When you need it
- Work in plantations, mines, forests, at sea or in rural areas.
- Inside buildings with poor coverage: basements, warehouses, plant rooms, hospitals.
- Sales and delivery staff moving through patchy coverage.
- Any task where waiting for a screen to load costs money or patience.
When you do not: office staff on reliable connections, or tasks that only make sense with live data, such as checking a bank balance.
How it works
Four parts.
1. Local storage. A database on the device holding the data this user needs: their assigned jobs, the customer list for their route, the item catalogue, their own recent entries.
2. A change queue. Each thing the user does offline is recorded as a pending change with a timestamp and a unique identifier.
3. Synchronisation. When connected, the app sends its queued changes to the server and fetches changes made elsewhere since the last sync.
4. Conflict handling. Rules for when the same data was changed in two places.
The user should barely notice any of this, apart from a small indicator showing whether everything has been sent.
What to keep on the device
A phone cannot hold the whole company database, and should not.
- By user and role. A technician gets their own work orders, not everyone's.
- By time. The next two weeks of jobs, the last month of history.
- By area. The customers on this route, the assets at this site.
- Reference data. Price lists, item codes, checklists, forms.
Decide deliberately. Too little, and the user is stuck in the field without the record they need. Too much, and sync is slow, storage fills, and more sensitive data sits on devices than necessary.
The five conflict cases
This is where projects succeed or fail, and the answers belong to the business.
1. Two people edit the same record. A supervisor changes a job's priority in the office while the technician adds notes in the field.
Options: merge field by field, since they changed different fields. Or the latest change wins. Or the supervisor's change wins by role.
2. Two people edit the same field. Both change the scheduled date.
Options: latest wins, with the overwritten value kept in history. Or show both to a named person to decide.
3. A record is edited offline after being deleted or closed on the server. The technician adds a report to a job that was cancelled an hour earlier.
Options: reject the change and tell the user. Or accept it and flag it for review. Silently discarding it is the worst choice.
4. Limited quantities. Two salespeople, both offline, each sell the last five units of the same item.
No sync rule fixes this. Options: allocate stock per salesperson in advance, accept offline orders as "pending confirmation", or allow overselling and handle the shortage by process.
5. Sequences and numbers. Invoice numbers must be consecutive, and two offline devices cannot both know the next one.
Options: give each device its own range or prefix, or assign the official number at sync time and show a temporary reference until then.
Write down the rule for each kind of data before development starts. "Latest wins" is acceptable for some data and dangerous for others.
Prefer designs that avoid conflicts
The best conflict is the one that cannot happen.
- Record events, not states. "Counted 48 at 10:15 by Andi" is an addition. Two additions never conflict. "Stock = 48" is an overwrite that can.
- Give records one owner. A job assigned to one technician is edited by that technician.
- Split records by responsibility. The supervisor edits the plan. The technician edits the report.
- Create, rarely edit. Field work is mostly new entries: inspections, readings, orders, photos.
A design built on appending new records needs very little conflict resolution.
Making sync dependable
- Unique identifiers created on the device, so a record has its identity before it reaches the server.
- Safe to repeat. If a sync is interrupted and retried, nothing is duplicated. The server recognises a change it has already applied.
- Small steps. Changes are sent in batches, so a dropped connection loses little.
- Resumable uploads for photos and files, and compression. Images are the heaviest part of most field apps.
- Sync on the right connection. Let users choose to send large files only on Wi-Fi.
- Device time is not trusted. Phone clocks are wrong surprisingly often. Record the device time, and let the server add its own.
- Order matters. A new customer must arrive before the order that refers to it.
Tell the user the truth
People trust an offline app only if it is honest about its state.
- A clear indicator: all sent, 3 items waiting, sync failed.
- Per-record status where it matters: saved on device, sent.
- Last sync time.
- A manual "sync now" button.
- Clear messages when something is rejected, with what the user should do.
- Never lose an entry silently. If the server refuses a change, keep it on the device and show it.
The worst outcome is a field worker believing a report was sent when it was not.
Security on the device
Data on a phone is data that can be lost with the phone.
- Encrypt local storage.
- Require a PIN or biometric to open the app, with a timeout.
- Limit what is stored to what the role needs.
- Remote wipe, or automatic removal of data when a device has not connected for a set period.
- Revoke access centrally when someone leaves.
- Decide how long offline login is allowed. A week without contacting the server may be reasonable. Indefinitely is not.
Personal data on devices falls under data protection law like any other, so keep the amount small.
Testing
Offline behaviour cannot be tested on office Wi-Fi.
- Airplane mode for hours of real use, then reconnect.
- Bad connections: slow, dropping, switching between networks mid-sync.
- Kill the app mid-sync.
- Two devices changing the same record, for each conflict rule.
- Large backlogs: a week of entries and two hundred photos.
- A full device, and a device with the wrong date.
- In the actual field, with the actual users, before rollout.
Questions to settle before building
- Which tasks must work with no connection, and for how long?
- What data does each role need on the device?
- For each kind of data, what is the conflict rule?
- What can be created offline, and what needs live confirmation?
- How large are photos and files, and when are they sent?
- How are numbers and sequences assigned?
- What happens to a lost device?
- How does the user know what has and has not been sent?
A vendor who starts building without asking these will discover the answers in production.
What it costs
Offline-first typically adds a significant share to the cost of a mobile app: local storage, sync, conflict handling and a much larger test effort. It also needs to be decided at the beginning. Retrofitting offline support onto an online app is close to a rewrite.
Against that, set the cost of the alternative: paper in the field, double entry in the evening, data that arrives a day late, and an app that staff quietly stop using.
We design field applications with offline operation as part of Enterprise Software, and often alongside sensors in IoT and Control Systems projects. If your team works where the signal does not reach, contact us.
Summary
Build for the place where the work happens. An offline-first app stores what each role needs on the device, queues changes and synchronises in the background. Decide the conflict rule for each kind of data in advance, and prefer designs based on adding records over editing shared ones. Make sync repeatable and resumable, show users exactly what has been sent, protect data on devices, and test in real field conditions before rollout.
Frequently asked questions
What does offline-first mean?
An offline-first app stores data on the device and works fully without a connection. The user reads and writes locally, and the app synchronises with the server in the background whenever a connection is available. Being offline is treated as a normal state, not an error.
What happens when two people edit the same record offline?
A conflict occurs when both devices synchronise. The system must apply a rule decided in advance: the latest change wins, changes are merged field by field, one role's change takes priority, or the conflict is shown to a person to resolve. Which rule fits depends on the data, and it is a business decision more than a technical one.
Is an offline-first app more expensive to build?
Yes, noticeably. Local storage, a synchronisation mechanism, conflict handling, and testing under poor connectivity all add work, and the design must be right from the start because adding offline support later usually means rebuilding. It is worth the cost when the work genuinely happens where connections are unreliable.
Related service
Website Development
Fast, SEO-ready websites that turn visitors into customers.
