Build or buy a POS accounting integration? The real 3-year numbers

Should you buy a standard connector to get your till sales into your books, or build your own POS accounting integration? My short answer: one or two stores, buy. A chain where every location is its own company with its own administration, build. For a 19-store bakery chain going from Lightspeed to Twinfield, my estimate came to about €4,950 to build versus about €12,450 for a per-store connector over three years.

That's an estimate I made in July, before I built anything. Not an invoice, not a promise. But the logic behind it holds for any chain, so here it is.

The 3-year math behind Lightspeed to Twinfield

The client is Vlaamsch Broodhuys, an artisan bakery chain with 19 stores. Every store is its own B.V. with its own Twinfield administration. In the summer of 2026 they moved every store, one by one, to the Lightspeed K-Series till.

Off-the-shelf connectors are priced per location: a setup fee per store and a subscription per store per month. Each line looks small. Then you multiply by 19 stores and by 36 months.

A custom build works the other way around. Most of the cost sits at the start, and adding store number 20 costs almost nothing. That's how you end up at roughly 40% of the connector cost over three years.

The spreadsheet missed two things, though. The connectors on the table didn't do what they needed. And the build did more than booking.

When building is not the right approach: buy the connector

I'll be honest, for a lot of businesses a connector is the right call. Aqqo, Omniboost and Combidesk all exist for a reason. Omniboost is even listed on Lightspeed's own Twinfield integration page.

Buy when:

• You have one or a few locations, booking into one administration. The per-store price never gets the chance to add up.
• Your exact till version is supported. Check this first. "Lightspeed" is not one product.
• A standard daily booking per ledger account is enough, and nobody needs anything else from that data.
• You don't want to own anything. A subscription comes with a vendor and a helpdesk. A custom build means someone has to own it, even if that's mostly watching an alert.

If you're looking for a Lightspeed Twinfield koppeling for a single shop, stop reading and buy one. Seriously.

When building your own POS accounting integration wins

At Vlaamsch Broodhuys the connector route failed on fit before it failed on price.

• Multiple locations, one administration each. The cheaper option couldn't split bookings per location. Nineteen stores means nineteen administrations and one booking per store per day.
• Your till isn't supported. Combidesk only supports Lightspeed C-Series. These stores run K-Series.
• Card fees. The controller posted Lightspeed Payments transaction costs by hand from Excel. Now they're booked per store per day, back-booked from 7 September.
• Tips. Booked on their own account, not mixed into revenue.
• Dashboards on the same data. Once the sales data flows reliably, the next question is always "can the store managers see it?" A store dashboard with a personal 07:00 email now runs on the same stream for about 50 users.

That last one is the real argument. A connector moves numbers from A to B. Your own integration gives you the data, and you decide what else plugs into it.

What makes an integration trustworthy

A kassakoppeling that books money into your books has one job: be right, every day, without anyone checking. Speed is easy. Trust is the work. This is what I built in, and what I'd ask any vendor about too:

1. It refuses unbalanced days. A day that doesn't balance, or isn't closed yet, doesn't get booked. Revenue groups without a mapping get refused too.
2. It never books twice. Before booking, it checks a register that is verified against what's actually in Twinfield. Run it ten times, you get one booking.
3. It re-checks the last 10 days on every run. If a day was refused or missed, it gets booked automatically once it's correct.
4. Bookings are provisional. The controller finalises them. The integration does the typing, finance stays in charge.
5. Unknown payment methods go to a fallback account with an alert, instead of breaking the booking.
6. A watchdog. A separate dead man's switch raises the alarm if the job ever stops running. Failures alert me directly.

Lightspeed closes the business day at 05:30. The job runs at 06:00, so every store's day is in its own administration before the bakery opens. Around 4,000 receipts a day, across 19 stores.

Certification: someone else checks your work

Here's something people don't expect: you can't just build a Twinfield integration and switch it on for production. Wolters Kluwer audits integrations that write into Twinfield.

For this one, the process started on 22 July. In the review I had to defend sending my own VAT amounts instead of letting Twinfield balance the VAT automatically. I wanted the books to show exactly what the till said. On 30 July 2026 Wolters Kluwer confirmed the certification, valid for two years.

A small warning if you go this route: Twinfield has no staging environment. Testing runs on live data, and test rows get deleted afterwards. Plan for that.

First real booking on 10 August. Fully automatic for every store since 11 August.

How to decide for your own chain

1. Count your locations and administrations. One administration for everything is connector territory. One per location starts to tip the other way.
2. Check your exact till version against each connector's support list.
3. Price the connector over 36 months: setup per location plus the monthly fee, times your number of locations.
4. List what you still do by hand after the booking. Card fees, tips, store reports. That's the work a connector leaves behind.
5. Ask who owns it when it breaks. If the answer is nobody, buy.

If the numbers point to building, the same pattern works for other tills with an API and other accounting packages. You can see more of what I've built on my work page.

Running several locations and stuck between a connector and a build? Let's do the math together in 30 minutes.

Is it cheaper to build a POS accounting integration than to buy a connector?

It depends on how many locations you have. Connectors are priced per location, so the cost grows with every store. For 19 stores my estimate over three years was about €4,950 to build versus about €12,450 for a connector. For one or two stores, a connector is usually cheaper.

Can Lightspeed K-Series be connected to Twinfield?

Yes. Lightspeed's API exposes daily financials per location, and Twinfield accepts cash-book entries per administration. Check support carefully, because not every connector supports K-Series. Combidesk, for example, only supports C-Series.

Can every location book into its own Twinfield administration?

Yes. In this case every one of the 19 stores books into its own administration, one booking per store per day, with revenue split per ledger account and VAT rate and each payment method on its own line.

Does a custom Twinfield integration need certification?

Integrations that write into Twinfield are audited by Wolters Kluwer before they go live on production. This Lightspeed to Twinfield integration was certified in July 2026, valid for two years.

What happens if a day doesn't balance?

It doesn't get booked. The integration refuses unbalanced or unclosed days, sends an alert, and books the day automatically on a later run once it's correct, without ever booking it twice.