Odoo as a headless ERP

A focused interface for the people doing the work, with the ledger running behind it. One pattern, four engagements, across more than a decade.

An ERP is designed for people who live in it. Most people who touch a business system do not: the operative moving a pallet, the bidder in an auction, the person taking a name at a trade stand. The usual answer is to train them all on the ERP, buy them all a licence, and accept the error rate that follows.

There is another way to arrange it. Keep the ERP as the system of record, and put a purpose-built interface in front of it for one specific job - integrated in real time, not synchronised overnight. The people doing that job get a screen that does exactly what they need and nothing else. The people who run the business keep the full system.

We have built it this way four times, and only one of those involved software of our own.

Four times, over more than a decade

DVLA timed auctions. Two open source products paired: Flexibase runs the auction, Odoo runs the accounting, integrated in real time. A bidder never sees Odoo, and what it is doing is not trivial - the winning bidder pays the hammer price plus buyer's premium and VAT, the VAT on the premium goes to HMRC, the hammer price and assignment fee go to the DVLA, deposits are taken above value thresholds and refunded to unsuccessful bidders, and documentation is released only once every lot in a purchase is paid. Flexibase is our own design and build.

CarShop. A dealer management system built on Odoo, with the website, mobile devices and third-party data feeds integrated into it, handling ex-fleet consignment stock that off-the-shelf dealer and accounting products could not accommodate. Delivered as an eighteen-month programme while the business grew from £100m to £300m. The Odoo Object Mapper - the library that lets an outside application drive Odoo without carrying its schema - came out of this programme. The plumbing for the pattern was built here.

JP Morris. A warehouse interface on phones and tablets over an Odoo Community Edition back end. Scan a lot number, choose the source location, confirm the move. Picking shows an order's demands with the lots that satisfy them, their locations, weights and expiry dates, and a package is assigned and validated in a few taps. The devices are deliberately low-cost, because the interface asks almost nothing of them.

A trade stand. Lead capture on lazysite, writing into Odoo CRM, with business card scanning - built for people using it occasionally, under pressure, in a noisy hall.

The boundary is the design, not a limitation

The JP Morris system will refuse to complete a pick when a lot has expired and needs manual confirmation. It says so, and sends the user to Odoo.

That refusal is the whole idea working correctly. The focused interface takes the work that is fast, repetitive and easy to get wrong. Judgement goes back to the ERP, where the full context, the audit trail and the people qualified to exercise it already are. A system that let a warehouse screen quietly override an expiry check would be worse, not better.

What it buys

Training for the focused task falls close to zero, because there is almost nothing to learn. Errors fall because the work is scanning rather than typing. Hardware is cheap, because the interface is undemanding. Licence costs do not scale with headcount, because occasional users are not using the ERP.

And there is still one ledger. Nothing has been copied into a second system that then has to be reconciled, and the ERP itself stays close to standard - the customisation lives in the surface, where it is cheap to change, rather than in the core, where it is expensive to carry.

What it is not

It is not a product we are selling. The front end has been different every time - a bespoke auction platform, a dealer management system, a mobile warehouse application, a web page - because the front end is the part that should be specific. What stays constant is the arrangement, not the software. Even the framework has been replaced: lazysite is the successor to Flexibase, and deliberately a much lighter thing.

It is not a replacement for the ERP, and it is not an argument that people should be kept away from it. It is an argument that the number of people who need to learn an ERP is smaller than the number who need to use the business.

All case studies → · Talk to us about your systems →