Half a billion spent, then abandoned

Lidl spent roughly €500 million over seven years building a SAP-based inventory system, then walked away from it in 2018 without going live. The number comes from German business reporting at the time of the cancellation; Lidl itself did not publish a figure. The project is now a standard citation in enterprise software circles for a category of failure that has nothing to do with bugs: the system worked, in the sense that it processed data correctly. What it could not do was match how Lidl thought about its own stock.

SAP's retail modules, like most ERP inventory systems, record goods at their purchase price — the price the retailer paid the supplier. Lidl had long managed its inventory at retail price, the price goods carry on the shelf. That is not an eccentric choice; many retailers use it, because it makes markdowns and margin calculations easier at scale. The problem was that migrating to SAP's approach would have required re-expressing every product, every warehouse movement and every historical comparison in a different unit of value. Seven years in, with a reported €500 million committed, the reconciliation still had not been resolved. In July 2018, Lidl cancelled the programme and announced it would return to its existing system, Wawi, which it had been running since long before the SAP project began.

The failure shares a structural cause with the Hershey go-live of 1999: not a technical breakdown, but a collision between the assumptions baked into a packaged system and the operational reality of the business that bought it. SAP's inventory logic is not a flaw; it is a design choice that works well for a majority of its retail customers. It simply was not compatible with Lidl's, and the question of whether that incompatibility was discoverable before seven years of development is the uncomfortable one. Enterprise software implementations typically surface deep-seated mismatches late, because the mismatch only becomes concrete when real data is loaded and real workflows are tested end to end.

What the record shows about cause

There is a word in German-language IT journalism for this class of problem: Anpassungsaufwand, the cost of making a standard system fit a non-standard process. The documented tension in the Lidl case was not that SAP refused to modify its software — SAP products carry a long history of customer-specific configuration — but that the modifications required to support retail-price inventory accounting were extensive enough to undermine the original argument for buying a standard package in the first place. A heavily modified SAP instance is harder to upgrade, harder to support and carries the same data-migration risk as a bespoke system when the vendor moves to a new architecture. Lidl was, in effect, being asked to build something custom inside a licensed shell.

The programme also ran across SAP's own product transition. The original implementation targeted an earlier SAP platform; by the time the project was mid-run, SAP had introduced S/4HANA as its strategic direction, adding another layer of architectural decisions to an already complicated programme. Whether the platform shift contributed materially to the cancellation is not documented in any published finding.

What is documented is the cost and the outcome. Roughly half a billion euros, seven years, no go-live. Lidl returned to Wawi and, from what subsequent reporting indicates, continued expanding internationally on the system it had been trying to replace. The episode did not stop Lidl growing. It did not bankrupt Lidl the way that the FoxMeyer collapse ended a company. But it sits in the record as evidence that vendor scale and product maturity are not sufficient conditions for a successful implementation. The mismatch between the system's data model and the customer's operating logic is always the real risk, and it is almost always underestimated at the point of signing.