Oracle Is Retiring GloriaFood. How does that affect the UK Restaurant market?

The event

Oracle published a document in March 2026 titled Oracle GloriaFood EOL Frequently Asked Questions, version 1.0, hosted on oracle.com. The first line does the work: Oracle will end-of-life all GloriaFood products, services and the partner programme on 30 April 2027.

Three product lines end that day — Online Ordering across both the free and paid tiers, the Point of Sale service and its hardware, and Oracle Payment. New direct agreements and new partner agreements stopped the moment the FAQ went out. Existing services auto-renew through the wind-down, but renewals close a month before the end date, and each account's service actually stops at the expiry of its final term. Oracle's own worked example: a 30-day term renewed on 3 March 2027 ends on 3 April, not 30 April. There's a 30-day data-retrieval window after that, and no migration tool.

It didn't feel official because Oracle ran no PR campaign. The notice went to the accounts that use the system, not the trade press. That's why it reads as rumour secondhand and reads as settled once you have the PDF.

Why Oracle walked

Oracle hasn't published a reason, so anything definitive here is inference — but the filings make the inference easy. Oracle's FY2026 10-K, filed 22 June 2026, describes a company pouring everything into AI compute: capital expenditure of $55.7bn, up 162% year on year; free cash flow of negative $23.7bn; headcount down around 13% to 141,000; a $1.84bn restructuring charge; roughly $70bn of capex guided for FY2027. Sitting behind that is a compute commitment to OpenAI reported at around $300bn.

Against that backdrop, a free ordering widget for independent restaurants is the definition of a line you cut. GloriaFood was acquired in June 2021 to bolt global online ordering onto Simphony, Oracle's enterprise POS. The attach strategy never needed the standalone free product to survive, and the independents who used it were never Oracle's customer. Oracle now markets Food & Beverage almost entirely at chains, stadiums, hotels and enterprise. The one genuinely SMB product in the portfolio is the one being switched off. That's consistent with a long Oracle pattern: buy a platform, keep the enterprise tier, discontinue the small-business end, let those customers scatter.

What this actually is, commercially

A forced-switch event with a hard date across an entire base at once.

Merchants don't normally move. POS and ordering are sticky by design, and most switching in UK hospitality is involuntary — venues closing rather than choosing to leave. The GloriaFood EOL breaks that. It puts a whole population in-market on a deadline, which is the rarest and most valuable condition a vendor selling into this space can encounter. Timing is set by Oracle's renewal mechanics, not the April headline — accounts on 30-day terms start losing service in staggered waves from early 2027, and renewals close a month before the end. In-market urgency is a 2026 story, not a 2027 one.

The catch is that the base splits into two populations that need completely different treatment, and most vendors will aim at the wrong one.

Population one: the independents

The larger group by headcount — independents who dropped a free ordering widget onto a website, and the web agencies who built products on the white-label programme.

Nobody can size it. Oracle publishes no user numbers. The "123,000 restaurants" figure circulating on migration blogs is unsourced and traces to nothing; GloriaFood's own peak marketing claim was "over 20,000 restaurants in 100 countries." Install crawlers, which only see live widgets, disagree by a factor of four — Wappalyzer around 2,300 sites, WebTechSurvey 4,733, Aguko 9,376. WebTechSurvey's country split puts roughly 296 live UK sites, 259 on .co.uk, fifth behind the US, Netherlands, Germany and Canada.

So the entire UK sizing question rests on one third-party crawl. Plan against hundreds you can name, not thousands you can't.

These merchants aren't in market for a till. They lost a free widget; a 36-month POS contract answers a question they didn't ask. The realistic destinations are direct-ordering platforms and low-cost all-in-ones, and the pitch that lands is commission. On a £22.50 delivered order, Deliveroo and Uber Eats take about 30% — roughly 36% after VAT — versus around 2% on an owned website. SmartPubTools puts the gap at £32,760 a year for a venue doing 100 orders a week. That's the number that moves an independent, not a feature list.

Population two: the Simphony and OPERA sites

Smaller, higher value, different rhythm. These are established Oracle POS and hotel customers who lose one integration, not their system. The GloriaFood online-ordering link stops at EOL; Simphony and OPERA carry on. Oracle points these customers at its own validated integration partners — Deliverect, QikServe, UrbanPiper, Tillster and others, several with real UK operations — rather than leaving them to search.

For a vendor already on that partner list, this is qualified demand arriving without a fight. For one that isn't, it's a channel that just narrowed. Worth noting: Flipdish, one of the larger UK/Ireland ordering players, isn't on Oracle's list. If you sell into the Simphony orbit and you're not validated, that's the gap to close.

The reseller layer nobody's fighting for

The partner programme ends on the same date, which kills the white-label engine UK agencies built ordering products on. Those agencies now hold a book of restaurant clients and no backend.

That's the highest-leverage target in the whole event and it's the least contested. One agency conversation can carry twenty or more venues. Most vendors are chasing restaurants one at a time; a couple of overseas players (Restolabs, Fleksa) are recruiting the displaced partners directly. The agencies made their money on the build and the hosting, not the software — so the winning offer is wholesale margin, multi-site admin and no per-order commission, which are exactly the gaps in the old GloriaFood deal.

The part that's a data problem

Every vendor above wants the same thing: the list of UK venues running GloriaFood, with a reachable decision-maker. Nobody has it.

Not Oracle, which has published nothing. Not the trade press — The Caterer, Big Hospitality, Propel and MCA hadn't covered it at the time of writing; the noise is all migration vendors selling replacements, which is where the inflated counts come from. And not the crawlers, which disagree by a factor of four and only see what a browser downloads.

That last point is the crux. Technographic detection systematically misses this segment: dormant free accounts with no live widget, agency-managed sites where the widget sits on a subdomain or loads through a tag manager, venues that only ever took orders through the GloriaFood-hosted page or Facebook, and the large cohort of independents with no website at all. The detectable UK count is a floor of a few hundred. The true base sits above it and well below the marketing numbers, and no public dataset closes that gap.

Closing it is a verification problem, not a scraping one. Peach Data catalogues UK hospitality merchants at record level with the technology signals attached and confirmed against real venues — which is the difference between a targeted displacement campaign and spraying the 296 sites a crawler happened to catch. If the honest answer to "which UK venues run GloriaFood" is currently a screenshot from one tracker, that's the thing to fix before committing pipeline to the deadline.

The three questions worth answering before you spend on this

How many UK sites can you actually identify as running GloriaFood today, by name and address — not as a crawler total, but as reachable venues?

Which of those sit on Simphony or OPERA, and are therefore a partner-referral conversation rather than a cold one?

How many are agency-managed rather than direct — because if they are, you're pitching the wrong party, and the agency is worth twenty of them.

The deadline is real and dated. The base is real and unlisted. The vendors who win it are the ones who solve identification first and sell second.