When an EPOS goes down on a Friday night, the first thing the operator does - once the shouting has died down - is open a spreadsheet and try to work out what the outage cost. Almost without exception, the number they arrive at is the takings they would have rung through the till during the dead time. They subtract that from the day’s target, attach it to an angry email to their support provider, and move on.

That number is real. It is also the smallest part of what just happened.

Over the past two years we have helped a number of London hospitality groups reconstruct what their downtime events actually cost them once everything is on the page. The headline lost-sales figure is rarely more than a third of the total. The rest is hidden in fallback labour, refunds, churn, payroll knock-on and back-office time.

All numbers below are ballparks based on what we see in mid-market casual dining and pub groups in London. The point is the structure, not the decimal places.

The Five Cost Components of an EPOS Outage

1. Direct lost sales

This is the obvious one. If the till is down and you cannot take orders or payment, some proportion of the revenue you would have taken in that window walks out the door.

Worked example: a mid-size casual dining site doing roughly 8,000 pounds a day across lunch and dinner. The EPOS goes down for two hours over the Friday evening peak - say, 19:30 to 21:30. Peak hours are not evenly distributed: those two hours probably represent something close to 25 to 30 per cent of the day’s revenue, not 2/14ths of it. Call it 2,000 to 2,400 pounds of attempted sales in the dead window.

You will not lose all of it. Some guests will wait. Some will pay cash. Some will accept a paper bill. But you will lose a chunk - typically 60 to 75 per cent in a peak service that depends on table turn. Direct lost sales: roughly 1,300 to 1,800 pounds. Call it 1,500 pounds for the example.

That is the number that ends up in the angry email. Now we add the rest.

2. Manual fallback costs

The shift does not stop when the till stops. Front of house starts writing tickets by hand. The kitchen tries to keep up without a KDS. Servers run cards on a backup terminal or take card numbers down on paper. Some guests get the wrong bill. Some get their meal comped because the team cannot be bothered to argue. Some items are forgotten entirely.

When we sit down with finance teams the morning after, the manual-fallback cost typically lands at 25 to 40 per cent of the direct lost-sales figure. It comes from:

  • Refunds and comps issued in the moment to keep the peace
  • Items that were served but never made it onto a paper ticket and so were never charged
  • Reconciliation errors discovered the next day when the manual tickets get keyed in
  • Card payments taken on a backup device that then have to be matched to covers by hand
  • Overtime for the manager who stays back to sort the float and the paperwork

For our example site, add another 400 to 600 pounds. Call it 500.

3. Guest churn

This is the hardest to measure and almost always the biggest long-term cost.

Some guests walked out during the outage. They will not come back this month, and a portion of them will not come back at all. Others stayed but had a bad night - slow service, wrong bill, no card machine, a manager apologising at every table. A few will leave a one-star review that sits at the top of your Google profile for six weeks. A handful of regulars, the ones who come every Friday, will quietly try the place down the road next week and discover they like it.

We model churn conservatively at 5 to 10 per cent of the affected covers being permanently lost, valued at their annual contribution rather than a single visit. For a site with strong repeat custom, even the bottom of that range gets uncomfortable quickly. On our example site, two hours of peak might touch 80 to 120 covers. Lose six of them as regulars worth 400 pounds a year each and you are looking at 2,400 pounds of annualised revenue gone from a single outage.

We will use 1,500 pounds as a deliberately cautious figure for the example. In reality this is the line item that should keep finance directors awake.

4. Staff morale and productivity

The team that worked the outage is tired and irritated the next day. The chef is still annoyed about the tickets that came through in the wrong order. The GM spent the whole service apologising instead of running the floor. Sales targets were missed, which means commission and bonus thresholds were missed, which means the conversation in the staff room next week is about the till that did not work.

Productivity in the days after a bad service measurably drops. In a tight labour market - which is most of London - a single bad night can cost you a key team member who decides they have had enough. Replacing a senior server is a 1,500 to 3,000 pound exercise.

Conservatively, add 200 to 400 pounds per outage as a steady-state cost. Call it 300.

5. Back-office cost

Finally, the bit nobody sees. The next morning, somebody has to reconcile the manual tickets against the card terminal log against the cash float against the stock movement against the EPOS audit trail (such as it exists for the dead period). That somebody is usually your most expensive back-office person, and it usually takes them most of a day.

Add management time for the post-mortem, the call with the EPOS vendor, the email to head office and the conversation with the auditor at year end. For a single site, single outage, we typically see 200 to 400 pounds of back-office cost. Call it 300.

The total

For one two-hour peak outage at one site:

  • Direct lost sales: 1,500 pounds
  • Manual fallback: 500 pounds
  • Guest churn (cautious): 1,500 pounds
  • Staff and productivity: 300 pounds
  • Back-office: 300 pounds
  • Total: roughly 4,100 pounds

The lost-sales figure in the angry email was 1,500. The actual cost was nearly three times that. And we were deliberately conservative on churn.

Scaling Up: A Worked Example for a 10-Site Group

Take a 10-site casual dining group. Assume each site suffers one two-hour peak outage per month - which, in our experience, is roughly what an estate on a “next business day” support contract running ageing hardware looks like over a year.

10 sites x 12 outages x 4,100 pounds = 492,000 pounds per year.

That is half a million pounds of avoidable cost, sitting inside an operating model that almost nobody on the finance team has ever quantified, because the only number that ever surfaces is the lost-sales line.

Even if you halve our estimates across the board, you are still looking at a quarter of a million pounds a year. That is a meaningful number against the cost of a proper EPOS support contract, a refresh of end-of-life hardware, or a managed network refresh that stops the till dropping off the LAN in the first place.

Why “Cheap” EPOS Contracts Are a False Economy

The cheapest EPOS support contract on the market in 2026 is roughly 40 to 60 pounds per terminal per month, with a “next business day” SLA and no on-site response. The difference between that and a proper hospitality-grade contract - 24/7 cover, on-site response inside four hours, proactive monitoring, integrated network and payment support - is usually somewhere between 30 and 60 pounds per terminal per month.

For a 10-site group running four terminals per site, that is at most 28,800 pounds a year of additional contract cost. Set against half a million pounds of avoidable downtime, the maths is not subtle.

The problem is that the cheap contract looks cheap on the line item finance reviews every quarter, and the half-million-pound cost is invisible because it is scattered across a dozen lines and a dozen sites. This is the central reason groups tolerate downtime for years before doing anything about it.

What Good Looks Like

A hospitality-grade support setup, in our view, has the following characteristics:

  • 24/7 cover, with named engineers who know your estate, not a generic out-of-hours desk
  • Target resolution time of one hour for a P1 (till down at peak), four hours for a P2
  • On-site response inside four hours for any London site, faster for flagship venues
  • Proactive monitoring of terminals, network, payments and integrations - so you know about a problem before the GM does
  • An integrated view of EPOS, network and security - because most “EPOS outages” are actually network or certificate problems wearing an EPOS hat
  • A written runbook for fallback that the team has actually rehearsed

The last point is the cheapest and most often skipped. Even with the best support contract in the world, you will have outages. The difference between a 4,100-pound outage and a 1,500-pound outage is whether the floor team knows what to do in the first ten minutes.

How CloudMatters Hospitality Customers Benchmark

Across our hospitality customer base, the groups that have moved from a cheap reactive contract to a proper managed service typically see EPOS-related downtime drop by 70 to 85 per cent in the first year, and the average outage duration drop from over two hours to under thirty minutes. The avoided cost - calculated using exactly the framework above - almost always pays for the contract uplift inside the first quarter.

If you want to run the same exercise on your own estate, the starting point is a downtime audit: every outage in the last twelve months, duration, peak or off-peak, root cause and the five cost components above. Most operators have never seen their own number on a single page. It tends to focus the mind.

If you would like help putting that page together, our team works with London hospitality operators every day - see IT support for hospitality for how we approach it, or get in touch and we will run the audit with you.