How to Measure the ROI of an Employee Meal Programme
A measurement framework finance will accept: the four metrics that matter, the baselines to capture before launch, and the vendor numbers to ignore.

Most meal programme business cases are built on a productivity percentage that somebody invented. You have probably seen the shape of it: assume a 15% afternoon productivity uplift, multiply by an average daily salary, multiply by headcount, and produce a number with six figures and no evidential basis whatsoever.
Finance directors have seen it too, which is why those business cases tend to die quietly.
There is a better way to do this, and it is not harder. It requires deciding what you are actually buying, capturing a baseline before launch, and measuring four things you can genuinely observe. Here is the framework.
Start by naming the outcome
A meal programme can be bought for at least four different reasons, and they are measured differently. Pick one primary outcome, or the measurement gets muddled and the review meeting turns into an argument about which number counts.
Office attendance. The commonest driver in London since hybrid settled. You are buying a reason to make the commute.
Retention, particularly of junior staff. You are buying a daily, visible signal of being looked after, in a population that leaves for reasons that are rarely purely financial.
Admin and expense reduction. You are buying the removal of a process: reimbursement claims, receipt chasing, month-end reconciliation of food spend.
Replacing or supplementing a canteen. You are buying variable cost in place of fixed cost, and menu variety in place of a rotating hot counter.
Secondary benefits are real and worth mentioning. They are not what the programme should be measured on.
Capture the baseline before you launch
This is the step almost everyone skips, and skipping it makes everything afterwards unfalsifiable.
Four weeks before go-live, capture:
- Attendance, by day of week, from badge swipes or desk bookings. Four weeks minimum, so day-of-week effects are visible.
- Current food spend, including reimbursed lunches on personal cards, ad-hoc catering, the fruit order, and any canteen subsidy.
- Expense claim volume and processing effort for food specifically. Number of claims per month, and an honest estimate of the finance time spent on them.
- Retention and engagement, at whatever granularity you already have: twelve-month attrition by team, and the relevant engagement survey items.
Without these, six months later you will be comparing your programme against a memory.
The four metrics that matter
1. Adoption
The leading indicator, and the earliest warning of trouble. Track weekly, by team and by site.
Define it before you count it. Uptake is orders placed as a share of eligible person-days (an eligible employee on an eligible day); weekly unique participation is the share of eligible employees who order at least once in a week. They answer different questions, so report both and say which one a figure uses. There is no universal healthy figure: set your own target from the pilot, and read the trend rather than a benchmark.
If adoption is low after the first month, the cause is usually one of three things, all fixable and all visible in the data: awareness (people forgot it exists), restaurant fit (the team's favourites are missing), or friction (login problems). Work out which before adjusting the budget. For restaurant fit, a pilot team can compare real menus, for example atis and The Salad Project, in a short preference survey before the subsidy changes.
Adoption also drives the cost model, which is the part finance cares about. A £12 daily budget for 100 people over 200 working days is £240,000 at full uptake, but around £160,000 at 67% uptake. Budget the first number, forecast the second.
Recording a lunch that went wrong
Ordit's own delivery fleet handles each order from the restaurant to the desk, so a pilot team has one operator to ask when a delivery misses the mark. Group meals arrive accurate and labelled with each person's name, so the pilot team can log handover issues apart from menu feedback. Ordit's on-time delivery guarantee applies to all deliveries. Employees can order meals delivered in around 30 minutes. See programme reporting or book a demo.
2. Attendance on programme days
The measurement is simple and most organisations already hold both series: meal orders by date, and badge swipes or desk bookings by date. Overlay them, control for day of week, and compare programme days against non-programme days and against the pre-launch baseline.
Two warnings. First, do not attribute the whole delta to food. Team norms, meeting schedules, weather and managers all move attendance. Second, be sceptical of any supplier who quotes you an attendance uplift percentage without access to your badge data. That includes benchmarks in articles like this one.
The published research is useful for shaping expectations rather than making promises. ezCater's 2025 Lunch Report found 75% of hybrid employees say they would work on site more often if their employer provided lunch, and its decision-maker survey found 78% agree that providing food encourages in-person work. DoorDash's 2026 report found Thursday workplace orders running 20% above Monday. Both are US samples. They tell you where to look; your badge data tells you what happened.
3. Admin cost removed
The most defensible line in the whole business case, because it is arithmetic rather than inference.
Count the food-related expense claims the programme eliminates, and cost them. The GBTA and HRS benchmark study puts a single expense report at $58 and 20 minutes of staff time, with the 19% that contain errors costing a further $52 and 18 minutes to correct. Even at a deliberately conservative £15 a claim, a programme that removes 100 claims a month removes £18,000 a year of invisible finance and manager time.
Add the reconciliation effect: one monthly invoice, coded by team, site and expense code, replacing a month of individual claims to audit. That is the line that makes finance an ally rather than an obstacle, and it is covered in detail in the hidden cost of meal expenses, which looks at expense administration. For reconciliation and coding, see finance operations and integrations.
4. Retention and sentiment
The slowest metric and the one with the largest financial weight, which is an awkward combination.
Track twelve-month attrition in participating teams against the pre-launch baseline and against non-participating teams if you piloted. Expect to need two to three quarters before the signal is readable, and expect the programme to be one contributing factor among many.
For sentiment, a two-question pulse at week four and at month six beats a suggestion box: does the programme make the office a better place to work, and would you notice if it stopped. The second question is more informative than the first.

A worked review, at six months
All figures here are illustrative. A 150-person London firm, £11 daily budget, in-office days only (Tuesday to Thursday), launched in January.
Cost. 150 × £11 × 3 days × 26 weeks = £128,700 at full uptake. Actual spend at 68% uptake: £87,500, about 100 meals a day. The example assumes they arrive as scheduled group orders from four restaurants a day, so delivery is 78 days × 4 × £15 + VAT = £5,616 with VAT. The 5% + VAT service fee on food and delivery adds about £5,590, so the all-in figure is about £98,700. Class 1A National Insurance at 15% is charged on the cost of providing the benefit, meals, delivery and service fee together (GOV.UK; HMRC EIM21102): 15% of about £98,700 adds around £14,800, because a recurring meal budget is a taxable benefit and salary sacrifice does not help.
Attendance. Core-day attendance up from 61% to 73% against the pre-launch baseline. Monday and Friday up 4 points, which would be a finding worth noting and not one the programme was designed to produce.
Admin, over the same six months. 380 food expense claims a month before launch, 0 after. At £15 a claim, that is 380 × £15 × 6 = £34,200 of manual processing removed in the six months, the same period as the cost above. Staff time released is not cash saved unless the hours are actually redeployed or hiring is avoided, so show the two lines separately.
Retention. Twelve-month attrition in the participating population down 2 points against baseline. Too early to attribute confidently, and worth saying so out loud in the review.
The honest conclusion from that review is not "the programme returned £X". It is: over six months the programme cost about £113,500 all-in, removed about £34,000 of processing cost, moved core-day attendance 12 points, and the retention signal is promising but unproven. That is a number set a CFO can act on, and it survives the second meeting. Choose the comparison period before launch and use it for every line: six months of cost against a full year of savings, or the reverse, would overstate or understate the result.
The numbers to leave out
Invented productivity multipliers. No credible study assigns a percentage of afternoon output to lunch habits. The research on lunch breaks is real, and narrower: breaks involving relaxation, autonomy and company predict lower afternoon exhaustion and higher engagement. Quote that if you want the mechanism, and do not convert it into pounds.
Supplier attendance benchmarks presented as forecasts. Useful as a hypothesis, worthless as a commitment.
Employer-brand claims you cannot evidence. "It will help us hire" is plausible, and ezCater's survey finding that 90% of decision makers now mention food perks during hiring supports the direction. It is not a line item.
Food waste savings against a canteen you have not costed. If you are replacing a canteen, cost it properly, including fixed overhead at actual usage rather than at design capacity. The canteen cost piece sets out the comparison.
What to instrument on day one
Everything above depends on data you either have or do not. Three things to insist on when choosing a platform:
- Order data by date, employee, team and site, exportable, so attendance overlays are possible without a support ticket.
- Spend against budget in real time, by team and site, so the forecast can be corrected mid-quarter rather than at year end.
- Expense codes captured at checkout and pushed to your finance system, so the admin saving is real rather than relocated.
Ordit's admin portal reports by order, item, restaurant, employee, team, department and expense code, and can push coded transaction data to SAP Concur, Coupa or Workday via SFTP. The point is not the feature list; it is that the measurement above becomes an export rather than a project.
For the programme design that sits underneath the measurement, see how to set up an employee meal programme in 4 weeks, and for the budget mechanics themselves, meal budgets.
Ready to feed your team properly?
Join hundreds of London offices already using Ordit for individual meals, group orders, and catering.



