Foodtech · build vs rent economics

When does owning your tech stack actually beat renting a POS?

We didn't want to argue with a slogan, so we did the math — and built you the same calculator we use. Plug in your network, move the sliders, and see for yourself where the line really is. No sign-up, no pitch.

The question that started it

“If Toast and friends take that much in processing and modules — maybe custom software is just cheaper? And if so, where exactly does it pay off?”

00 — Where the question came from

The “8% margin leak” everyone repeats

You've seen the line in your feed: all-in-one restaurant platforms (Toast and the rest) can take up to 8% of revenue — processing, plus software, subscriptions, and a pile of “mandatory” modules. The framing is always that the box is a leak and custom is the rescue.

We decided not to trust the slogan and to actually run the numbers instead. Break the 8% apart, line it up against the cost of building your own, and find where the real line sits. Who does the math genuinely favor — and who is honestly better off renting?

The headline up front: “we're cheaper on commission” is the wrong argument. The real upside lives somewhere else — and that's exactly what decides whether building makes sense for you.

01 — Myth #1: “custom means cheaper fees”

You don't win on processing

This is the first thing that falls apart under the numbers. Take “your own software + Stripe” against a Toast-class boxed system. Public US market rates (2026):

Transaction typeYour software + StripeBoxed system (Toast-class)Who wins
Card present (dine-in, counter)2.70% + $0.052.49% + $0.15box
Online / delivery / keyed2.90% + $0.303.50% + $0.15custom

On in-person payments Stripe is actually a touch more expensive. Online, it flips and custom + Stripe comes out cheaper. The break-even sits around a 26% online share: below it, processing plays for the box; above it, for custom.

The honest part

A large chain on a negotiated enterprise contract can get processing from Toast below these rates. So against a well-negotiated contract, you won't “win on commission.” Don't build your decision on that.

Rate sources: public Stripe and Toast pricing; summaries from NerdWallet, Merchant Insiders, Merchant Cost Consulting, UpMenu (2026). Fixed per-transaction fees are left out of the model — those need an actual transaction count.

02 — Where the money actually is

The module tax that grows with every location

That “3.5% on top of processing” is recurring module fees: the POS plan, online ordering, loyalty, KDS, kiosk, a branded app, reporting. Each module is a separate line on the bill, charged per location, every month.

Typical market rates: online ordering ~$75/location/mo, loyalty ~$50–75, KDS ~$50–165, POS plan ~$69–110. A full stack lands at $350–600 per location per month. That's exactly what building your own replaces.

The key asymmetry

The box scales linearly: +1 location = +$400/mo, forever. Custom is one upfront CapEx plus near-fixed hosting for the whole network. The more locations you run, the wider the gap. “Cheaper to build at 15 locations than to retrofit at 80” isn't a slogan — it's the shape of the curve.

03 — The hidden multiplier

Your own data and ML — the thing the box won't sell you

This is the real argument. In a boxed system your data lives inside the vendor's stack and, as you grow, turns into a pile of isolated instances. Your own platform pulls it all into a single data warehouse — and that unlocks things the box simply can't do:

  • Demand forecasting (ML) — cuts waste and out-of-stock. With food cost ~30% of revenue, trimming losses by even 1.5% returns more money than all your software savings combined.
  • One customer identity — app + web + kiosk + dine-in = a single profile. Real loyalty personalization, not dashboard theater.
  • Upsell engine at the register — real-time recommendations on your own data.

It's the data and ML that drop the payback threshold: what breaks even at ~75 locations without them breaks even at 30–50 with them. And that's the one argument you can't buy from a vendor at any volume.

04 — Methodology

How payback is actually calculated

We compare two cost trajectories over time. Custom is a big one-time CapEx plus a small OpEx. The box is zero upfront, but a recurring module tax × number of locations. Where the curves cross = payback.

# Monthly savings from switching to custom
Savings = (N·M − OpEx) + N·ΔProc + ML

# Processing delta per location (sign depends on online share d)
ΔProc = Revenue · [(1−d)·(0.0249−0.0270) + d·(0.0350−0.0290)]

# Data multiplier (if counting ML)
ML = N · Revenue · 0.30 · %waste reduction

# Result
Payback(months) = CapEx / Savings    Net over 5 yrs = Savings·60 − CapEx

N — number of locations · M — module tax per location/mo · OpEx — hosting and platform support for the whole network/mo · d — online order share · CapEx — development cost. This is a directional model: it gives you direction and order of magnitude, not accounting precision. Not financial advice.

05 — What a build is made of

The cost of your own ecosystem and what it replaces

Each block is a one-time build cost and, at the same time, a set of modules you'd otherwise rent monthly and per location. Toggle blocks on and off and edit the numbers — the bottom recomputes the module tax M that feeds the payback calculator.

Build cost & replacement builder

Left — what you build (CapEx). Right — what it replaces in a boxed system ($/location/mo).

Build blockBuild, $Box equivalent$/location/mo
Website + customer app + custom loyalty
replaces online ordering + branded app + loyalty module
paid add-on
Kiosk (software)
separate license/device in the box
paid add-on
Menu board software (TV)
digital menu boards, often limited
paid add-on
POS register + admin panel
the box core + mandatory processing
paid add-on
Kitchen display system (KDS)
module/device, priced per location
paid add-on
Courier app · offline-first, real-time GPS
no proper box equivalent, usually
limited
Platform layer: data warehouse + integrations + ML
no box equivalent at all — enter your project price
not in box
One-time build (CapEx)
$285,000
Box module tax · per location/mo
$445

▲ Blocks with no real box equivalent: Courier app — these are where custom wins not on price, but on capability itself.

Open parameter

The platform layer (single data warehouse + integrations + ML models + shared admin + PM/QA) defaults to $0 — drop in your real number for the project. This is the layer that pushed earlier rough estimates of a full ecosystem toward ~$500K. The box has no equivalent for it, at all.

06 — Payback calculator

Where the line of profitability runs

CapEx and the module tax come from the builder above. Here you add your network parameters and see at how many locations the curves cross. Move the sliders to match your own setup.

Payback: custom vs rent

Where the lines cross = the moment you've paid off the build and gone net positive against the box.

60

The main lever. More locations, the harder the module-tax asymmetry works.

$80,000

Affects the processing delta and the size of the ML effect (via food-cost share).

30%

Above ~26%, processing starts playing for custom. Delivery-heavy networks win here.

$10,000

OpEx of your own platform across the whole network — infra, support, and ongoing development.

Verdict
Strong case to build
Payback under 1.5 years. This is the core profile.
Payback
1y 4mo
Module savings / mo
$16,700
Processing delta / mo
$1,584
Net over 5 yrs
$812,040
Box — cumulative costCustom — build + run

Left of the crossover you're paying off the build; right of it, the custom platform is pure margin against renting — and the gap widens with every new location.

07 — The substance, beyond price

Custom vs boxed POS: what matters besides cost

The green column is custom. We've marked in yellow where the box genuinely wins — that's what keeps the comparison honest.

CriterionYour own ecosystemBoxed system
Fit to your processesbuilt to specbend process to the product
Data ownership (single warehouse)fulldata in vendor's stack
Per-location module feesnone — one CapExgrows linearly with locations
Choice of payment processorfreeoften mandatory, can't switch
ML on your data (forecast, personalization)yesno / surface-level
Courier app (offline, real-time GPS)nativeusually none, via third parties
One ecosystem across all channelsa single layera set of different-vendor integrations
Speed of change across the networkfrom one panelthrough scattered admins
Lock-in and contractsnone2–3 years + exit penalties
Upfront capitalrequirednot required box wins
Support, updates, PCI scopeyour responsibilityincluded box wins

08 — The non-financial case

Why this was never really about saving money

  • A data asset. Owned customer data is what stays with the network for good and works for every initiative after it. The app is just a channel; the value is the data underneath.
  • Margin leverage through ML. Demand forecasting and purchasing optimization hit cost of goods — your single biggest line. That's not software savings, it's margin growth.
  • Speed at scale. One promo rolls out to every location at once, instead of being assembled over weeks across dozens of isolated admins.
  • Vendor independence. Freedom to switch processors — and any component — without penalties or re-integration.
  • Closing the box's gaps. Offline-first operation and native real-time courier tracking — the things boxed systems either lack or bolt onto third parties.
  • Operations on your terms. Loyalty logic, upsell, and reporting built around how the network actually runs, not the other way around.

09 — Who this is for

Where the math is on your side

This fits you if

  • 50–150 locations (payback starts around ~40–50)
  • A growing network — the asymmetry works for you over time
  • A real online share (>30%) — then processing turns positive too
  • Already a patchwork multi-vendor stack — your real tax is higher
  • An ambition to use your data — the ML upside is real

It probably doesn't (and that's fair)

  • Fewer than ~30 locations
  • Dine-in only, little online
  • No capital for upfront CapEx
  • Simple operation, standard processes
  • You value “all-inclusive, someone supports it”

The honest version

“You won't beat a boxed system on processing — a good vendor contract will undercut us. But no one will sell you ownership of the data layer where ML trims your waste and personalizes at the customer level. The money is there, and it grows with every new location instead of leaking out as a module tax.”

Ran your numbers and the math leans toward building?

That's usually where the real conversation starts. We're happy to pressure-test your assumptions with you — honestly, even if the answer turns out to be 'rent the box for now.'