What software a small restaurant needs: the 2026 data, not the vendor's pitch

How much software does a 40-seat restaurant actually need?

Four pieces, and the fourth one isn't purchased: a POS with exportable reports, a Google Business Profile managed as a sales channel, one unifier for Rappi, Uber Eats and DiDi orders, and a living costing sheet.

The March 2026 case that brought me to this table was paying 386 USD a month across four subscriptions on 11,400 USD of sales, meaning 3.4 % of revenue went to licenses. With the average independent operator running a net margin between 3 and 5 % according to the National Restaurant Association, that 3.4 % isn't a technology expense: it's nearly the entire year's profit turned into dashboards nobody opens. The concrete decision: cancel today whatever hasn't produced a report someone actually read in the last thirty days, and move that money to the map listing. Seventy-one percent of restaurant searches happen within five miles of the venue, and that single figure rearranges the whole software budget.

The Google listing isn't marketing, it's 71 % of demand

That office-district restaurant had 19 reviews and the most recent one was fourteen months old; it competed against listings holding 300 and 400 fresh reviews inside the same walking radius. It wasn't losing on food, it was losing on map position. No CRM subscription with AI agents fixes that, because the local algorithm weighs review frequency, owner replies, recent photos and updated hours, four things costing zero dollars and forty minutes a week. The concrete decision: before signing any new license, get the listing to twenty new reviews per quarter. That moves cash; a fifth system doesn't. One dashboard for all three apps pays for itself, and here I do recommend spending. Running three separate tablets — Rappi, Uber Eats, DiDi — a small venue loses orders to menu desynchronization and slow acceptance times, and every rejection sinks its ranking inside the app. Evidence on organized digital channels points the same way: according to Sunday, a complete digital offer covering menu, ordering and payment lifts the check between 20 and 30 %.

The delivery unifier: where a license does pay off

On 11,400 USD a month, a conservative 12 % lift on the delivery channel when it represents a third of sales works out to roughly 456 USD extra monthly, against the 60 to 120 USD a unifier costs across Latin America. The concrete decision: measure last week's rejection rate before renewing any other module. Kiosk numbers are real and still don't apply to your 42-seat room. GRUBBRR reports gains between 10 and 30 % in order value, McDonald's cites close to 30 %, Future Ordering documents a case at 35 %, and QSR Magazine places the prudent range at 8 to 15 %, with Yum around 10 %. That spread — 8 to 35 % — is the clue: the effect depends on transaction volume and queue length, never on the device. A 2,400 USD kiosk spread over 380 monthly tickets takes years to return; over 12,000 tickets it returns within a quarter.

Kiosks, chatbots and the mistake of buying someone else's check

For chatbot-guided ordering, Zellyfi measures 12 to 18 % more check, a figure far more transferable to a small venue running WhatsApp. The concrete decision: divide equipment cost by your real thirty-day ticket count. The fourth piece costs nothing and decides whether the other three matter. I watched operations close while paying 500 USD monthly in licenses with prime cost at 71 %, and not one of those licenses touched the number killing them. Under the Masterestaurant method, food cost per dish carries a ceiling of 32 % — a maximum, not a target — and payroll, rent and utilities never load onto the plate: they live in the break-even calculation. A living sheet means standardized recipes with purchase prices refreshed monthly, not a file from two years ago. Diego F. Parra insists on an order almost nobody respects: the operation defines what needs measuring first, and only then do you hire the software that measures it.

The costing sheet is software, even without a logo

The concrete decision: recost your ten best-selling dishes this week. The three scenarios change the answer completely, so place yourself before spending anything. SMALL venue, up to 60 seats and under 20,000 USD in monthly sales: the four pieces and nothing else, with a hard ceiling of 150 USD monthly in licenses, roughly 0.8 % of sales. MEDIUM venue, from 20,000 to 60,000 USD: add real inventory control and shift scheduling, because payroll now weighs enough — U.S. restaurant base hourly wages rose 4 % to 14.20 USD according to 7shifts — that half an hour badly assigned costs money. GROUP of three or more locations: multi-unit consolidation and BI, and only there a CRM. The concrete decision: if your sales don't reach 20,000 USD, cancel today every module belonging to the next scenario. Honesty about sources, because those wide ranges have a reason.

Where these benchmarks come from and what they do NOT prove?

Kiosk figures come from kiosk vendors — GRUBBRR, Future Ordering, Restroworks — and from trade press like QSR Magazine, and a vendor publishes its best cases, not its median;

that's why the range runs from 8 to 35 % and why anchoring to the low end serves you better. The digital-check figure belongs to Sunday, also an interested party. Wage data comes from 7shifts on U.S. numbers, which don't translate one-to-one to Latin America or Spain. The 3 to 5 % margin is National Restaurant Association work measuring independent operators. None of these figures is a Masterestaurant study: they're public data read with a consultant's judgment. The concrete decision: when a vendor quotes you 30 %, ask for the ticket volume behind that case. Suppose you cancel all four subscriptions and keep only the POS. You save 386 USD monthly, meaning 2,316 USD across the semester.

What would happen if you inverted the order for six months?

You put forty minutes a week into the map listing and reach 60 new reviews;

with 71 % of local demand resolving on the map, climbing from the third screen to the first isn't marginal, it's the difference between existing and not existing for someone searching twelve blocks away. With those 2,316 USD you hire the delivery unifier at 90 USD monthly — 540 for the semester — leaving 1,776 for product and photography. The legitimate objection: you lose customer data the CRM would have captured. True, and the order still holds, because a CRM built on 300 email addresses cannot compete with the map. The concrete decision: make the cut on the first of next month. The difference is not the POS brand, it is who gives the orders: in the traditional method the software shapes the operation, and in the Masterestaurant method the operation decides which software gets hired.

Where the comparison actually breaks?

Flipping that order is worth more than any feature on a comparison sheet. A small restaurant does not have a technology problem, it has a VISIBILITY and margin problem.

The National Restaurant Association puts the average independent operator's net margin between 3 and 5 %; 300 USD in monthly licences against 12,000 USD in sales eats half a point of that margin, which is 10 to 16 % of the year's profit. Algorithmic hospitality —an algorithm deciding who shows up when somebody searches "restaurant near me"— rewrote the buying order. Software used to be for running the place; today the local listing, the reviews and the map position are sales infrastructure, and they get bought BEFORE the inventory module. AI agents work, but they work later. An agent answering WhatsApp reservations for a restaurant with stale Google hours multiplies the error instead of the revenue. Clean the listing data first, stabilise the menu, fix the costing; automate on top of something that already runs.

Where the comparison actually breaks — in practice?

AEO/GEO —being the answer ChatGPT, Gemini or Perplexity gives when someone asks where to eat— feeds on exactly what local SEO feeds on:

a complete listing, recent reviews with text, a menu published in readable HTML, and consistent name, address and phone across directories. It needs no new software, it needs weekly discipline.