What Glovo, Pyszne.pl and Uber Eats commission really costs — run the numbers
Polish trade round-ups for 2026 put aggregator commissions roughly between 10–15% (pickup / own courier) and 25–35% with the platform’s rider. At 40 orders a day that is not “a few zloty” — it is a cook’s wage or the rent.
Restaurant owners rarely ask for a “web app”. They ask how much Glovo takes, whether Pyszne.pl is worth it, and whether they can sell delivery without giving away a third of the ticket. Below are numbers you can defend — and a calculation you can plug your venue into. Contract rates are individual; we quote publicly repeated 2025–2026 trade ranges, not a secret chain tariff. None of the aggregators publishes one national Polish rate, so anyone selling you an “official Glovo 27%” is guessing or quoting someone else’s annex.
This piece does not tell you to switch Pyszne off on Monday. The apps really do bring orders a new kitchen will not get from the pavement. The problem starts when commission becomes a standing cost of selling to people who already know you. Then you pay for access to a list you will never get back — and for marketing that does not build your own.
What Polish trade round-ups say
Hospitality sites (Foodyo, WMenu, Gastroninja, UpMenu and others) describe a commission model. The exact percent depends on the city, the package (platform rider vs own courier vs pickup) and whether you buy featured placement. Roughly, from publicly repeated 2025–2026 bands:
- Pyszne.pl (Just Eat Takeaway) — often about 10–30% of order value; lower with your own rider, higher with full platform fulfilment. Small venues rarely see below ~20% if they do not deliver themselves.
- Glovo — the most cited band is 25–35% with a Glovo rider; pickup packages are often materially lower (older package write-ups mention around 15% for pickup). That is not Glovo’s price list — it is a cluster of trade blogs.
- Uber Eats — usually in a similar band to Glovo, about 25–30% (some round-ups go to ~35%), plus in-app promotion costs.
Older reporting (including Business Insider coverage when fees were a public fight) spoke of 25–30% on delivery and even higher once marketing was stacked on. We do not treat that as a 2026 official tariff. The 10–35% envelope covers the whole model: pickup / own rider at the bottom, full platform logistics at the top. Your contract can sit anywhere in that band — and only that number hits the payout.
On top of the headline cut sit costs that are not called commission: featured placement, a higher in-app price (to “fit” the percent), disputes settled by the platform, and late third-party riders that land on your rating. Some round-ups add 1–3% for payment handling and “–20% first order” discounts that the kitchen usually funds, not the portal.
A worked example: 30 orders a day at 80 PLN
A cautious lunch-and-evening pattern in a county town or a big-city district: 30 orders a day, 80 PLN average basket, 22 selling days, 25% commission (the middle of the cited range, not the top). This is not an Instagram dining room in central Warsaw. It is volume many pizzerias, kebab shops and Asian kitchens do without posting about it.
- Platform turnover: 30 × 80 PLN × 22 = PLN 52,800 a month.
- 25% commission: PLN 13,200 a month.
- In a year (12 months, no seasonality): PLN 158,400 to the middleman.
- At 30% (upper Glovo/Uber band): PLN 15,840 / month, PLN 190,080 / year.
Thirteen thousand zloty is not “a marketing fee”. It is often a cook’s wage or the rent. And that assumes you do not mark up in-app prices — if you do, some guests drop off and regulars learn that “you are more expensive in the app”. Dual pricing (venue vs app) is rational in the books and ugly in the head: the same person compares receipts.
Per order: an 80 PLN basket at 25% is 20 PLN to the platform before food cost, the box and gas. With food cost around 30% of net and a few zloty of packaging, what remains is often about half of the same plate on the floor. The kitchen works the same. The difference goes to the middleman and the plastic.
What if it is only 10 orders a day?
10 × 80 PLN × 22 × 25% = PLN 4,400 a month, PLN 52,800 a year. Still more than running your own ordering page, a BLIK gateway and SMS. The difference: at 10 orders the aggregator may be the only source of new guests — so you do not switch it off overnight. You build your own channel beside it, not instead of it.
What if you run 40 orders a day, as in the lede?
40 × 80 PLN × 22 = PLN 70,400 of platform turnover. At 25% that is PLN 17,600 a month, PLN 211,200 a year. At 30%: PLN 21,120 / month. That is no longer “once we grow”. It is a salary, a second site, or a quiet year without a loan on the oven. The higher the volume, the more commission behaves like a tax on success: the better you do, the more you pay for the same slot in the app.
What to check in the contract before you trust a blog
Round-ups give an order of magnitude. The payout follows the annex. Before you compare “25% from the internet” with your P&L, open the contract and mark four places venues most often get wrong:
- What the percent is taken from — gross or net. With 8% VAT on food service the gap is about two percentage points of real cost.
- Whether commission includes the delivery fee the guest paid. Sometimes it does; then “25% of the basket” is not what you think.
- Who funds promotions. “–20% first order” in the app usually comes off your margin, not the platform’s budget.
- Whether there is a price-parity clause — a ban on a lower price in your own channel. If there is, a “10% cheaper on the site” plan needs another lever: free delivery, a dessert, points.
Add the tablet, the joining fee, the “super partner” pack and list placement. That is not commission, but it is a channel cost. If you buy a boost once a month, add it to the PLN 13,200 before you say “the app pays because there is traffic”.
What the aggregator will not give back
Commission hurts in the till. The larger cost sits in the data. After three years on Glovo/Pyszne/Uber you have turnover, not a relationship. The platform knows who orders on Friday at 21:17, with which extra, and how often they return. You get a bag and a ticket. If they change the algorithm or the rate tomorrow, you have no list to write to: “order direct, skip the app”.
- The customer list. Phone and history stay in their system. Switching the account off kills visibility overnight.
- The relationship. A birthday coupon means paying for a campaign in their panel.
- Margin on regulars. A guest who already knows you still generates 25% for the platform if they order in the app out of habit.
- Ad data. You cannot cheaply see which QR on the box worked, because the box sends people back to their app, not your site.
- The rating. A late third-party rider lands on your stars. A dispute “settled by the platform” does too.
When an aggregator makes sense (and when it is expensive)
At opening, in a new city, or with a kitchen nobody searches by name — the apps are a shop window. You pay for people typing “kebab nearby” at 9.30 pm. That is an acquisition cost, not a cost of sale. It gets expensive when 60–80% of delivery is repeat addresses in your district and still goes through the app because they cannot order on your site.
An aggregator also makes sense in dead hours: Tuesday 3 pm, the kitchen idle, a thinner-margin order is better than none. It makes sense when you have no rider of your own and eight deliveries a day will not fund a wage. It makes sense as a second tap beside your own. It stops making sense as the only tap through which the neighbour on the third floor has been ordering lunch for two years.
Harvard Business Review (citing Bain & Company) has long used the order of magnitude: acquiring a new customer costs 5 to 25 times more than keeping one, and a 5% lift in retention has been associated with 25–95% more profit (industry-dependent). Source: Amy Gallo, “The Value of Keeping the Right Customers”, HBR, 2014, drawing on Frederick Reichheld’s Bain work. In restaurants, “keeping” means your own channel: a box coupon, SMS, a loyalty card — not another cut on the same person.
Put that on the PLN 13,200. If even half of 30 daily orders are repeats (the address comes back), about PLN 6,600 a month is spent on people you do not need to “acquire” through the app. That is a budget for your own site, a gateway, SMS and 10% off “skip the middleman” — and you still keep more than at 25%.
Your own ordering system: what stays of those 13 thousand
Your own ordering site (menu, cart, BLIK/card, pickup and delivery zones) is paid once — plus cheap SMS and a payment-gateway fee, a fraction of the aggregator. The flow is in our article on a own online ordering system. The same panel later takes a QR menu and loyalty.
We do not promise that PLN 52,800 will move from the platform to your domain tomorrow. Guests have a habit: they open Glovo, they do not type your name. Habits break with a coupon on the box, a shorter wait (your rider knows the stairwell), a different price or points. The platform stays for strangers. Regulars you invite in. After a season the numbers show what share moved — and whether 13 thousand fell to 8 or to 4.
- Take your real contract percent (not a blog’s) × last month’s platform turnover.
- Estimate how many of those orders are repeat postcodes.
- Put a QR on the box and receipt to your site with 10% off “skip the app”.
- Keep the aggregator for strangers. Move regulars.
- After 90 days compare your own channel’s share of delivery. Keep both taps or close the dearer one.
What phase one looks like here (no suite)
An MVP is not a native app in two stores and a GPS courier fleet. It is a menu, a cart, payment, pickup, delivery zones, an “order accepted” SMS and a kitchen panel that does not need a Glovo tablet to see the ticket. Your own rider or a partner — phase two, once phase one already sells. A till / POS integration — when volume justifies it, not on day one.
If you know your annex rate — send it with order count and average basket. We will run the MVP payback on your numbers, not on an internet average.
Margin on one 80 PLN ticket: VAT, food cost, the box
Trade write-ups (WMenu, Foodyo) break down not only the percent but what is left after VAT and food. On an 80 PLN basket at 25%, the platform takes 20 PLN of gross. With 8% VAT on food service the net base is about 74 PLN; food cost at 30% of net is another ~22 PLN, plus a few zloty of packaging. The same plate on the floor leaves more, because there is no 20 PLN middleman. That is why the PLN 13,200 from 30 × 80 × 22 is not “marketing” — it is the gap between a cook’s wage and a fee for a slot in the app.
Do not mix this with one “official Glovo rate”. There is none. The 10–35% envelope is pickup / own rider at the bottom and a platform rider at the top. If your annex is 28% of gross plus a “–20% first order” funded by the kitchen, you keep less than in the 25% example. Plug your annex into the same 30 × 80 × 22 = PLN 52,800 of turnover and see whether it is 13,200 or closer to 15,840 as at 30%.
A quarterly test: how many guests come back through the app
Once a quarter export 90 days of platform orders and count how many addresses / phones repeated. A useful trigger: about 20% repeats. Many repeats = you pay 25% on people you do not need to acquire (HBR/Bain: a new customer is 5–25× dearer than keeping one; a 5% retention lift has been tied to 25–95% more profit). Few repeats = the app really acquires, but the bag gives no reason to return on your site: no QR, no coupon, no shorter wait with your own rider.
- Repeat address: 10% off “skip the app” on the box and receipt (8 PLN of 80 vs 20 PLN commission).
- One-off guest: keep the platform; do not switch it off on Monday.
- After 90 days compare your own channel’s share. The goal is not 0% Glovo — it is cutting the PLN 13,200 spent on regulars.
PLN 13,200 a month: a wage, the rent, or your own rider
At 30 orders a day at 80 PLN and 25%, PLN 13,200 leaves each month, PLN 158,400 a year. At 40 orders from the lede: 17,600 / 211,200. That is a cook’s wage or county-town rent. Some venues prefer to pay 25–35% for logistics because eight deliveries a day will not fund a rider — rational at opening. It stops being rational when most of those 30 tickets are the same district. Your own channel (site, BLIK, SMS) costs a gateway fraction, not 25% of the basket. The aggregator stays the shop window. Margin returns to the kitchen.
Frequently asked questions
- How much does Glovo take from restaurants in 2026?
- Trade round-ups most often cite 25–35% with a platform rider and materially less for pickup. The binding number is in your contract. Glovo does not publish one national tariff — we do not quote an “official percent” that does not exist.
- What about Pyszne.pl?
- Cited ranges are about 10–30% of order value, depending on the delivery model. A small venue using Pyszne riders rarely sees the bottom of that range. Historically the ordering system alone was cheaper than full “orange backpack” delivery.
- What does Uber Eats take in Poland?
- 2025–2026 round-ups put Uber Eats in a similar band to Glovo: usually about 25–30%, some write-ups to ~35%, plus in-app promos. Again: the contract, not the blog.
- Will my own system replace aggregators?
- It can, but it does not have to. A sane plan is your channel for regulars and the platform for new guests — until the numbers say otherwise. Switching off overnight kills visibility you do not yet have on your own domain.
- How long to ship an ordering site?
- Menu, cart, payment and pickup is usually weeks. Delivery, couriers and a native app — phase two, once phase one already sells. We do not start with a store app if there is no cart on the web.
- Will 10% off on my own site eat the margin?
- 10% of 80 PLN is 8 PLN. 25% commission is 20 PLN. Even after the discount you keep more than on the platform — unless the contract blocks a lower price. Then you give an extra or free delivery, not a percent.
- From what volume does an own channel make sense?
- Even at 10 orders a day at 80 PLN and 25%, PLN 4,400 a month leaves — more than a site and a gateway. At 30 orders (PLN 13,200) MVP payback is months, not years. The threshold is not a magic number; it is your annex × your share of repeat addresses.
Describe your project