Everyone can see the fee on a booking-platform invoice. The harder number is the one nobody prints: how many covers would disappear if the platform went. That uncertainty is what keeps restaurants paying commission on tables they suspect they would have filled anyway. Nobody wants to cancel a channel on a Monday and discover a quiet Saturday at the end of the week.
So the safe way to cut booking commission is a sequence, not a switch. Keep the platform running while you work out which of its covers it actually created and which it merely processed. Build a direct route alongside it. Make that route the easy one for people who already know your name. Then — and only then — decide what to renegotiate, reduce or leave, with your own numbers in front of you.
Nothing in this guide asks you to remove a source of covers on day one. The commission-free explainer makes the economic case for owning a direct route — flat cost instead of a cost that rises with your busiest nights. This guide is the operational half: how to move demand towards that route over eight weeks without gambling a full Saturday to find out what the platform was really doing for you.
Why this is a sequence, not a switch
A booking platform does two different jobs, and it charges you the same way for both.
The first job is discovery: a diner browsing the marketplace finds a restaurant they had never heard of and books. That is customer acquisition, and paying for acquisition can be rational — you would pay for it in other forms anyway.
The second job is processing: a diner who already knew your name searched for it, landed on a platform listing or a platform-controlled booking button, and completed the booking there. The platform handled the transaction, but it did not create the demand. Your sign, your food, your regulars and your reputation did.
Cancelling the platform outright treats both jobs as worthless. Staying put and grumbling treats both as indispensable. Neither is true, and you cannot tell the two apart from the invoice — the fee line looks identical for a cover the marketplace found and a cover your own name found. The only way to know is to measure, which is where the sequence starts.
First, split your covers into two piles
Open the platform's reporting for one complete, recent month. Look for fields named source, discovery, marketplace, direct, website or similar, and export the records if the interface allows it. Count covers, not reservations, so a table of six is not weighted the same as a table for one.
Write down four numbers:
- total covers recorded through the platform;
- covers attributed to marketplace or discovery activity — the covers the platform found;
- covers attributed to your own website, profile or name search — the covers you found;
- covers with no usable source — unknown.
Resist the urge to force the unknowns into either pile. An honest "unknown" line is worth more than manufactured certainty, and it tells you how much the platform's own reporting can support the decision you are about to make. If the report cannot separate sources at all, ask the supplier in writing what breakdown it can provide, and start tagging the routes you control so your own records improve from this month on.
Alongside the cover split, note the actual fees paid — from the invoice, not from memory. This guide deliberately quotes no headline commission rates: fee models differ by platform, product and contract (per-cover fees, per-booking charges, subscriptions, paid placement, or a mixture), several platforms publish pricing only on request, and any figure printed here would age badly. Checked 24 July 2026, that remains the position: your current agreement and your last invoice are the only rates that matter to this decision.
At the end of this step you have a baseline: two piles, an unknown line, and a real monthly cost. Everything else in the sequence is measured against it.
Build the direct route before you change anything
The diner does not pay your commission, so "it's cheaper for us" is not a reason for them to change behaviour. The direct route has to win on clarity and convenience, and it has to exist before you promote it.
Start with a phone in your hand. Search your restaurant's exact name and follow the result a diner is most likely to tap. The owned website should load quickly, fit the screen, and put the booking action within one vertical scroll: a tappable phone number and a short form that asks only for what the team needs — party size, date and time, contact details, and any preferences worth knowing about, such as seating or an occasion.
Be plain about what happens next. A direct booking on an owned site is typically a request the restaurant confirms: the guest states what they want, the enquiry lands with your team, and a person replies to confirm the table. Say so next to the form. A clear expectation is part of an easy route, and it is more honest than dressing an enquiry up as an instant confirmation it is not.
If you already run a reservation system that works for you operationally, keep it — the owned site can link to it or embed it. The goal is not purity. It is to stop making the commissioned route the only obvious action for people who already sought you out by name.
Finally, decide your definitions before the first enquiry arrives, and keep them for the whole eight weeks:
- direct enquiries received — every request through the owned route;
- direct enquiries confirmed — the ones a person confirmed;
- covers requested vs covers confirmed — counted separately, always.
An enquiry is never a cover until someone has confirmed it. If you let those two numbers blur, week eight's comparison will flatter the direct route and you will make the final decision on bad data. The restaurant website building guide covers the owned site itself — structure, menu, mobile behaviour — in more depth than this guide needs to.
Give regulars a reason to change the habit
A booking habit formed on a platform does not move because you built a form. It moves because the physical restaurant keeps pointing at the direct route, gently and consistently:
- a small line on the printed menu — "book direct at yourname.co.uk";
- a card presented with the bill;
- a QR code on the receipt or the collection bag;
- a line in a lawful post-visit message to guests who agreed to hear from you;
- a one-sentence script at the host desk for the team to use when regulars leave.
Give the prompt a reason where you can afford one: earlier notice of a new menu, a small direct-only allocation on busy nights, or a modest gesture funded from avoided commission. Do the arithmetic first — avoided commission minus the real cost of the incentive and the work of honouring it — and write the terms plainly. A saving that becomes an unsustainable offer is not a saving.
The same logic extends to delivery if that is part of your commission problem: keep the marketplace for genuine discovery while inviting existing customers towards an owned ordering route, and track ordering and booking as separate operations, because they are. Online ordering on TableSpark is early access.
Keep the discovery you can attribute
This is the part most "ditch the platforms" advice skips, and it is why covers get lost.
Keep attributable marketplace discovery. Shift name-led demand to the owned route. The first pile — genuinely new demand from marketplace browsing — still has a job, and commission on it may be a fair acquisition cost, judged the way you would judge any marketing spend: by contribution after food, labour and the fee, not by booking count alone. The second pile — people who searched for your name and got routed through a charged intermediary — is where the owned site, your search profile and your physical prompts should start doing the work instead.
Two cautions keep this honest. First, review contribution, not volume: a channel can send plenty of covers and still be weak after discounts and fees, while a smaller channel earns its keep by filling quiet services. Second, do not react to one good week. Seasonality, events and school holidays all move the mix; compare like-for-like periods and annotate anything unusual rather than declaring a trend from seven days of data.
The eight-week sequence
Each fortnight has actions and a measurement. If you skip the measurement, you have redecorated, not decided.
Weeks 1–2 — baseline and route
Do: pull one complete month's platform report and build the three piles: platform-found, name-led, unknown. Record covers and invoiced fees. Then make the owned route usable on a phone — booking action within one scroll, tappable phone number, a form that states clearly that a person will confirm, and several test journeys completed by the team, including every confirmation message and reply path.
Measure: the baseline numbers, written down and dated. Nothing is removed from the platform.
Weeks 3–4 — make name searches land on the owned site
Do: claim and complete your Google Business Profile. Check that name, address and phone match the owned website and your major listings, point the profile's website field at the owned domain, and review the booking links the profile shows — removing only those you are authorised and operationally ready to remove.
Measure: search your exact restaurant name on a phone and keep a dated note of what appears and in what order. This is an observation, not a ranking promise — results vary by location, device and time — but a dated note in week 3 gives week 8 something to compare against.
Weeks 5–6 — add the physical prompts
Do: introduce one or two prompts the team can deliver consistently — bill card, receipt QR, menu line — all pointing at the same owned URL. Brief the team on the wording so the prompt is the same promise at every table. If there is an incentive, record its cost and redemption terms from the first day.
Measure: direct enquiries received and confirmed, week by week, using the definitions fixed in weeks 1–2.
Weeks 7–8 — measure, then decide
Do: run the same platform report for the same period length. Put four numbers side by side with the baseline: platform-found covers, name-led covers, unknowns, and confirmed direct covers. Check cancellations and no-shows only if both routes define them the same way.
Decide: with the evidence in front of you, choose the next step — keep both routes as they are, renegotiate the rate, drop a paid placement, change tier, or begin a planned exit. Whatever you choose, you have not removed a source of covers before building and observing the alternative. That is the whole point of the sequence.
Eight weeks is a floor, not a ceiling. A seasonal business or a small sample of covers may need a second cycle before the pattern is trustworthy. The sequence loses nothing by running longer; it only loses when a step is skipped.
What to renegotiate before cancelling anything
The measurement changes the supplier conversation, because you arrive with a source split instead of a feeling.
Ask for a rate review based on the demand the marketplace demonstrably creates — the first pile, not the total. Ask whether a premium placement or promotion has produced incremental bookings, and for the evidence. Check minimum terms, renewal dates and notice periods before you set any exit date, and ask two questions about data: who owns the guest relationship under the contract, and can you export your guest list and bookings in a usable format at any time? Then export a test file, because a promise about exports is worth exactly one successful export.
Leaving is the last step, not the opening move. A smaller contract, a different tier or a dropped placement often preserves the discovery that is genuinely working while the owned route absorbs the name-led demand you were paying to process.
One exception to "no deadlines": if your booking platform is closing, the sequence has a hard end date and the order of operations changes. Quandoo, for instance, takes its last new reservations on 30 September 2026 and switches off remaining infrastructure on 31 December 2026 (per Quandoo's own notice, checked 24 July 2026). If that is you, follow the Quandoo closure guide — a time-boxed migration is a different exercise from a commission decision, and mixing the two serves neither.
Where TableSpark fits
TableSpark's job in this sequence is the owned route: a website built around your live menu, with direct booking capture at a flat monthly price. A guest states party size, date and preferences; the enquiry lands in your inbox and your own guest list; a person on your team reviews and confirms it. Guest data exports as CSV whenever you want it — it is yours. Restaurants that need operational depth beyond direct capture should run a specialist reservation system alongside the owned site, and the site can link to it.
Building is free until you publish — a timing offer, not a free plan. Published plans are £19, £39 and £69 per month, excluding VAT, with on-site reservations included from the £39 Growth plan; the pricing page has the current detail. For any card payment taken through the platform: 0% TableSpark commission · Stripe’s standard card-processing fees apply.
If the direct route is the missing piece of your sequence, start building free — nothing is published, and nothing is paid, until you choose to go live.
Frequently asked questions
Will I lose covers if I leave a booking platform?
You can, if you remove the route before understanding what it contributes — that is the failure this guide exists to prevent. Run the direct route alongside the platform, separate platform-found covers from name-led demand, and measure both for a consistent period. No sequence can promise a particular number of covers survives the move; what it can do is make sure you decide from evidence rather than from an invoice and a guess.
Can I run direct bookings and a platform at the same time?
Yes, and for most restaurants that is the right configuration for a long while, sometimes permanently. The platform keeps doing the discovery you can attribute; the owned site captures the guests who already know your name. Keep the source definitions clean and never count an unconfirmed enquiry as a cover.
What commission do UK booking platforms charge?
There is no single answer worth printing. Models differ — per-cover fees, per-booking charges, subscriptions, paid placement, or combinations — terms differ by contract, and several platforms publish pricing only on request (checked 24 July 2026). Your current agreement and most recent invoice are the figures to base the decision on, which is why the sequence starts by writing them down.
Who owns my guest data?
Whatever the relationship feels like, the contract and the privacy roles decide it. Ask the platform whether your guest list and bookings can be exported in a usable format without a support ticket, and test it before you rely on it. Store any export securely and contact only guests for whom you have a lawful basis. On an owned route this question disappears: enquiries land in your inbox and your list from the start.
How long before I can decide?
Eight weeks is the minimum for a fair comparison: two to baseline and fix the route, two for name searches, two for in-room prompts, two to measure. A strongly seasonal restaurant, or one with a small monthly cover count, should run a second cycle before treating the pattern as real. The cost of running longer is patience; the cost of deciding early is a guess dressed up as a measurement.
Build before you cut
The order is the whole method: measure the source, build the direct route, change the habit, measure again, then decide. Every step is reversible except the one this sequence tells you to take last. Start with the owned site — the website building guide covers that groundwork — and let the platform keep every cover it can prove it found.