Restaurant Marketing OS
For independent and growing UK hospitality
Every restaurant marketing tool will help you make a calendar. None of them will tell you that the calendar is a fantasy. This one compiles your campaign from playbooks built on real hospitality lead times, gives every task a date, an owner and a dependency — and then says plainly whether the plan you have asked for fits the time you have left.
No signup, no billing, no credit card. You pick a person and you are them.
“With twenty-six days you cannot develop, cost, approve, design and print a festival menu. Drop the special menu and the photography. Keep the website page.”
Restaurant marketing used to be shared across brand, campaigns, social, CRM, digital, PR, content, events, partnerships, website, paid, reporting and project management. Independent operators cannot afford that department, so it lands on one or two people. What follows is not a shortage of ideas. It is marketing chaos and no operational structure.
The owner, the GM or whoever answers the phone does the marketing between shifts. There is no plan, only whatever is urgent this week.
A single junior marketer is the whole function — brand, social, CRM, PR, website, paid, events — and spends ninety per cent of the week on Instagram and email because those are the jobs that shout.
Two marketers covering six restaurants. That is not a capacity problem you fix by working harder. It is an allocation problem, and one site is always quietly being neglected.
Consultants deliver the strategy deck. Nobody turns it into dated work with names against it, so by March it is a PDF nobody opens.
All four fail the same way. Seasonal work gets started too late to be done properly, and nobody finds out until the printer says no. The point of this product is to move that discovery to the day you plan the campaign, instead of six weeks before the event.
A playbook is not a plan. It is a thing that emits plans. Give one an event date and a set of ticked tactics and it produces the dated, owned, diagnosed campaign for exactly that combination — twenty-nine tasks or fifty-nine, depending on the boxes.
Five steps, in order: what the restaurant is, an honest marketing audit, the findings, what the year is actually for, and only then campaigns. The order is the argument — picking campaigns first is how you end up with a calendar full of activity that ladders up to nothing.
A special menu. Photography. Long-lead PR. An email push. A website page. Every tick box is a real block of work with real dependencies behind it, not a checklist item.
The playbook holds no dates at all — it holds anchors, offsets, effort in working days, and which calendar those days are counted in. Supply the event date and out comes the dated, sequenced, assigned plan for exactly the boxes you ticked.
Deliverable, tight, or not deliverable — and if it does not fit, precisely which tactics to drop to make it fit, or the news that dropping things will not save it and the date has to move.
Below is the local food festival playbook, with the same tactics ticked both times, compiled against today’s date. Nothing here is a screenshot or a mock-up — the page ran the compiler when you loaded it, and these are its answers.
Asked in good time
Deliverable95 days out
Festival opens 4 Jan 2027
The verdict. Deliverable — the plan fits the time available.
Asked the way it usually happens
Not deliverable26 days out
Festival opens 27 Oct 2026
The verdict. Not deliverable — the campaign spine itself does not fit. You need about 48 more days.
Nothing is hard-coded. The right-hand column fails because the menu chain — develop, cost, approve, design, print — genuinely needs about sixty days, and menu development only happens on quiet kitchen days. Change the date and the answer changes. Untick the special menu and the answer changes. That is the whole machine.
It is also the argument you can send to your owner as a link, because the state lives in the URL. “Look — with these dates we cannot print.”
Any tool can produce a plausible-looking schedule. Four things make this one survive contact with a kitchen.
Kitchens develop menus on quiet days, Monday to Wednesday. Designers and printers work Monday to Friday. The restaurant trades all seven. Effort is counted in the calendar the work happens in, which is why the dates hold up.
Untick photography and the tasks behind it are removed — then everything that depended on them is rewired onto the surviving work. You never orphan a constraint by simplifying a campaign.
A forward pass gives the earliest each task can start. A backward pass from the event gives the latest it can start. The difference is float, and the zero-float chain is your critical path — the work where a two-day slip costs you the campaign.
An approval is a task with an approver on it. That is why "decisions waiting on you" and "what is holding up the print run" are the same question, answered on the same screen, sorted by the delay they are causing.
17 playbooks covering the hospitality year, 666 tasks and 140 tactic switches between them. Each one is a versioned file, and each lead time in it is a falsifiable claim about how hospitality works — which means you can argue with it over a table, and we can be wrong in public and fix it.
The number most operators get wrong is at the top of this table. Christmas party season needs 168 days. If you start it in August, no amount of effort recovers it — and the product will say so on the day you create the campaign rather than in November.
| Playbook | Lead time | Tasks | Tactics |
|---|---|---|---|
| Christmas party season | 168 days | ||
| Brighton Fringe | 115 days | ||
| Brighton Pride | 100 days | ||
| OctoberBest | 97 days | ||
| Local food festival participation | 90 days | ||
| New Year's Eve | 86 days | ||
| Chef collaboration / supper club | 80 days | ||
| New menu launch | 70 days | ||
| Summer al fresco | 70 days | ||
| Local partnership / producer collaboration | 66 days | ||
| Veganuary | 61 days | ||
| Valentine's Day | 57 days | ||
| Easter family lunch | 53 days | ||
| Mother's Day | 52 days | ||
| Father's Day | 51 days | ||
| Halloween night | 47 days | ||
| Review and reputation drive | 29 days |
Read live from the playbook files. Correct a lead time in one of them and this table changes.
Compiling the plan is the beginning. The rest of the product is the operational discipline a marketing director would bring — the bit that is normally missing between a strategy deck and a Monday morning.
A marketer needs to know what to do next and what has run out of float. An owner needs to know what is running, what needs a decision from them, and what it produced. Showing an owner a task list is how these systems lose their sponsor in week three.
The point of an annual planner in hospitality is not seeing what is happening — it is seeing what has not been started. Christmas sits eleven months away on the grid, which is roughly how far away it is when the work should begin.
Everything waiting on a sign-off, ordered by how much delay it is causing rather than when it arrived. A menu approval with no float ranks above a social post with three weeks of it. Approving writes a name and a timestamp.
Multi-site operators get one question answered properly: which site is being neglected. Compared on overdue work, work with no float, sign-offs sitting unanswered, and campaigns in preparation — plus the honest column, sites with nobody named on them.
What we said we would do against what happened. Every KPI is a number a human looked at and entered, so nothing on the page is a dashboard artefact.
"We allowed seven days for print and it took eleven." That learning attaches to the playbook, not to the person or the campaign — so it resurfaces next year on the screen where somebody is about to make the same decision again, even if they have never met whoever wrote it.
Four places, all of them named:
What it does not do is decide whether your plan is deliverable. A campaign the AI invents goes through exactly the same compiler as a hand-authored one, is held to the same lead times, and gets the same verdict — and if it produces something that will not compile, the errors go back to it to fix. The judgement stays with the corpus, and the corpus is a set of files a human can review.
Every AI feature has a deterministic fallback. With no API key configured the product is still completely usable — it just stops sounding like a person.
Worth knowing before you spend an afternoon on it.
The demo is seeded with three Brighton operators and eighteen people, and two of the campaigns in it are deliberately broken — because a demo where everything is fine demonstrates nothing. A three-site Thai group where nobody is mapped to designer, photographer or PR. A one-site restaurant doing well with no slack anywhere. A six-site group where one site is quietly being neglected.
Pick any of the eighteen and you are them. Start with the head of marketing at the six-site group, then sign in as their CEO and see how differently the same data reads.