The restaurant social media playbook: 30, 60 and 90 days
This is an operating plan, not a theory of marketing. It assumes you have a kitchen to run, one person with limited time, and no appetite for a channel that quietly dies in month two.
Most restaurant social media plans fail in the same place: week five. The launch enthusiasm is gone, the person who volunteered has picked up extra shifts, and the account goes quiet without anyone deciding to stop.
The fix is structural. Decide who owns capture and who owns approval, and make those two different jobs. Set a cadence low enough to survive a bad month. Pick two numbers to watch and ignore the rest. Then leave the plan alone for ninety days, because thirty days of restaurant social data is noise.
What follows is a plan you can run with or without software. If you are running Bolta, the drafting and scheduling steps collapse into an approval queue, but the ownership decisions and the weekly rhythm are still yours to make.
Days 1-30: set it up and make it survivable
The first month is about removing every reason the system could stop. Do not chase growth yet.
Start by picking your platforms honestly. For most restaurants that is Instagram plus one — Facebook if your guests skew local and older, TikTok if you have a kitchen worth filming and someone willing to film it. Two platforms done properly beats five neglected ones. Connect them and check that every profile has current hours, the right address, a working reservation or ordering link, and a phone number someone answers.
Then split the two jobs. Capture is a kitchen and floor job: someone shoots three things a day into a shared album. Approval is an owner or general manager job: someone with the authority to say no to a price, a claim or a photo. Never assign both to the same overworked person, because when they get busy you lose the whole channel at once.
Write down the voice in five lines. What you sound like, what you never say, whether you use exclamation marks, how you refer to guests, and what topics are off the account entirely. This document is short on purpose. If you are running Bolta, this is exactly what the agents work from, and it is what makes the first drafts usable rather than generic.
Set the cadence at three posts a week and hold it there for the full month. Do not increase it because a post did well.
Days 31-60: build the rhythm and start replying
Month two is about the loop, not the output. By now you should know which capture habit sticks and which one you invented on day one and never did.
Move to a fixed weekly meeting, fifteen minutes, same slot every week — a slow afternoon, not a Monday morning. Three things happen in it: review what went out, approve the next fortnight's queue, and note anything coming up that needs a post. That meeting is the entire management overhead of the channel.
Start treating replies as part of the job. Comments and direct messages on a restaurant account are usually questions about hours, dietary options, parking or reservations, and answering them within a day converts more covers than any post. Assign this to the floor, not the kitchen, because the floor already answers these questions on the phone.
Begin the second platform properly if you have not. In month one it was enough to have it live; in month two it should get content shaped for it rather than a copy of the first platform's post.
This is also the month to look at your own edits. If you rewrite the same thing every time — cutting hype, adding the price, changing how you name a dish — that is a voice instruction you have not written down yet. Add it to the five-line document. In Bolta, those edits also feed the voice model directly, so the pattern gets absorbed rather than repeated.
Days 61-90: measure two things and decide
Ninety days is the first honest read. Judge it on two numbers and ignore the rest.
The first is saves and shares, not likes. A save on a restaurant post is somebody planning to come. A share is somebody recommending you. Likes are ambient. If saves are climbing month over month, the content is working even if follower growth is flat.
The second is inbound: direct messages, reservation clicks, calls that mention something you posted, catering enquiries. Ask the host stand to note when a guest mentions a post — it is unscientific and still the most useful signal you will get.
Then decide one of three things. If both numbers are rising, keep the cadence and improve the photography, which is usually the biggest remaining lever. If saves are flat, the posts are probably announcements rather than specifics — check whether each one names a concrete thing. If the channel is not being fed at all, the problem is ownership, not content, and the fix is reassigning capture to someone whose shift allows it.
What you should not do at ninety days is start over. Restaurant accounts compound slowly, and the operators who win are the ones still posting three times a week in month nine.
A review checklist
A process for the person who approves posts. Adapt it to the policies your restaurant already has, and check anything you are unsure about with whoever owns compliance where you work.
Confirm the facts before approving
Hours, prices, dates, capacity and any ingredient claim. Each of these has an owner in the building — the chef for what is in a dish, the manager for hours and pricing. Check with the owner, not from memory, and do it at the approval step so it happens once.
Check consent for every face
Guests, staff and anyone in the background. If someone is identifiable and has not agreed, either get the agreement or use a different frame. Keep a simple note of who has said yes to being featured so you are not re-asking the same team member every week.
Check reposts for permission and credit
Before approving any guest photo, confirm the guest agreed, the credit is attached, and the image has not been cropped in a way that changes what they showed. Have a standing rule that any request to remove a repost is honoured the same day.
Read it as a guest who has never been in
The most common failure in restaurant posts is assumed context. Would someone who has never eaten with you understand what the dish is, where you are and how to come in? If not, add the missing sentence before approving.
Confirm the image matches the caption
Wrong plate, old menu version, a photo from a different season. Scheduled queues drift, and an image mismatch reads as carelessness. Check the pairing at approval, not at publish.
Check the timing against your actual calendar
Closures, private events, holiday hours and staff shortages. A scheduled post promoting a service you are not running that day is the single most avoidable own goal in restaurant social media.
Pause the queue when something goes wrong
An incident in the neighbourhood, a bad news cycle, a closure. Have one person with the authority to pause everything scheduled, and make sure they know how. Cheerful scheduled content during a bad moment is worse than silence.
Review edits monthly, not post by post
Once a month, look at what you changed most often across your approvals. That pattern is a voice rule you have not written down yet. Add it to your voice document.
Where Bolta fits
Bolta’s agents research and draft posts, then hold them for review. The approval workflow is designed so that a human reviews every post before publication. Bolta does not decide what is appropriate for your restaurant to say, and it does not replace the review your restaurant already does. Follow the policies you already have, and consult your compliance officer or counsel on anything specific to your situation.
Frequently asked questions
Who should own social media in a restaurant?
Split it. Capture — photos and clips — belongs to whoever is in the building during service, usually a cook or a floor lead. Approval belongs to an owner or general manager with authority over pricing, policy and staff photos. Assigning both to one person is why most restaurant accounts go quiet in month two.
How long before restaurant social media shows results?
Plan on ninety days before you judge it. Thirty days of restaurant social data is noise, and the compounding effects — saves, shares, people who followed months ago and finally book — take longer than a single quarter to appear. Set a cadence you can sustain for nine months rather than one you can sprint for six weeks.
What should a restaurant measure on social media?
Saves and shares first, because a save means someone is planning to visit and a share means someone is recommending you. Second, inbound: direct messages, reservation link clicks and enquiries. Follower count and likes are the least useful numbers available. Ask the host stand to note when guests mention a post.
Should we pause posting when something goes wrong?
Yes, and decide in advance who can do it. A local incident, a closure or a public complaint makes a cheerful scheduled post read badly. Name one person with the authority to pause the queue, make sure they know how, and treat resuming as a separate decision once the situation is clear.
How does Bolta fit into a restaurant social media playbook?
It replaces the drafting and scheduling steps, not the ownership decisions. The agents write and queue posts from your context and photos; your approver clears the queue in a couple of minutes a day. The weekly meeting, the capture habit and the two numbers you measure stay exactly as described here.
See what Bolta would write for your restaurant
Bolta’s agents research, draft and schedule posts. Nothing reaches a public account until a person on your side approves it. Paid plans start at $19/month, and there is a free Starter tier.
Last updated: 2026-07-27