Ask an owner which dishes carry the business and the answer comes fast and confident. Then you pull the sales report, and the "bestseller" is fifth. It happens all the time, and it is not a lack of experience: memory keeps the busy Friday and the dish a regular raved about, not the quiet Tuesdays.
Deciding your menu by feel is expensive. It keeps dishes that only cost you prep time and hides the ones quietly paying the bills. The fix is not a complicated system. It is a simple habit of asking a question, checking the data and testing one change at a time.
Why your memory misleads you
The dish you love is not always the one that sells. The dish people talk about is not always the most profitable. And the item that "always goes out" may leave almost no margin once you count ingredients and prep time. None of this is a failing; it is just how memory works. Your head stores impressions, your sales data stores counts.
The data you already have
Most restaurants have more information than they use:
- Sales by item from your POS or ordering system, by day and time.
- Ingredient costs from your supplier invoices, to work out each dish's margin.
- Dish reviews and comments from customers, which tell you why something sells or does not.
- Your team's notes: what guests ask for that is not on the menu, what slows the kitchen, what comes back half eaten.
- What you throw away, which often points at the dishes that do not move.
For the classic four-way split of dishes by popularity and margin, see our guide to menu engineering. Here we focus on how to make and test decisions.
A simple decision routine
- Ask one clear question. "Do we need four risottos?" is better than "Is the menu working?".
- Look at the last four to eight weeks. One week can be distorted by weather, a holiday or a football match.
- Decide on one change and what you expect it to do.
- Run it for a fixed period, say four weeks, without changing anything else in that part of the menu.
- Compare like with like: the same weekdays, similar weather, the same number of weeks.
- Keep it, adjust it or reverse it, and write down what you learned.
A worked example
Say you offer four risottos. You are sure they all sell. Over the last six weeks the report shows two of them account for most risotto orders, one sells a few a week and one barely moves. The quiet ones also need two ingredients no other dish uses.
Dish reviews add colour: the mushroom risotto is often praised, the seafood one is mentioned as "a bit pricey".
So you test: keep the two strong risottos, drop the slowest, and move the seafood one to a weekly special at a slightly lower price. After four weeks, check three things: total risotto orders, how much of those two special ingredients you threw away, and how the line felt on busy nights. If total orders held and waste fell, you have your answer. If customers kept asking for the dropped dish, bring it back as a special.
Mistakes to avoid
- Changing five things at once. You will never know which change worked.
- Reacting to one bad week. Wait for a pattern.
- Looking only at popularity. A dish that sells a lot at a thin margin may need a new price, not more space.
- Ignoring the kitchen. A dish that slows every other order has a hidden cost.
Do this this week
Pull last month's sales by dish and list your bottom five. For each one, write the question you would want answered. Pick one, decide your test and put the review date in your diary.
Where GetFoodies fits
On your own ordering site, the dashboard shows the orders and customers from your channel, dish reviews tell you what people love, and the digital menu can be changed whenever you decide, without reprinting. GetChef can also correct your menu text for you to approve. If you would like a menu that is easy to test, you can create your free site.