A guest who cannot decide does not leave. They order the thing they already understand — the burger, the chicken, the pasta — and they order it without a starter, because they have used up their willingness to deliberate.
That is the whole cost, and it does not appear anywhere in your review scores. It appears in your average spend per head, which is not a number anyone connects to guest feedback.
The arithmetic, because it makes the case better than the argument
Take a venue at 120 covers a night and £34 average spend. Suppose one guest in six is a defaulter in the sense above: they order a recognisable main, no starter, no recommendation taken. Move half of them to an ordinary outcome — one starter or one upsold main, call it £7 — and you have £70 a night. About £25,000 a year.
You will not get all of it, and the figures are illustrative rather than measured. But it puts decision friction in the same conversation as a labour line, which is where it belongs and where it never is.
How it reads in reviews
It is easy to miss, because guests do not experience it as a problem worth reporting. It arrives as a mild aside inside an otherwise positive review — the kind of hinge clause discussed in the four-star piece:
- "Took us a while to work out what to have"
- "Menu's a bit overwhelming"
- "Ended up just getting the burger"
- "Wasn't sure how big the sharing plates were"
- "Would have liked a recommendation"
- "Couldn't tell what came with what"
Notice how little heat there is in any of them. This is why menu confusion survives for years in venues that respond conscientiously to complaints: it never looks like a complaint.
The last two are the most valuable. "Wasn't sure how big" is a sizing information failure, and "couldn't tell what came with what" is an inclusion failure — both are single-line menu fixes, covered in the menu language piece.
The four causes, in the order they are worth checking
Too many options in one section. Past roughly eight items in a section, guests stop comparing and start scanning for the familiar. Eleven mains is not twice as good as six; it is worse, and it costs more in prep and waste.
No hierarchy on the page. If every item has identical visual weight, the menu offers no guidance and the guest has to do all the work. One highlighted item per section changes ordering behaviour more than any wording change.
No size or intensity information. Sharing plates with no indication of how many they serve, "small plates" that vary from three bites to a meal, no heat level. Uncertainty resolves toward the safe choice.
Nobody at the table is recommending anything. The most powerful lever and the one that has nothing to do with the menu document. A server who says "the shortrib is what I'd have" removes the decision entirely, and it converts at a rate no printed design change approaches.
That last point is worth a moment. "Would have liked a recommendation" is a staffing and confidence finding, not a design one. Servers who have not tasted the food cannot recommend it credibly, so they stop trying — and the fix is a staff tasting, which most venues do at menu launch and never again.
What to do with it this week
Two moves, both cheap.
One: cut the longest section by a third and highlight one item in each. Pick the three lowest sellers in your biggest section and remove them. Nobody will write a review about their absence, and the section becomes readable.
Two: brief the floor on one recommendation per section, and have them taste it. Not a script — a genuine opinion each server is willing to state. Then listen for whether the phrase "would have liked a recommendation" stops appearing.
Measuring whether it worked
This is the unusual case where reviews are the weaker measurement. The signals to watch are average spend per head and starter attachment rate, both of which you already have in the till and neither of which needs a guest to write anything.
Reviews are the diagnostic — they tell you the friction exists and roughly where. The till tells you whether you fixed it. Operators who try to prove this one with review sentiment usually conclude it did not work, because "took us a while to decide" was never frequent enough to show a clean before-and-after.
The reason it goes unnoticed is that the evidence arrives politely, once a fortnight, filed under positive. OMMU counts those asides along with everything else, so ordering friction turns up as a theme with a number attached rather than as a phrase you half remember reading.