Two servers, identical wording, same shift, and one gets four times the reviews. This is the normal outcome, and it is why script-focused programmes disappoint — the variance between people asking is much larger than the variance between forms of words.
Worth understanding what actually drives it, because the usual explanation ("some people are just better with guests") is both true and useless.
The three things that make an ask land
Whether the guest has been served by that person. A review request from someone who has looked after you all evening is a small favour between two people who have a relationship. The same request from the person who happens to be at the till is a marketing prompt from a stranger. This is the largest single factor and it is a rota question, not a training one.
Whether the asker has done something specific for them. The server who remembered the allergy, sorted the wrong drink, or recommended the dish the table loved is in a completely different position from one who took the order correctly. Recovery and recommendation both create the conditions for a successful ask, which is why the manager who fixed a problem is often the strongest asker in the building.
Whether the asker sounds like they mean it. Detectable instantly. A server who thinks review-begging is beneath the venue will transmit that in four words, and guests respond to the discomfort rather than the request.
None of these is about charisma, which is the useful part. All three are arrangeable.
Why the best server is often the worst asker
Consistently observed and initially confusing. The most skilled front-of-house people are frequently the most reluctant to ask.
Two reasons. They have a strong sense of the service they provide as complete — asking for something at the end feels like undercutting it. And they are usually the ones running the busiest section, so the payment moment is the moment they are furthest behind.
Meanwhile the person who converts best is often a mid-tier server with a warm manner and an unhurried section. Which means the ask is a section-planning decision as much as anything else: whoever is asking needs three spare seconds at the table, and your strongest server never has them.
What to do with that
Do not make it universal. A venue where all fourteen staff are supposed to ask produces less than one where three people ask reliably. Universal requirements dissolve into nobody's responsibility, and inconsistent asking is worse than none because guests occasionally get asked twice.
Pick two or three people and make it theirs. Choose for warmth and for section load, not seniority. Tell them explicitly that it is their job and that nobody else is doing it.
Give the recovery ask to managers. Anyone who resolves a complaint is in the best position in the building afterwards, and it produces the most useful reviews you can get — the specific wording is in the scripts piece.
Let the reluctant ones opt out, openly. A senior server who says they would rather not is telling you something accurate about how it would come across coming from them. Forcing it produces a bad version.
The tasting problem underneath all this
The staff who ask most confidently are, almost without exception, the ones who have eaten the food.
This is not sentimentality. A server who has tasted the shortrib can say "the shortrib is what I'd have" and mean it, which drives both spend and the conditions for a successful ask. One who has not is performing enthusiasm, and performed enthusiasm is what makes a review request feel like a script.
Most venues do a staff tasting at menu launch and never again, so anyone hired in the following eight months has never eaten anything. If you fix one thing from this piece, fix that — it is cheap, and it moves ordering friction and review volume at the same time.
How to tell whether it is working
Not by review count alone, which moves for too many reasons. Ask the two or three designated people, weekly, roughly how many times they asked. An approximate self-reported number is enough, because what you are watching for is the number going to zero — which is what happens around week three, for reasons covered in the durability piece.
And read the reviews you get for whether they name people or dishes. Reviews prompted by someone who served the table are specific. Reviews prompted by a QR code on a receipt are generic. If your incoming reviews all read "great food, will return", the asking is happening at the till rather than at the table, and the fix is who rather than what.
Tracking whether volume actually shifted after a change like this is the part that needs someone counting across months rather than weeks. OMMU does that alongside the themes, so "review volume is up a third since you gave it to Maria and Tom" is a sentence with evidence behind it.