The usual way an owner decides whether training worked goes like this. Someone pulls the ticket medio for May, pulls it again for July, sees it went up by three euros and says the course paid for itself. July has more British families ordering puddings, more German couples ordering a second bottle of Binissalem, and a menu that was probably repriced in June. None of that is training.
If you trained your team, or you are about to, you want a number you can believe. That means comparing like with like, taking price changes out, and looking at each server and each moment of the meal separately. None of it needs special software. You need your till exports, a spreadsheet and about an hour a week.
Compare periods that are actually comparable
The fairest comparison is the same weeks, the same days and the same service. Four weeks before training against four weeks after is a sensible minimum. One week each side is too easy to distort with a single 14-person birthday table or a wet Tuesday that emptied the terrace.
Some rules that save arguments later:
- Lunch and dinner separately. A menú del día lunch and an à la carte dinner are different businesses. If lunch grew as a share of covers after training, your blended spend per cover falls even if every server sold better.
- Avoid calendar traps. Easter, local fiestas, a cruise ship day in Palma, the last week before a hotel closes. If one side of the comparison has one of these and the other does not, move the window.
- Same cover rules both sides. If you counted children as covers in the baseline, count them after. If bar stools without a table number were left out before, leave them out now.
- Look at total covers too. A higher ticket medio on fewer guests can still be a worse month. In August 2026 Juan Miguel Ferrer, president of CAEB Restauración, told Mallorca Magazin that restaurants are taking in less even as tourist numbers rise, and forecast August revenue down 12 to 14 percent on last year. Spend per cover is one number among several.
If you can, keep a comparison group. That might be one server who has not been trained yet, or the dinner service if you only coached the lunch team. If both groups rise by the same amount, the season did it, not the training.
Take the price rise out before you celebrate
This is where most before-and-after numbers fall apart. If the pa amb oli went from €9.50 to €10.50 and the house red by the glass from €4 to €4.50 between the two periods, your spend per cover rises even if every guest ordered exactly the same thing.
The fix is simple. Take the items sold in the before period and price them at the new menu prices. That gives you what the before period would have earned if nothing had changed except the prices. The gap between that figure and your real after-period figure is the part that came from guests ordering more or ordering differently.
A worked hypothetical, for dinner service only:
| Figure | Spend per cover |
|---|---|
| Before training, at old prices | €31.40 |
| Before-period items, repriced at new prices | €32.30 |
| After training, actual | €33.90 |
| Price effect (32.30 minus 31.40) | €0.90 |
| Ordering effect (33.90 minus 32.30) | €1.60 |
The headline rise is €2.50. The honest figure for what changed at the table is €1.60. Still worth having, and now you can defend it.
An export of sales by item for a date range is all you need. If you only changed five prices, you can do this by hand on those five lines. INE publishes the monthly IPC with breakdowns by autonomous community if you want a general inflation check, but your own menu is a far better price index for your own restaurant.
Track spend per cover by server
Many TPV systems can report sales by camarero or by user login; check yours before training starts. Export it for both periods and build a small table: covers served, spend per cover before, spend per cover after.
| Server | Covers before | Covers after | Spend before | Spend after | Change |
|---|---|---|---|---|---|
| Marta | 610 | 590 | €30.80 | €33.10 | +€2.30 |
| Tom | 540 | 560 | €32.10 | €32.60 | +€0.50 |
| Jordi | 580 | 600 | €31.00 | €33.80 | +€2.80 |
| Lena | 95 | 110 | €29.50 | €35.20 | +€5.70 |
These are invented names and numbers to show the method. Reading a table like this, a few things matter more than the ranking:
- Small samples lie. Lena's jump looks like the star result, but on 110 covers the whole rise is about €630 (110 × €5.70). Two large groups that each ran up a few hundred euros on cava and wine could account for most of it. Give a server at least a few hundred covers before you draw conclusions.
- Sections are not equal. If Jordi moved from the inside room to the front row of the terrace with the sea view, his tables changed, not just his selling. Note section changes in the same sheet.
- Who rang it up? If servers share a login, or the bar rings in the drinks for the terrace, the per-server figures mean little. Fix the login habit first, then measure.
- Flat is information too. Tom barely moved. That might mean he already sold well before, or that the approach did not suit him. Both are worth a quiet conversation, not a public league table.
Track the moments, not just the total
Spend per cover tells you something changed. It does not tell you what. For that, count how often each suggestion lands. These are attachment rates, and they are easy to pull from an item report.
| Moment | What to count | Divide by |
|---|---|---|
| Water | Bottles of water sold | Tables |
| Starters | Sharing starters (croquetas, pimientos de padrón, calamares) | Tables |
| Second drink | All drinks sold (the till cannot tell first rounds from second, so watch the per-cover figure rise) | Covers |
| Dessert | Desserts, including ensaïmada or gató with ice cream | Covers |
| Sobremesa | Coffees, carajillos, hierbas | Covers |
If you trained mainly on dessert and coffee, those two lines should move most. If dessert attachment is flat but spend per cover rose, look elsewhere: maybe a new fish of the day at a higher price is carrying the number.
Run them by server as well. A server whose dessert rate doubled while their water rate stayed flat has learned one habit and skipped another, which is a far more useful thing to know than a single average.
Decide what counts as a real result before you look
The most common mistake is deciding the rules after seeing the numbers. Write down, before training starts, which weeks you will compare, which services, how covers are counted and how you will handle price changes. Then stick to it, even when the result disappoints.
A few signs that a rise is real rather than noise:
- It holds after the price effect is removed.
- It shows up across most servers, not one.
- The attachment rates for the trained moments moved in the same direction.
- It holds for a second four-week block, not just the first enthusiastic fortnight.
That last one matters. Teams often sell harder in the weeks straight after a session and drift back by the third or fourth week. Measuring a second block tells you whether the habit stuck or whether you need a refresher in the pre-shift briefing.
This is also how we would want to be judged. Otra Ronda's results-based option charges €690 plus 12% of measured uplift, ex-IVA, and "measured" only means something if the baseline, the comparison weeks and the price adjustment are agreed in writing before anyone is trained. With an outside trainer or without one, the same discipline protects you from paying for, or praising, a rise that the menu reprint delivered.
A one-page routine
- Every Monday, export last week's sales by item and by server, split lunch and dinner, into the same sheet and columns.
- Keep a notes column for price changes, section moves, big groups and weather.
- At the four-week mark, reprice the baseline, compare by server and by moment, and write three lines on what changed.
November is a good month to set this up. Covers are lower, the team has time, and by the time the terraces fill again next spring you will have a method you trust instead of a July number you want to believe.