Picture an owner in Port de Pollença who books staff training in April. By late May the till is up on April, and everyone agrees the training worked. Except May is always up on April. The terrace is open every day, the boats are back in the bay and the British half-term crowd has arrived. The training may have done something, or nothing at all. Nobody can say, because nobody wrote down what normal looked like before the team was told anything.
That is what a baseline week is for. Seven days of clean numbers, taken from your own till before anyone is trained, so that every later figure has something honest to be compared against.
Turnover alone tells you very little
Total sales move for two reasons: how many people sat down, and how much each of them spent. A busy week with weak spend and a quiet week with strong spend can land on the same total. If you only look at the total, you cannot tell which lever moved.
This summer showed the problem in public. In August, CAEB Restauración, the restaurant association, said turnover in Mallorca's bars and restaurants fell about 14% in the first two weeks of the month, and the Majorca Daily Bulletin, reporting it, described fewer customers and a lower average spend than the year before. In September the same association described busy streets with the money not reaching restaurant tills. By its account, footfall and spend came apart, and a single turnover figure hides that.
So the baseline has to split the two. Covers on one side, spend per cover (the ticket medio) on the other, and the categories that make up that spend underneath.
What a baseline week actually is
Seven consecutive days, every service you open, recorded the same way each time. Normal menu, normal prices, normal rota. No new specials, no briefing about selling, no hint to the team that anyone is watching the dessert line.
That last point matters more than owners expect. The moment a server hears that the boss is counting coffees, coffees go up for a few days. That is nice for the till and useless for the measurement. Pull the numbers quietly, from reports the system already produces.
For each service, write down:
| Field | Why it matters |
|---|---|
| Net sales, ex-IVA | One consistent figure. Mixing totals with and without IVA wrecks every comparison. |
| Covers | The divisor. Decide once how you count children, bar stools and drinks-only tables, then never change it. |
| Spend per cover | Net sales divided by covers. The headline number. |
| Sales by category | Drinks, starters, mains, desserts, coffee and digestifs. This is where suggestive selling shows up first. |
| Units of key items | Desserts sold, bottles of water, coffees, second rounds. Counts are harder to fool than euros. |
| Sales per server | If your POS logs who opened the table. Useful later, not essential now. |
| Notes | Weather, events, big groups, anything unusual. The column people skip and later regret skipping. |
Leave tips out of the sales figure, and keep invitations separate. The chupito de hierbas on the house is good hospitality, but it is not a sale.
A rainy Tuesday is not a fiesta Saturday
The usual mistake is not bad arithmetic. It is comparing days that were never comparable.
Mallorca makes this harder. One wet afternoon closes a terrace and pushes the guests who do come inside, where the tables turn differently. The town's patron saint fiesta fills the square with locals ordering beers and pa amb oli, which drags spend per cover down while covers go through the roof. A cruise ship day in Palma changes who walks past your door at lunch. A market morning in Sineu or Santanyí, a match on the television, a puente around a bank holiday: each one shifts both covers and spend, and rarely in the same direction.
A few rules keep the baseline honest:
- Compare service with service. Tuesday lunch against Tuesday lunch, Saturday dinner against Saturday dinner. Never the week's average against one good night.
- Write down the weather. One word per service is enough: sun, wind, rain, heat. You will want it in three months.
- Flag events, do not delete them. If a fiesta landed on Saturday, keep the number but mark it. When you measure again, either find a comparable Saturday or leave that service out of both sides.
- Set big groups aside. A table of 18 on a closed set menu tells you about the set menu, not about your servers. Record it, but calculate spend per cover with and without it.
- Freeze prices and menu. If you add €1.50 to every main the week after the baseline, spend per cover rises without anyone saying a word. Change prices before the baseline week or long after the comparison, not in between.
What one week can look like
A worked hypothetical: a 60-seat restaurant in Sóller, open lunch and dinner, measured in a quiet week in spring. Figures are invented to show the method.
| Service | Covers | Net sales (ex-IVA) | Spend per cover | Note |
|---|---|---|---|---|
| Tue lunch | 34 | €843 | €24.80 | Rain, terrace closed |
| Tue dinner | 41 | €1,394 | €34.00 | Normal |
| Fri dinner | 78 | €2,808 | €36.00 | Normal |
| Sat lunch | 112 | €2,128 | €19.00 | Market day, many drinks-only tables |
| Sat dinner | 86 | €2,709 | €31.50 | Group of 16, set menu |
Look at Saturday lunch. Most covers of the week and the lowest spend. Average it in with everything else and you get a figure that describes no real service. If the comparison week has no market crowd, spend per cover will appear to jump and someone will take the credit. Same with Saturday dinner: strip out the 16 on the set menu and the remaining 70 covers might tell a different story from the total.
The useful output is not one number for the week. It is a spend per cover for each regular service, plus a short list of services marked as unusual.
Small things that break the numbers
- A server rings a table of six as one cover because the cover count field was left at the default.
- Staff meals or a family dinner put through the till as normal sales.
- Voids and comps handled differently by different shift managers.
- The bar and the dining room on separate tills, with one forgotten in the report.
- A day's Z report missing, then "estimated" from memory.
None of this needs software beyond what you already have. A notebook by the pass and the end-of-day report are enough, as long as the same person fills both in the same way every day.
Winter numbers and summer numbers
A baseline taken now, in November, is a good baseline for winter. Regulars, lunch trade, maybe a few long-stay visitors. It is not a baseline for August, and it should never be used as one. If you plan to train before next season, take a second baseline in the first steady week after opening, before the team is briefed. Or, if your POS kept last year's data, pull the same weeks from last season and compare service by service, keeping in mind that prices and the visitor mix will have moved.
The quiet months are still the right time to set up the habit. You have time to decide how covers are counted, check that the category reports work, and run one practice week so the real one is clean.
Why we start here
Every Otra Ronda engagement will begin with a baseline week measured from your till, for a simple reason: without it, any claim about training is a guess. That includes ours. On the results-based option, part of the fee (12% of measured uplift, on top of a €690 base, ex-IVA) depends on that measurement, so the baseline is the line both sides agree on before anyone sets foot on the floor.
Measuring the after-week, and reading it server by server, is its own job. But it only works if the before-week was honest. Seven days, the same columns every service, the weather in the notes, and nobody on the team told a thing.