Techniques for Increasing Sales in Foodservice: From Menu Engineering to CRM
You can grow restaurant sales without new guests in two ways: raise prices, or change what each guest orders. In 2026 the first route is hitting a wall of falling traffic. The second rests on menu engineering, service standards, CRM data and the right staff incentives. Below are the metrics, formulas and a worked calculation.
The 2026 market: higher checks, fewer guests
Poster analysed anonymised sales data from 11,555 Ukrainian venues. For January–July 2026, the average check rose 17% and attendance fell 14% against the same period of 2025. Revenue gained only 3%. For every 100 visits last year, about 86 remain.
One nuance from the same report matters: the check did not grow because guests order more dishes, but because venues repriced menus to cover food, wages, rent and logistics. Revenue growth therefore does not equal profit growth. Further price increases will support the check but will shrink traffic even more.
The practical conclusion: price has a ceiling, while order structure does not. That means how many items a guest takes, how many of them carry margin, and how often the guest returns. That is what the techniques below address. For pricing itself, see "Menu Pricing: Price Psychology and the Profit Check Formula".
Measure first: five metrics every technique depends on
Upselling without measurement becomes "it feels better". Pull 8–12 weeks of POS data and calculate:
Item attach rate = checks with the item / all checks
Items per check = items sold / number of checks
Margin per guest = (revenue − food cost) / number of guests
Server spread = (best server check − weakest server check) / average check
Compare servers on the same shifts and in the same zone. In venues without sales standards the spread can reach 30–40%, and it is the cheapest reserve you have: lifting average servers toward the best requires no capital. Accurate costing is the foundation, covered in "We calculate Food Cost correctly".
Menu engineering: what to promote and what to drop
The popularity-and-margin matrix was proposed by Michael Kasavana and Donald Smith in 1982 and remains the core tool. An item counts as popular if its share of category sales is at least 70% of an even share (1/N). It counts as high-margin if its contribution margin is at or above the category's weighted average.
An example with six main courses (1,650 portions a month, weighted average margin 222.75 ₴, popularity threshold 11.7%):
| Dish | Portions | Share | Price | Food cost | Margin, ₴ | Class |
|---|---|---|---|---|---|---|
| Burger | 520 | 31.5% | 285 | 34% | 187 | Workhorse |
| Pasta carbonara | 430 | 26.1% | 290 | 19% | 235 | Star |
| Sous-vide chicken | 360 | 21.8% | 260 | 40% | 155 | Workhorse |
| Ribeye | 140 | 8.5% | 790 | 42% | 460 | Puzzle |
| Grilled salmon | 110 | 6.7% | 520 | 48% | 270 | Puzzle |
| Mushroom risotto | 90 | 5.5% | 295 | 27% | 215 | Dog |
Look at the salmon: a 48% food cost looks bad, yet it earns 270 ₴ per portion, more than the burger or the chicken. Rank items by absolute margin and use food cost percentage as a control. What to do with each class:
- Stars: leave the recipe and price alone, give them prime placement and keep quality consistent.
- Workhorses: raise the price 5–8% or trim the portion by a few grams. Pair them with a high-margin side or sauce.
- Puzzles: rewrite the description, move them to the menu's focal zone and make them the server's recommendation. If 10% of workhorse orders shift to ribeye and salmon, that is 88 portions × (376 − 174) ≈ 17,800 ₴ extra margin per month.
- Dogs: remove or rework them. The exception is an item that serves a distinct need, such as a vegetarian option.
Eight upselling techniques for the floor
Each technique is tied to a moment of service and measured by its own metric. That is what separates a standard from "be more proactive".
| Technique | How it works | Sample phrase | Metric |
|---|---|---|---|
| Choice instead of yes/no | A two-option question lowers the refusal rate | "Tiramisu or cheesecake with your coffee?" | Dessert attach rate |
| Chef's recommendation | One specific item with a reason, not a menu recital | "Tonight's ribeye is 21-day aged; the chef suggests medium" | Share of Puzzles |
| Pairing | A drink presented as part of the dish | "A dry white goes well with the salmon, shall we do a glass?" | Drinks per guest |
| Size-up / premium | A larger or better version for a small surcharge | "Double espresso for +20 ₴" | Premium option share |
| Price anchor | An expensive item makes its neighbour look reasonable (anchoring and decoy effects) | Ribeye at 790 ₴ next to salmon at 520 ₴ | Target item share |
| Welcome aperitif | The drink decision is made before the menu is opened | "Shall we start with prosecco or a lemonade?" | Aperitif attach rate |
| Dessert after plates are cleared | The offer lands in a pause, not with the bill | "Shall I bring the dessert menu now or a little later?" | Checks with dessert |
| Set menu or business lunch | A fixed bundle raises items per check | "Soup, main and a drink, better value together" | Items per check |
One principle holds throughout: the suggestion must be specific, short and leave the guest free to decline. Pressure buys a short-term check bump and loses repeat visits.
CRM and POS: prompts built on data, not hunches
When guest order history sits in the POS, the server sees a trigger on screen and makes a suggestion that has already worked before. Five scenarios worth configuring first:
- first visit: introduce a signature item and record preferences;
- a guest's favourite dish: suggest its pairing, not the dish itself;
- an allergy or restriction in the notes: offer a safe alternative instead of silently avoiding the topic;
- no visit in 45 days: a personal message with a reason to return;
- birthday: a dessert on the house that costs you no more than its food cost.
Why this matters economically: according to research by Frederick Reichheld (Bain & Company, Harvard Business Review, 1990), a 5% increase in customer retention raises profit by 25–95% depending on the industry. For a restaurant with falling attendance, winning back an existing guest is cheaper than acquiring a new one. For the link between POS and margin, see HoReCa Automation 2026.
Collect personal data only with the guest's consent and in line with applicable data-protection law, including Ukraine's Law on Personal Data Protection. Store only what you actually use for service.
Staff incentives: bonus on margin, not revenue
If a server is paid on revenue, they will sell expensive dishes with a high food cost. If they are paid on margin per guest, their interests align with the venue's P&L. A practical scheme for a pilot:
- Metric: shift margin per guest. The bonus pays for exceeding the baseline of the previous 8 weeks.
- Safeguard: the bonus is paid only if guest ratings and complaints stay at or above a threshold you set.
- Team component: split part of the bonus across the whole shift to avoid fights over "profitable" tables.
- Training: a five-minute pre-shift briefing (dish of the day, one Puzzle, one drink pairing) and tastings of new items for the team.
- Transparency: show the team why a sauce at 15% food cost is worth more than a second beer at 45%. People who understand the logic sell more naturally.
Sales pressure without support quickly leads to burnout and turnover, which cost more than any upsell earns. See "Mental health and burnout prevention".
Kitchen and the 86 list: the costliest losses at the upsell stage
The best script is useless if the item is unavailable or takes too long. Take an item with a 200 ₴ margin ordered 10 times a day: every day on the 86 list costs 2,000 ₴ of margin, roughly 60,000 ₴ a month. The warewash area works the same way: when a clean glass takes too long, servers stop suggesting wine. An under-counter machine cycle usually runs one to three minutes depending on the programme, and at peak that difference becomes a difference in drinks per guest.
That is why equipment decisions should use TCO (total cost of ownership), not just purchase price. For the trade-offs of new versus used equipment, see "Choosing between new and used equipment".
A 90-day plan
- Days 1–14. Audit. Export the POS report, calculate the five metrics, build the menu matrix and pick one shift for the pilot.
- Days 15–45. Pilot. Introduce two techniques (for example, the welcome aperitif and dessert after plates are cleared), run briefings and compare the pilot shift with a control shift.
- Days 46–70. Scale. Add CRM triggers, rewrite the Puzzle descriptions and raise the price of the workhorses.
- Days 71–90. Lock it in. Launch the margin bonus, update the standards, repeat the measurements and compare against the baseline.
Common mistakes
- counting revenue instead of margin per guest;
- promoting the whole menu at once instead of two or three items;
- using discounts instead of recommendations, which erodes margin and trains guests to wait for deals;
- overloading servers with dozens of on-screen triggers;
- neglecting quality: upselling a mediocre dish does not bring the guest back.
Calculator: how much margin upselling adds
Enter your own figures. The starting values come from the example in this article.
Want your POS report reviewed?Write to us: we can help build the menu matrix, choose techniques for a pilot and estimate the expected effect on margin.
Discuss collaborationFrequently asked questions
Sources and method
Poster, study of 11,555 venues, January–July 2026 versus the same period of 2025 (as reported by The Page, 12 August 2026). Kasavana M., Smith D. Menu Engineering, 1982. Reichheld F., Sasser W. "Zero Defections: Quality Comes to Services", Harvard Business Review, 1990. Huber J., Payne J., Puto C. "Adding Asymmetrically Dominated Alternatives", Journal of Consumer Research, 1982. Law of Ukraine "On Personal Data Protection". The menu, check and calculator examples are an illustrative model with rounded values, not data from a specific venue. The uplift ranges (5–8% and 10–15%) are working targets for a pilot.
Author: Ruslan, Independent HoReCa Analyst, exclusively for eeat.com.ua
