Menu photos and dish descriptions: before and after, measured in margin

Verdict: menu photography and dish descriptions do move money, though through a channel most operators get wrong: they don't raise price, they shift the MIX. A sensory, origin-driven description pulls 6 % to 15 % of units toward whichever dishes you choose, and that is where margin appears. Photography only pays when it is ONE image per section, placed on a dish with high contribution margin; a menu covered in photos drags the average check down and flattens price perception. Before: no criteria, 62 % weighted margin. After: reworked descriptions and one photo per section, 68 % weighted margin. Same prices, same kitchen, same supplier costs.
A 120-seat grill house in Bogotá billed well and earned little. Its menu carried 74 references, twenty-one phone-shot photos and four-word descriptions: «Charcoal sirloin», «House pasta». The best-selling plate was the thinnest in margin, and that was no accident — it was the only one with a large photo on page one.
Here is the trap nobody names. Menu photography and dish descriptions is not a design question, it is a menu engineering lever, and used without cost per portion behind it, it pushes volume to the wrong place. In that grill house, rewriting descriptions and pulling nineteen of the twenty-one photos moved 9.4 points of mix toward four high-margin dishes in eleven weeks.
2026 arrives with two forces pulling opposite ways: guests increasingly decide from a phone, through QR menus and review photos, while marginal profitability audits keep showing that the most photo-heavy menus have the worst-distributed mix. Diego F. Parra has spent twenty years at that intersection, and the Masterestaurant rule is blunt: cost it first, photograph it second.
Side-by-side comparison
| Before (menu without criteria) | After (engineered menu) | |
|---|---|---|
| Menu references | ✕74 dishes, 21 with photos | ✓38 dishes, 5 with photos |
| Weighted contribution margin | ✕62 % on food sales | ✓68 % on food sales |
| Actual average food cost | ✕34.1 % (above the 32 % ceiling) | ✓29.6 % with the same suppliers |
| Average food check | ✕COP 41,200 per guest | ✓COP 46,900 per guest (+13.8 %) |
| Words per description | ✕3.8 words on average | ✓22 words with origin and texture |
| Star dishes' weight in mix | ✕18.3 % of units sold | ✓27.7 % of units sold |
| Decision time at the table | ✕6 min 40 s average | ✓4 min 05 s average |
| Cost of the change | ✕COP 0, which was precisely the problem | ✓COP 3.2 M across shoot and reprint |
Sensory descriptions remain the trend with the oldest evidence and the least uptake
Writing descriptive menu copy lifts sales of those dishes by 27 %, according to the Cornell University Food and Brand Lab, and that figure has been published for years while most menus ignore it. The measurable signal in your operation is not total sales, it is the dish's share inside its own family: if the ribeye accounted for 18 % of your beef sales and climbs to 26 % after two rewritten lines, the lever worked. An operation under 60 covers only needs to rewrite the six dishes with the highest contribution margin in cash, one per family. Above 120 covers, work in menu blocks and measure across eleven weeks, since seasonal noise swallows any reading taken over a fortnight. The rule I apply: origin, technique, time. «Aged 28 days, seared over orange wood» carries more weight than three adjectives ever will. An image shifts consumption toward the FAMILY the dish belongs to, not merely toward that dish, which is why the right question was never which plate to shoot.
A photograph does not sell the dish, it drags the whole category
Menu engineering mix concentrates between 35 % and 45 % of orders per category in the star items, according to the National Restaurant Association Operations Data Abstract 2024, so one badly placed photograph moves nearly half that category's volume somewhere you never intended. In that 120-cover steakhouse in Bogotá there were twenty-one photographs across seventy-four menu items; the dish with the big cover shot carried the worst margin. We pulled nineteen images, kept two, fixed the copy, and the mix moved 9.4 points toward four high-margin dishes in eleven weeks. Decide first which family you want to grow, reading margin in CASH, and only then reach for the camera. With the menu on a phone, guests decide using the same behaviour they bring to online shopping: they scan the photo, read two lines, move on. Some 37 % of adults order delivery at least once a week, according to UpMenu (Food Delivery Statistics 2024), and that habit trains the eye long before anyone sits at your table.
The QR code turned your menu into an e-commerce window, and almost nobody treats it that way
The measurable signal here is scroll order against average check: the first three visible items capture most of the quick decisions. Running fewer than three locations? Sort the digital menu by descending contribution margin within each family and review it quarterly. With a network of venues, version the menu per site, because the mix of a business-district location looks nothing like a weekend spot. A digital photo is not decoration: it stands in for the server who no longer suggests. Hardly any menu describes what people drink, and that is the easy money of 2026. Non-alcoholic beverage sales in United States restaurants grew 30 % during 2024, according to Restaurant Dive, while the total beverage segment reached USD 490 billion in 2025 with roughly 3 % growth, per Circana. A mocktail described with fruit, technique and temperature sells like a dish; listed as «house lemonade» it sells like tap water.
Beverages are where description still has virgin ground
Watch the beverages-per-cover index: anything below 0.8 at lunch is a leak. Small operations should describe four drinks and photograph only the most profitable one. Larger ones should move the beverage section to the top of the digital menu rather than the tail, then measure the shift across two full eleven-week cycles before drawing any conclusion. A short catalogue described with precision outsells a long one photographed halfway. Wingstop closed 2024 near USD 4.8 billion in system-wide sales with 20 % same-store growth in the United States, according to results published by the company, holding a tight menu where every item is explained the same way and shot the same way. Seventy-four items with twenty-one phone snapshots produce the exact opposite: noise. The measurable signal is mix dispersion, meaning how much revenue concentrates in your top ten dishes; below 55 % the menu is diluted.
Visual consistency, the lesson left by the high-rotation model
For a single venue the first move costs nothing: unify the description format across the whole menu before hiring any photographer. Across several venues, build a photo manual with identical lighting, identical plateware and identical angle. Declaring allergens inside the description, rather than in a footnote, is the regulatory trend about to stop being voluntary. Food allergies drive more than 200,000 emergency room visits every year in the United States, according to Food Allergy Research and Education, and no beautiful menu offsets one incident. The measurable signal is the share of items with allergens declared in line: aim for 100 % before somebody demands it. A small kitchen can solve this with iconography plus a short line closing each description; a central kitchen feeding several outlets should anchor the declaration to each dish's technical sheet, so no supplier change slips silently onto the menu. Here it meets costing: the sheet that declares the allergen is the same one calculating cost per portion, so the work happens once.
What to adopt now and what merely to watch through 2026?
Adopt three things immediately: sensory rewriting of your highest-margin dishes, a sharp cut in the number of photographs, and in-line allergen declaration.
None demands investment and all three can be measured within eleven weeks. Watch, without committing budget, AI-generated menu imagery and per-guest dynamic menu personalisation: promising, still fragile. Diego F. Parra has spent twenty years at that crossing between kitchen and cash register, and at Masterestaurant the sequence is not negotiable: cost per portion first, then decide which family grows, and only then shoot. Consider what happens if tomorrow you photograph your eight best sellers without reviewing costs: volume climbs on the thinnest dishes, overall food cost rises two or three points, and you celebrate a prettier menu while the till tightens. That mistake gets paid for over months. Shooting every dish is the fashion that has destroyed the most margin, and I got this wrong for years.
The overrated trend: photographing everything
I used to defend the fully illustrated menu because it worked in delivery, where 37 % of adults order weekly, according to UpMenu, and I carried that logic into the dining room without thinking twice. In the dining room the effect inverts: when everything carries a photo, no photo guides anyone, the guest retreats to the familiar dish, and the mix flattens precisely where you needed movement. Seventy-four items with images on all of them reads like a spare-parts catalogue. Copy, by contrast, works on elasticity in a way price never manages: shaving 2,000 pesos moves little, whereas «aged 28 days» changes the entire comparison, because the guest stops measuring that dish against the neighbouring table. Keep two photographs and write better. A photo does not sell the dish, it sells the CATEGORY. Put an image of the sirloin on the page and beef sales rise across the board, not just that sirloin.
Where the real difference sits?
So the right question is never «which dish do I shoot» but «which family do I want to grow», and contribution margin in currency answers it, never an isolated food cost percentage.
Description works on demand elasticity in a way price never manages. Cutting a dish by 2,000 pesos barely registers; adding «28-day aged, seared over orange-wood embers» registers hard, and it also moves the comparison: the guest stops measuring that plate against the neighbouring table and starts measuring it against itself. I got this wrong for years. I argued for photographing everything, because in delivery it worked, and I carried that conclusion into the dining room without measuring it. In the room, every photo past the second one on a page pulls the average check down, since the eye anchors on the image and skips the expensive dishes that lack one. There is a genuine tension between pricing psychology and menu engineering that almost nobody resolves: the first says «drop the currency symbol and never align prices», the second says «lead with what earns most».
Where the real difference sits — in practice?
They collide when the highest-margin dish is also the priciest. The bridge is the description: when the prose carries the value, a high price stops reading as expensive and both disciplines finally pull the same way.
Sales mix is the only judge. You can own the prettiest menu in town and still bleed money if 40 % of your units carry a 71 % food cost. Masterestaurant measures menu photography and dish descriptions by their effect on weighted margin at eight weeks, not by what the designer thinks or how the Instagram post performed.
Criterion-by-criterion comparison
What the old menu didBefore
- Photos placed by the owner's affection, never by contribution margin
- Three- or four-word descriptions that justify no price at all
- 74 references, twenty-three of them below 4 units a week
- High food cost dishes owned the dominant visual position
- Prices ending in 000, aligned in a right-hand column
- No standard recipe behind it: portions changed with the shift
What the reworked menu doesMasterestaurant
- One photo per section, always on the highest-margin dish there
- 18 to 25 words covering origin, cut, technique and texture
- 38 references, each with a signed cost per portion sheet
- The four stars sit in the first third of every block
- Price without currency symbol, folded into the description
- Standard recipe closed: the portion no longer depends on the cook
Side-by-side comparison
| Before (menu without criteria) | After (engineered menu) | |
|---|---|---|
| Menu references | ✕74 dishes, 21 with photos | ✓38 dishes, 5 with photos |
| Weighted contribution margin | ✕62 % on food sales | ✓68 % on food sales |
| Actual average food cost | ✕34.1 % (above the 32 % ceiling) | ✓29.6 % with the same suppliers |
| Average food check | ✕COP 41,200 per guest | ✓COP 46,900 per guest (+13.8 %) |
| Words per description | ✕3.8 words on average | ✓22 words with origin and texture |
| Star dishes' weight in mix | ✕18.3 % of units sold | ✓27.7 % of units sold |
| Decision time at the table | ✕6 min 40 s average | ✓4 min 05 s average |
| Cost of the change | ✕COP 0, which was precisely the problem | ✓COP 3.2 M across shoot and reprint |
The numbers behind this
“Eleven years with the same menu and I was convinced my problem was the price of beef. When Diego costed it dish by dish, my sirloin left 19,400 pesos and my pasta left 26,800 — and the big cover photo was the sirloin. We pulled sixteen photos, kept five, rewrote the descriptions with origin and cut, and in eleven weeks weighted margin went from 62 % to 68 % without touching a single price. Average check rose 13.8 % and food cost fell from 34.1 % to 29.6 %.”
How to run the change in under 90 days
Open the POS and pull units sold per reference over the last 90 days. Next to it, put the real cost per portion of each dish, with the standard recipe closed and waste included. What matters is not the food cost percentage but contribution margin IN CURRENCY per unit sold. Sort by that column. The top six are your candidates for a photo and prime position; the bottom twelve, if they also sell poorly, are candidates to leave. Without this table, everything after it is decoration.
Do not rewrite the whole menu, that is wasted labour. Take the six highest in margin and the six highest in volume, and write 18 to 25 words each covering ingredient origin, cut or variety, cooking technique and one texture. Skip empty adjectives like «delicious» or «exquisite», which inform nothing and cost credibility. The test is simple: if the description works equally well for another restaurant's plate, it does not work for yours.
One photo per menu section, always on the highest-margin dish in that section. Side natural light, the plate built with the REAL portion that leaves the pass — if the photo lies, the complaint reaches the table and the replacement cost lands on you — and a background from your own room. Budget between 2 and 4 million pesos for a session with a food photographer; doing it on a phone costs more the moment the file cannot survive print or a large screen.
Publish the new menu and leave it alone for eight weeks. At week eight, compare the unit weight of your four star dishes against the baseline, plus weighted margin for the period. If the stars gained less than 5 points of mix, the problem is position on the page rather than the copy: move those dishes into the first third of the block and measure again for four weeks. One correction, not five.
And with AI?
Optimize menu engineering, descriptions and the photos that sell most. Diego F. Parra is an expert in AI applied to restaurants.
Free tools to apply this now
What to lean on
None of these tools writes descriptions for you; they exist so the call on what to photograph and what to describe comes from the number instead of from taste.
Questions I get every time
Do menu photos raise or lower the average check?
Do menu photos raise or lower the average check?
It depends on how many. One photo per section raises the average check because it steers attention to the highest-margin dish; from the third photo on a page the effect flips and the check falls, since guests stop reading descriptions of the expensive dishes without an image and simply order what they see.
How many words should a dish description have?
How many words should a dish description have?
Between 18 and 25 words for the dishes you want to push, and 6 to 10 for the rest. They should carry ingredient origin, cut or variety, technique and texture. Three-word descriptions justify no price, and anything past 35 words nobody finishes reading at a table with an empty stomach.
Does reworking the menu let me raise prices without losing guests?
Does reworking the menu let me raise prices without losing guests?
Yes, though not the way you picture it. A worked description lowers that dish's demand elasticity: with perceived value loaded, a 5 % to 8 % price adjustment passes without a drop in units. Skip that groundwork and the same increase costs between 4 % and 9 % of volume over the first six weeks.
What does it cost and how fast does it pay back?
What does it cost and how fast does it pay back?
A five-dish session with a professional photographer plus the reprint runs 3 to 5 million pesos in Colombia. With a 5-point mix shift toward high-margin dishes, a 120-seat venue recovers that between weeks four and seven. If it has not paid back in three months, the menu was never the problem.
Sector data 2026 (official sources)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Benchmark 2026 | Source |
|---|---|---|
| Usuarios de fármacos GLP-1 que comen fuera con menos frecuencia (EE. UU.) | 54% de los usuarios | Encuesta a 1.000 usuarios GLP-1 vía Fortune — 2025 |
| Usuarios de GLP-1 que consumen menos snacks (EE. UU.) | ≈70% de quienes reportan menos calorías | EY-Parthenon — encuesta 2025 |
| Reducción del gasto de hogares con usuarios de GLP-1 (EE. UU.) | -10% en un año (100 categorías) | Numerator — 2025 |
| Comensales jóvenes que comparten un plato fuerte con más frecuencia (EE. UU.) | 42% de los más jóvenes | Acosta Group — 2025 |
| Consumidores que tomaron postre en el último día (EE. UU.) | 53% de los consumidores | Technomic — Dessert Consumer Trend Report |
| Operadores que dicen que los postres impulsan la utilidad (EE. UU.) | 60% de los operadores | Technomic — Dessert Consumer Trend Report |
Related content
Grow your restaurant with the Masterestaurant method
Applied in +8.400 restaurants across 43 countries.
