Reels vs TikTok for restaurants: 41% lower guest acquisition cost and +2.3 EBITDA points after rebuilding the funnel with the Masterestaurant Demand Radar

The case file: 32 tables, 74,000 followers and an empty Wednesday

The steakhouse billed between 500,000 and 1 million USD a year and still could not fill the dining room from Tuesday to Thursday, which is exactly the symptom that sends an owner looking for help. The starting numbers: 96 seats, 21 employees across kitchen and floor, an average check of 28.40 USD, seven years of operation in a Colombian city of 900,000 people, and 68% of sales concentrated in the dining room, with the rest split between in-house delivery and aggregators. On social, 74,000 followers across Instagram and TikTok, daily posting, 800 USD a month between a freelance editor, paid media and props. Not one line of the P&L explained what that money bought. The owner put it in the sentence I have heard in forty-three countries with different accents: we post every day and nothing happens. Something was happening, of course. It was happening in the expense account.

The market squeeze: prices up, guest acquisition cost through the roof

No Colombian restaurant was competing in 2025 with the funnel it had in 2018, and that single fact changes the whole conversation. According to ACODRES (2025), restaurants in the country raised menu prices 9.8% starting in February of that year just to sustain 98,000 jobs, which means margin was defended with the menu, not with volume. At the same time, according to Marqii (2025), guest acquisition cost grew 222% over the eight years through 2025. Do the subtraction yourself: the check rises less than ten points while the cost of bringing in whoever pays it triples. Margin is not being eaten by the kitchen, it is being eaten by the funnel, and while the owner argued food cost with his chef, the real leak sat two meters away, inside a phone, with no denominator measuring it. That was the entry diagnosis. TikTok does not sell reservations, it sells attention, and its honest unit of measure is cost per qualified profile visit.

What TikTok actually buys: cheap, messy attention?

The moment we measured it that way, the number stopped lying: 0.11 USD per qualified visit on TikTok against 0.38 USD for the same movement on Reels (per the restaurant's own operating data, August 2025).

Three and a half times cheaper to pull a stranger into the restaurant's orbit. What TikTok never did, not once across fourteen weeks, was produce a clean pattern of same-day attributable bookings. Its audience arrived from other cities, sometimes from another country, with roughly the intent of window shopping. There sits the tension almost every owner resolves badly: the platform that brings the most cheap traffic seats the fewest people, and that does not make it useless, it makes it a discovery channel with a different job assigned. Reels charges more because it sells something else: trust within a fifteen-minute drive. The Instagram audience already knew who the restaurant was, had walked past the door, had friends tagged in the photos, and the only missing piece was a concrete reason to choose a Tuesday.

What Reels buys: expensive trust, with a postal code?

So the creative changed at the root. Trending audio was dropped in favor of a dish demonstration with the price visible on screen, the service hour stated, and one operational call.

At 0.38 USD per qualified visit, Reels stayed the expensive channel per click and the cheap one per seated guest, because 68% of sales lived in the dining room and dining rooms fill with neighbors, not with spectators. National reach does not pay payroll in a mid-sized city; neighborhood reach does. That nuance is worth the entire ad budget. The problem was never the platform, it was the bookkeeping, and that is why no comparison between channels meant anything: the denominator was missing. Those 800 USD a month came out of petty cash, with no line of its own in the P&L, no cost per guest attributed, no split between production spend (the editor) and distribution spend (the paid media).

The real error was accounting: marketing paid out of petty cash

With that disorder, debating Reels against TikTok is debating the color of a wall that is falling down. According to Toast (2026), 33% of industry professionals cite attracting and retaining guests as their top challenge, and I will risk an uncomfortable judgment: a good share of that 33% does not have an acquisition problem, it has a measurement problem, because nobody improves what they record as generic expense. I got this wrong for years, recommending more content before demanding an accounting line. Denominator first. Creativity after. The tool we applied was the Channel-to-Cash Matrix from the Masterestaurant method, which forces every platform to carry ONE measurable job and a single cash KPI, never several. TikTok was assigned discovery, and its only metric became cost per qualified profile visit, capped at 0.15 USD. Reels was assigned local conversion, measured by attributed reservations and by thirty-day repeat visits, capped at 6.00 USD per reservation.

The Masterestaurant intervention: the Channel-to-Cash Matrix

A marketing line was opened in the P&L, split between production and distribution, and one hard rule was agreed, the same rule Diego F. Parra imposes in every audit: nothing gets published unless somebody can say which table it was trying to fill. Paid media was redistributed 70% to Reels on Tuesdays, Wednesdays and Thursdays, and 30% to TikTok across the month, without touching the 800 USD total. With the same 800 USD a month, Tuesday-to-Thursday occupancy climbed from 41% to 63% on average across fourteen weeks (per the restaurant's own booking records, close of November 2025). Reservations attributed to Reels went from a figure nobody counted to 118 a month at an average cost of 4.74 USD, well under the 6.00 USD cap and against a 28.40 USD check. TikTok kept its role: 0.09 USD per qualified visit, better than the 0.11 USD baseline, plus follower growth that nobody confuses with sales anymore.

The result at fourteen weeks: same money, different cash

The freelance editor kept the same fee while shooting half as many pieces, longer ones, with the price on screen. What changed was neither the budget nor the talent. What changed was the question each video was answering. Copy the method, not the figures, because your first concrete step depends on what you bill per year. Under 500,000 USD: this week, open a marketing line in your P&L and put everything in it, with no new paid media, until you know your cost per guest acquired. From 500,000 to 1 million, the band in this case: assign ONE job per platform and a cost-per-reservation cap no higher than 20% of your average check. Above 1 million: separate production from distribution in the books and demand attribution by location, never aggregated. Above 5 million: audit whether your national reach is paying for seats in markets where you actually have tables.

Transferable lessons by annual revenue band

And above 10 million, the media-chef archetype running a multi-unit group: the face buys reach for free, so paid media should go almost entirely to the markets with soft midweek occupancy. This result does not travel to every operation, and saying so is part of the craft. First limit: a 100% delivery business or dark kitchen has no dining room to fill on Tuesdays, so the Reels advantage inside a fifteen-minute radius evaporates and TikTok's cheap discovery weighs far more in the mix. Second limit: a low check, say under 8 USD, cannot carry a 4.74 USD cost per reservation anywhere close, because the game there is frequency and loyalty, not booking; according to Paytronix (2024), the best QSR operators enroll around 110 new loyalty members per store per month, and that is the right engine for that profile. Third limit: a seasonal tourist city breaks the local logic entirely, because its audience genuinely comes from somewhere else.

Limits of this case: where I would NOT expect these numbers

Start with your denominator, not with your camera. TikTok buys cheap, messy ATTENTION, and its honest unit of measure is cost per profile visit, not a reservation. Once this case measured it that way, the number stopped lying: 0.11 USD per qualified visit against the 0.38 USD the same move cost on Reels. Reels buys expensive, localized TRUST. The audience already knows who you are, lives fifteen minutes away and needs a concrete nudge — a dish, a time, a reason. That is why the Reels creative stopped being a trend and became a dish demonstration with the price on screen. The underlying error was accounting, not platform choice. As long as restaurant marketing gets paid out of petty cash and judged by views, no comparison between channels means anything, because the denominator is missing. Online reputation works as a cross amplifier and almost nobody builds it deliberately: a diner who arrives through TikTok and leaves a photo review feeds next week's Reels.

What actually changes once you split the two funnels?

That loop cut paid spend without cutting reach. Delivery conversion behaves nothing like the dining room. TikTok moved discounted aggregator orders; Reels moved in-house channel orders carrying 22 more margin points per ticket.

Same footage, two opposite economics.