Delivery Algorithm Optimization: The Mistake That Sinks Your Ranking vs the Right Method

What delivery algorithm optimization means?

Optimizing the delivery algorithm means actively managing the variables Rappi, Uber Eats and DiDi Food use to decide your rank when a customer searches within three kilometers:

acceptance, cancellation, menu photos, dynamic pricing. No platform reveals its full formula, but the model weighs between 9 and 12 measurable variables, and here's the upside: most of them cost nothing to fix. Across the 140-plus ghost kitchens I've audited at Masterestaurant, four operational factors carry 60% of that weight: accepting in under 45 seconds, keeping your promised prep time, holding cancellation below 2%, and sustaining the rating from your last 100 orders. The other 40% rides on menu photography, how often you update prices, and how many orders you complete without a refund. The algorithm doesn't care about your concept or your cooking: it tracks operational behavior, in real time, every single day. Acceptance time carries more weight than most owners realize: confirm the order in under 45 seconds from the moment it lands.

The four variables the algorithm actually measures

Promising 15 minutes of prep and delivering in 28 hurts you more than promising 22 and delivering in 21; that's the trap, because the algorithm doesn't forgive broken promises even when the absolute time is short. Cancellations, all of them, including the ones the customer starts, need to stay under 2%. Cross into 6% and you lose up to 30% of organic visibility, based on the 140 restaurants I audited between 2024 and 2025. And the rating that counts is from your last 100 orders, not your lifetime average: ten straight five-star orders can undo months of mediocre reviews within a few weeks. Spending on advertising should raise sales. Layered on a kitchen that misses its own timers, it does the opposite: it speeds up the fall. That's the paradox I run into again and again in consulting, and here's how it resolves: a restaurant spending $800,000 COP a month on Rappi ads, with 28 real minutes of prep against the 15 promised on its profile, isn't buying visibility.

The structural mistake: paying for ads on a broken operation

It's buying a more visible penalty. Customers feel waits of 35 to 40 minutes at peak hours, right when the average ticket is highest, and they cancel. Cancellations climb to 6-8%, and the algorithm buries the restaurant deeper than before it ever ran an ad. Sixty-five percent of organic positioning comes from free metrics, not paid budget. At Masterestaurant we only turn on advertising once the operation already meets the four baseline variables, with a hard ceiling: never above 3% of delivery sales. DiDi Food still rewards aggressive pricing above almost anything else; Rappi and Uber Eats don't. That's the first lesson of 2026: each platform punishes and rewards different things, and treating them the same is an optimization error on its own. Cut your average ticket 8% and DiDi Food can move you up 2-3 positions in under 30 days. On Rappi and Uber Eats, what matters instead is acceptance speed and rating consistency over the last 100 orders, not the full historical record.

How each platform diverged in 2026: Rappi, Uber Eats, and DiDi Food?

I've seen a restaurant rise 4 positions on Uber Eats in three weeks just by cutting acceptance time from 95 to 38 seconds, without touching a price.

The same business, same effort, gained only 1 position on DiDi Food, and only moved once it adjusted prices by that 8%. Each algorithm needs its own 30-day plan with separate metrics: what works on one platform can do nothing on the other. Almost no owner takes menu photos seriously, and that's exactly where they lose points from the secondary 40% of the ranking. Platforms measure CTR, clicks over impressions, for every item in the search carousel. An item with a professional photo pulls between 25% and 40% more clicks than one with no photo or a blurry image. More clicks bring more orders, and more orders feed a positive signal the algorithm reads and rewards. At Masterestaurant I measured cases where updating 8 product photos on Rappi lifted average menu CTR 31% in 14 days, without changing price or prep time.

Menu photography as a ranking signal, not just conversion

Frequency counts too: a menu that goes 60 days without a price or photo change loses algorithmic relevance against competitors updating weekly. I recommend auditing and rotating at least 3 menu items a month, with a fresh photo, as the bare minimum for ranking upkeep. Before spending a single peso on in-app ads, measure four figures from your last 30 days: average acceptance time in seconds, the share of orders delivered within the promised window, total cancellation rate, and average rating over the last 100 orders. If acceptance runs past 60 seconds, if you miss more than 20% of your promised times, if cancellations top 2%, or if your rating drops below 4.5 out of 5, the algorithm is already penalizing you, and no ad budget reverses that. The MASTERESTAURANT method gets the operation to those thresholds first, holds them for 21 straight days, and only then turns on a campaign capped at 3% of delivery sales.

How to calculate your ranking score before investing in advertising?

Among the 140 restaurants we audited, the ones that followed that order grew organic sales by an average of 22% before spending a single peso on paid ads.

Few owners check, day to day, the share of orders that close clean, no refund, no dispute, and it's exactly the variable they ignore. Platforms log every return request: incomplete order, wrong item, excessive wait. Cross 4% of orders refunded over the last 90 days and you get an automatic ranking penalty, no matter what your visible star rating says. I've audited kitchens where 70% of disputes traced back to the same 3 items, almost always the most complex on the menu; simplifying or cutting them dropped disputes to under 1% within 45 days. Accumulated volume also works as a trust anchor: a restaurant with 5,000 clean orders competes with a structural edge over a newer one running just 200, even if both carry the same rating today.

30-day plan to climb in ranking without increasing your ad budget

The 30-day plan I run in consulting follows this order. Week 1: turn on order notifications on your fastest available device and set the acceptance protocol under 40 seconds; if you're currently past 90, this single change can move you 2 to 4 positions up on Rappi and Uber Eats. Week 2: audit the menu and pull, even temporarily, any item whose real prep time beats its promise by more than 5 minutes, then update the posted time so it matches your kitchen's reality. Week 3: refresh photos on the 5 items with the highest CTR potential. Week 4: review disputes and refunds. Any item with more than 3 disputes that month gets suspended or reworked, no exceptions. At the close of the 30 days, compare ranking before and after, measure the delta in organic orders, and only then decide whether to turn on ads with a 3% budget over delivery sales.

The 4 Differences That Weigh Most in the Algorithm's Ranking

A restaurant that accepts in 40 seconds outranks competitors running a 15% higher average ticket: the platform prioritizes zero friction for the customer over almost everything else. Nothing punishes harder than cancellations. Going from 2% to 6% cuts organic visibility by up to 30%, based on the 140 restaurants I audited over the last 18 months. Keeping your word on prep time matters more than the raw number. Promise 20 minutes and deliver in 19, and you'll outrank a restaurant promising 12 and delivering in 22. Ten straight five-star orders can undo, in three weeks, a prior slide from 4.3 to 4.7, because the algorithm weighs the rating from your last 100 orders, not your lifetime history. Tuning by platform matters: the same fix that gains 4 positions on Uber Eats might move you only 1 on DiDi Food without a parallel price adjustment of around 8%.