Cloud Kitchen Marketing Agency

Win the aggregators (Zomato & Swiggy)

Most orders start on the aggregator. We optimise your listing, menu, imagery, ratings and ad placements to lift visibility and conversion, and read the unit economics so you promote the dishes and combos that actually make money, not just move volume.

Build own-channel orders that dodge commission

Every order you take on WhatsApp or your own site saves the aggregator cut. We build that habit, QR inserts, first-party offers, and automated re-order nudges, so a growing share of demand comes back to you directly. In India, WhatsApp is the channel that works.

Multi-brand & repeat-order strategy

Cloud kitchens often run several virtual brands from one kitchen. We plan the brand mix, cross-promote sensibly, and run repeat-order automation on our restaurant automation stack, part of the wider restaurant marketing engine.

What a cloud kitchen engagement actually covers

Most operators come to us after six months of running listings on Swiggy and Zomato with nobody actually owning the numbers. The kitchen is running. Orders arrive. Somebody is discounting to keep the order count from falling, and nobody in the building can tell you what a ₹400 order leaves behind after commission, discount share, packaging and rider adjustments. That is the first workstream, and it is not a marketing workstream at all, it is a spreadsheet built from your own aggregator payout reports so that every decision after it has a floor to stand on.

Second workstream is the listing itself. Photography that matches what actually leaves your kitchen, dish names people search for rather than the clever ones the chef likes, category placement, description copy, and the tagging that decides which of your items show up when somebody in a two kilometre radius types ‘biryani’ at 8:40pm. Listing rank inside an aggregator app is not a black box you cannot touch. It responds to acceptance rate, preparation time, cancellation rate, rating, and how aggressively you are participating in the platform’s own promotional inventory. Some of those levers sit with your kitchen manager, not with us, and we will say so plainly.

Third is menu engineering. Which items carry margin, which items are there only to get you into a search result, which combos raise average order value without raising your packaging cost, and which items you should quietly kill because they are slow to cook and drag your preparation time up across every other order in the queue. Fourth is discount strategy, meaning the split between what the platform funds and what you fund, and at what point a funded offer stops buying incremental customers and starts subsidising people who were going to order anyway.

Fifth, and this is the one operators put off longest, is building an ordering channel you own. A WhatsApp reorder flow, a simple site, a printed insert in the bag with a code on it. It will never replace the aggregators. It does not need to. It needs to take the repeat orders, the ones you already paid to acquire, and move them somewhere the commission does not apply.

How the number is arrived at, and what moves it

Pricing for this category is driven by four things: how many kitchens, how many brands running out of those kitchens, how big each menu is, and how contested your delivery radius is. A single kitchen with one brand and a thirty-item menu in a suburb of Gurugram is a fundamentally different amount of work from three kitchens running four virtual brands each across Delhi NCR, where every brand needs its own listing hygiene, its own photography, its own offer calendar and its own read on which competitor just went aggressive on discounting.

Menu size matters more than people expect. Every item is a photograph, a description, a category decision, a price point that has to survive the commission maths, and a candidate for deletion. A ninety-item menu is not three times the work of a thirty-item menu, it is worse, because ninety items means the long tail is dragging your kitchen speed down and someone has to sit with the operations lead and argue about it dish by dish.

Competition moves the number because a contested radius means the offer calendar has to be watched weekly rather than monthly, and because paid visibility inside the aggregator app costs more when four other brands want the same slot. Ad spend on the platforms is separate from the fee, always. We do not mark it up and we do not take a cut of it, because an agency that earns more when you spend more has a reason to tell you to spend more.

What does not move the number: promises about order volume. Anyone quoting you on the basis of a guaranteed order count is quoting on something they do not control, and the way that engagement ends is with heavy discounting to hit the number, which you fund.

One more thing that changes the quote in your favour. If your kitchen already produces clean daily data, meaning item-level sales, wastage and preparation times that somebody actually records rather than estimates, a chunk of the first month disappears from the brief. Most kitchens do not have this. Building it is worth doing whether or not you hire anybody, because every decision described on this page depends on it, and a decision made on a guessed food cost is a guess with a spreadsheet wrapped around it.

Working the arithmetic on one order

Take a ₹400 order. Say the aggregator commission is a fifth of the order value, so ₹80 leaves immediately. Say you are running an offer where the customer sees ₹60 off and the platform funds half of it, so another ₹30 comes out of your side. Packaging is ₹18 for the container, the bag and the seal. Payment gateway charges apply on the gross. You are now looking at roughly ₹270 of realised revenue against an order the customer thinks was ₹400, and food cost has not been touched yet.

Now subtract food cost. If your raw material on that order is ₹120, contribution is around ₹150 before any of your fixed costs, meaning rent on the kitchen, the two cooks, the packer, electricity and the gas. Divide those monthly fixed costs by ₹150 and you have your break-even order count for the month. That single number is more useful than every dashboard the platforms will show you. Run it for yourself tonight with your own payout report, because your commission slab, your offer split and your packaging cost will not match the ones in this example, and the conclusion changes when they do.

Here is where it gets interesting. Push average order value from ₹400 to ₹520 with a combo that adds ₹40 of food cost and nothing to packaging, and contribution per order rises much faster than order value does, because commission scales with the ticket but your packaging and your rider adjustment do not scale the same way. That is the whole argument for menu engineering over discount depth. One raises the base. The other rents volume.

Run the same order through a direct channel. No commission, no funded discount, delivery arranged through a third-party rider aggregator at a flat per-order cost. Contribution roughly doubles. This is why the direct channel is worth building even at small volume, and why the sensible target is not replacing the platforms but shifting a portion of your repeat base off them.

Five questions to ask before you sign anything

Ask what happens to your contribution margin per order if the plan works. If the answer is about order growth and never touches margin, the plan is a discounting plan wearing a marketing costume. Volume bought below contribution is a machine for converting your working capital into someone else’s dinner, and it looks like success on every chart until the payout hits your bank.

Ask who owns the aggregator dashboard logins. The answer should be you, permanently, with the agency holding access rather than custody. An agency that resists this is protecting its position, not your account. Ask the same about your photography files, your menu copy and your customer contact data from any direct channel. You paid for all of it. You should be able to walk with all of it.

Ask what they will not do. A vague answer here is the tell. Somebody who has actually run these accounts has a list ready: they will not run offers below a contribution floor you set together, they will not add menu items that push preparation time past a threshold, they will not touch pricing without the operations lead in the room. Somebody who says they can do everything has thought about the pitch and not about the kitchen.

Ask how many virtual brands they think your kitchen should run, and why. The honest answer depends on your equipment overlap, your peak-hour capacity and whether you have prep space, and the honest version of that answer often is fewer than you were hoping. Four brands running from a kitchen that clogs at 8pm on a Saturday means four listings with rising preparation times and four ratings sliding at once.

The mistakes that cost the most

Discounting as a default setting. Not as a launch tactic, not as a targeted push into a slow slot, but as a permanent condition because switching it off makes the order graph dip and nobody wants to explain the dip. The dip is information. It tells you what share of your demand was ever really yours.

Ignoring preparation time. It is the least glamorous number on the dashboard and it quietly governs everything. Slow kitchens get downranked, get more cancellations, get riders waiting and marking delays, and then get rated on food that sat in a bag for eleven minutes. We have seen operators spend on visibility while their preparation time was the actual reason their listing was sinking, which is paying to be seen more often in a worse light.

Treating ratings as a vanity metric. Inside an aggregator app the rating is a distribution input, not a report card, and the difference between a four-point-one and a four-point-four is not pride, it is placement. The fastest way to move it is not asking for reviews, it is fixing the two dishes that generate the complaints and packaging the gravy separately so it does not arrive on the lid.

And launching a second virtual brand to fix a first brand that is not working. The second brand inherits the same kitchen, the same speed, the same packer and the same cook at peak hour. It doubles the listing maintenance and halves the attention. Fix the first one or shut it.

The costliest mistake of all is never separating the two questions that look like one. The first is whether people want your food. The second is whether your unit economics survive the way you are currently selling it. A kitchen can pass the first test and fail the second badly, and it will keep failing quietly for a year because the order graph looks healthy, right up until the working capital runs out and someone finally opens the payout statements properly.

What the first ninety days look like

Weeks one and two are diagnostic and dull. We pull twelve months of payout reports if they exist, rebuild the contribution maths per item, audit both listings line by line, photograph the menu properly, and sit in the kitchen during a Friday dinner rush to watch where the queue actually breaks. Nothing gets launched in this period. Operators find that frustrating and it is still the right sequence, because changing the offer calendar before you know your contribution floor is guessing with your own money.

Weeks three to six are listing surgery and menu cuts. New photography goes up, item names change to match search behaviour, the slowest and thinnest-margin items come off, combos go on, and the offer calendar gets rebuilt around a floor rather than around whatever the platform account manager suggested on a call. Expect order count to wobble here. That is the point.

Weeks seven to twelve are where the direct channel gets built and the paid visibility inside the apps gets tested properly, one variable at a time, with the contribution maths running underneath every test. By day ninety you should be able to answer three questions without opening a dashboard: what does an average order leave you, what share of your orders are repeat, and what would happen to profit if you switched every funded offer off tomorrow. Most operators cannot answer those on day one. That is usually the real problem, and it is fixable.