Who should you trust with GrabFood and GoFood: an in-house manager, a freelancer, a digital agency or a specialist agency?
The difference is the result, not the price
Look at what the specialists you will be offered instead of an agency actually promise. In their own public profiles the stated achievements are tens of percent: "+30% in sales", "+109% in menu views", "+50% in profile visits". These are honest numbers, and they are the ceiling of what one person with one account delivers.
Our published cases come from GrabMerchant and GoBiz dashboards, with screenshots:
A caveat we make ourselves: these are published cases, meaning our best, not the median. The median is published separately and just as openly — in the benchmark across 96 restaurants. Promising everyone 21x would be a lie. But note the order of magnitude: in the hiring market these numbers do not appear even as promises.
Why you cannot simply hire this person
The profession does not exist. Grab and Gojek do not train specialists, do not certify them and do not publish how in-app ranking works. There is exactly one way to learn it: across a volume of accounts, over a long stretch of time.
A simple example of something you cannot derive from one restaurant. Grab and Gojek are not ad networks. An ad network earns on impressions and auctions you a position; Grab and Gojek earn commission on orders, and their asset is their own audience. So ranking follows how well a restaurant monetises that audience: a bid amplifies a position, it does not create one. Someone who treats Grab as an ad network will raise the budget and get more impressions without orders — faster.
A second example: norms a single account simply does not have. Ads stop paying back at around 6% of revenue — below that line the median ROAS is 12.1x, above it 8.6x. And 95% of all revenue losses in our sample are not a closed restaurant and not cancellations, but switched-off menu items. Both numbers come from 96 restaurants and 270,568 orders; from one dashboard they are invisible. Benchmark 2026, the full method.
The comparison that matters
| In-house manager | Agency Delivery Booster | |
|---|---|---|
| What you are buying | One person’s hours. They accumulate the knowledge on your restaurant, at your expense. | A method run across 96 restaurants and published in full. |
| What the numbers are compared against | Last month at the same restaurant. There is no other data and nowhere to get it. | Medians across 96 restaurants and 270,568 orders — published openly. |
| The result on offer | In the public profiles of such specialists: tens of percent — +30% to sales, +50% to profile views. | Multiples: x2.6 to x21 in revenue, with numbers from the dashboards. |
| When the work starts | After the search and three to six months of platform learning, funded by you. | In week one. There is nothing to learn — the method is already written. |
| Holiday, sickness, resignation | Nobody watches the stop-list or reviews on those days, and the knowledge leaves with the person. | A team, not a person: cover is internal, the method stays. |
| How payment works | Fixed. Paid in the month revenue drops, and while the person is still learning. | 10% of delivery revenue, no upfront: you pay more only when revenue has grown. |
Where you do need your own person
We say this plainly, because otherwise the page cannot be trusted. Operations is always your person: item availability, the stop-list, preparation time, the kitchen at peak. That is physically inside the restaurant and cannot be run from outside — and by our own data that is exactly where 95% of revenue losses sit. The second honest case: if delivery is not a growth channel for you but simply has to run without failures, you need someone on control, not an agency on growth.
The hybrid that works
The most common working arrangement in chains is to split by the nature of the work. Your person holds operations. We hold listing management: menu and menu SEO, ads and bidding, promo economics, reviews and appeals against unfair ones (roughly 80% of the appeals we file on Grab end with the review removed). The split follows the data exactly: the losses are operational, the growth is in listing management. Different hands, different skills.
And if not an agency — then who?
Before handing delivery to an agency you will almost certainly look at cheaper options. There are five, and all five are real: a marketplace freelancer, a digital agency, a paid-ads specialist, a former Grab employee, your own hire. Here is each one — what you are actually buying and when it is the right choice, because for some jobs the right choice is not us.
What the five have in common is not a lack of skill. The market sells the launch. Orders come from the running. Registering the restaurant, building the menu, shooting the photos, handing the logins to the owner — that is a normal service and it saves weeks at the start. But ranking, the stop-list, bidding, promo economics and reviews are what happens to the account every week after the launch. Who owns that, and what they compare their numbers against, is the whole question.
A marketplace freelancer
Look at what is actually for sale on Fastwork and Sribu under GrabFood: "Jasa Daftar GoFood/GrabFood" — registration, menu and photo setup, then handover of the account to the owner with training so they can carry on alone. That is one-off work with a defined end, and as one-off work it is honest.
The trouble starts where one-off work is used to cover a continuous job. And especially where it is used to cover the rating. This part cannot be short, because it is the one place in this whole comparison where the mistake costs neither money nor months, but the business.
On the same marketplaces, right next to "optimisation", reviews and ratings are openly for sale: from Rp 99,000, with a money-back guarantee and a note that they may not stick, since platform policy is outside the seller’s control. That is the cheap answer to the rating problem, and people buy it because it looks like a solution.
Now the part the seller will not tell you and the contractor most likely does not know himself — and that is not a figure of speech: neither the marketplace freelancer nor the self-described "marketplace specialist" usually has any idea how the platform handles ratings, or that an account gets closed for faking them.
Grab and Gojek watch review manipulation closely and systematically. To them a fake review is not a minor rule breach but damage to their own asset: they earn commission on orders, orders follow trust in the ratings, and someone who lies in the ratings is taking money out of their pocket. The platform holds data neither you nor your contractor has: which device, which account, in what order and at what speed the reviews arrive. From there, manipulation is as visible as a bonfire at night.
A restaurant owner came to us — not our client at the time; his account was being run by exactly this kind of outside "specialist". He came shouting, because he had received an email from Grab: a second and final warning for fake reviews. His reviews were being posted from one and the same account — the crudest scheme there is, and the platform spotted it instantly. "Second and final" means exactly what it says: he had missed the first one and there would be no third. The next step is not a ranking penalty and not the removal of reviews, but a closed merchant account — along with the order history, the rating, the audience built up over time and the position the restaurant had earned over months. If delivery was the main channel, along with the business.
He did not buy himself a ban. He bought a service for a couple of hundred thousand rupiah from someone who did not know this could happen. That person was cheap right up until that email.
The legal answer to the very same problem looks duller and works better: the appeal. Around 80% of the appeals we file on Grab end with the unfair review removed — on two conditions: file fast, and cite the facts of the specific order rather than general disagreement. The difference is simple. After an appeal you have one unfair review less. After buying reviews you have one warning more.
When it is the right choice: a one-off job with a defined end. Registration, menu photography, translating descriptions. Anything with a finish date rather than a weekly rhythm.
A digital agency
You recognise it by the phrasing: "we’ll try running Grab for you". "Try" is an honest word, and it is also the diagnosis: they have twenty lines on the price list, and delivery is one of them.
Their competence is real, it just sits elsewhere: websites, social media, photography, Meta ads, branding. Ranking inside a marketplace does not transfer from those disciplines — different mechanics, different norms, and nowhere to collect them: to see that ads stop paying back at around 6% of revenue you do not need twenty services, you need ninety-six restaurants.
When it is the right choice: when you need branding, a website, photography and social media. We do not do that and do not take it on.
A paid-ads specialist
The most understandable and the most expensive mistake, because it sounds reasonable: "ads are ads, I’ll set them up".
A paid-ads specialist optimises an ad network auction. There you buy an impression, and the craft is buying it cheaper and more precisely. Grab and Gojek do not earn on impressions — they take commission on orders, and their asset is their own audience. So ranking follows how well a restaurant monetises that audience: a bid amplifies a position, it does not create one. Someone with an ad-network reflex will raise the budget — and on our own chart will drive past 6% of revenue, where the median ROAS falls from 12.1x to 8.6x. He is not failing as a paid-ads specialist. He is solving the wrong problem.
When it is the right choice: when you are driving traffic to your own channel — website, WhatsApp, the dining room. That is his auction and his job.
A former Grab employee
The most convincing candidate in an interview, and here it pays to be precise.
There is indeed an official agent role inside the Grab ecosystem — Agent Reference Merchant: the agent recommends a restaurant, helps it through registration, and receives a bonus for the signup plus a commission on the merchant’s transactions. The role ends at signup: no menu, no bidding, no rating in it. So the official role the platform created and pays for is sales, not growth.
An employee saw the processes, the rates and the promo programmes from the inside, and that is genuinely valuable. But he saw them from the platform’s side, not from the merchant dashboard, where revenue is assembled out of the stop-list, listing conversion and bidding. Two different windows onto the same screen.
When it is the right choice: if he actually ran merchant dashboards rather than account-managed. Three questions settle it — they are in the FAQ below, and we published our norms precisely so they can be applied to anyone, us included.
An in-house manager
Covered separately above: the short answer is that the profession does not exist as a profession, there is nowhere to learn it, and the knowledge only accumulates across a volume of accounts. But operations is always your person — and by our own data that is where 95% of revenue losses sit.
When it is the right choice: operations — always. Item availability, the stop-list, preparation time, the kitchen at peak.
Who solves what — in one table
| Who | What you are actually buying | When it is the right choice |
|---|---|---|
| Marketplace freelancer | A one-off launch: registration, menu, photos | A one-off job with a defined end |
| Digital agency | Brand and outside channels; delivery is a line on the price list | Website, photography, social, Meta |
| Paid-ads specialist | Craft in an ad-network auction | Traffic to your own channel |
| Former Grab employee | The view from the platform’s side | If he actually ran merchant dashboards |
| In-house manager | One person’s hours, the learning funded by you | Operations — always |
| Delivery agency | A method across 96 restaurants and accountability for revenue | When delivery is a growth channel |
Not one line in this table says the person is bad. All six say what job they are able to do. The mistake is expensive not because the contractor is weak, but because in the three or four months it takes to test the hypothesis the account accumulates a history the algorithm then remembers.
Frequently asked
I’m told to find a freelancer on Fastwork — isn’t that cheaper?
Cheaper, yes, and for some jobs it is the right choice: registering the restaurant, building the menu, shooting the photos. That is mostly what is on sale there — a one-off launch with the account handed back to the owner. There is no ongoing running, and orders come from the running: the stop-list, bidding, promo economics and reviews happen every week, not once. But the main warning is not about money, it is about the rating. On the same marketplaces reviews and ratings are sold from Rp 99,000, and the buyer is usually not the owner but the hired contractor, who does not know how it ends. Grab and Gojek track manipulation systematically: they see the device, the account and the speed at which ratings appear. A restaurant came to us holding a second and final warning from Grab for fake reviews — they had been posted from one and the same account. The step after a final warning is the merchant account being closed, with the entire history. The legal route to the same goal is appeals: around 80% of ours on Grab end with the unfair review removed.
My digital agency says it can run Grab too. Should I agree?
Ask what they will compare your numbers against. Ranking inside a marketplace does not follow from experience in websites, social media and Meta — different mechanics, different norms, and norms only come from a volume of accounts. Easy to check: our medians are published — ROAS 10.4x in Bali and 22.8x in Phuket, ads at 5.6% of revenue, one negative review per 138 orders. If the answer to "what share should go to ads" is "the more the better", you are looking at an ad-network specialist, and Grab is not an ad network.
What about hiring a paid-ads specialist — they know how to run ads?
They do, but a different kind. An ad network sells an impression at auction, and the specialist’s job is to buy it cheaper. Grab and Gojek take commission on orders, their asset is their own audience, and ranking follows how well a restaurant monetises it. A bid amplifies a position, it does not create one. The practical consequence: a network specialist raises the budget, and payback breaks at around 6% of revenue — below that line the median ROAS is 12.1x, above it 8.6x. 42% of the restaurants in our fleet are already past it, so the mistake is a common one.
Is my own hire cheaper than paying you 10% of revenue?
The arithmetic is simple: an in-house hire is cheaper once delivery revenue exceeds their fully-loaded cost times ten — because our fee is exactly 10%. The problem is that there is nothing to put into the formula. A person you can actually hire on this market does not know this work: it is not taught, there is no course and no certification, and the knowledge only accumulates across a volume of accounts. So the question is not "is it cheaper" but "will it produce the same result". Judging by the public profiles of such specialists: they promise growth in tens of percent, our cases show growth in multiples.
Why can’t I just find a good specialist myself?
Because the profession does not exist as a profession. The platforms do not train specialists, do not certify them and do not publish how ranking works. Everything we know comes from the dashboards of a hundred-plus restaurants over three years — including things you cannot derive from a single account: that ads stop paying back at around 6% of revenue, or that 95% of all losses come from switched-off menu items rather than a closed restaurant. Someone with one restaurant will never see that, however hard they try.
How do I check a candidate who says they have run GrabFood?
Ask three things and check the answers against our published norms. One: what share of revenue should go to ads, and why — if the answer is "the more the better", they think Grab is an ad network, and it is not. Two: how many hours their menu items spent on the stop-list last month — if they do not know, they were not managing it. Three: what their ROAS was and what they compared it against. Our fleet medians are 10.4x in Bali and 22.8x in Phuket; we published the norms openly precisely so they can be applied to anyone, us included.
Will my restaurant grow like the ones in your cases?
Our best, and we say so plainly. Published cases are the ones where there was something to show and permission to show it. The typical numbers are published separately and just as openly: medians across 96 restaurants, including the boring ones. Promising everyone 21x would be a lie; the difference is that multiple-fold growth happens at all in our work and is backed by dashboard screenshots, while in the hiring market such numbers do not appear even as promises.
When do I actually need my own person instead?
Always — but for operations, not for revenue management. Item availability, the stop-list, preparation time, the kitchen at peak: that is physically inside the restaurant and cannot be run from outside. By our data that is exactly where 95% of revenue losses sit. The second case is when delivery is not a growth channel for you but simply has to work: then you need someone watching that nothing breaks, and you do not need an agency.
Can I keep my person and bring you in as well?
Yes, and in chains this is the main working arrangement. Your employee holds operations, we hold listing management: menu and menu SEO, ads and bidding, promo economics, reviews and appeals. The split follows the data exactly: the losses are operational, the growth is in listing management. Different skills, and they almost never sit in one person.
What we find on an account like yours
This is not about whether your person is good or bad. It is what only shows up across a volume of accounts and stays invisible from a single dashboard.
Our fleet medians for Bali: check Rp 250,000, ROAS 10.4x, ads at 5.6% of revenue, cancellations 0.35%, one negative review per 138 orders. An in-house manager can only compare your month with your previous month — there is no other data and nowhere to get it.
It sits at roughly 6% of revenue: below it the median ROAS is 12.1x, above it 8.6x. 42% of the restaurants in our fleet are already past it. You cannot derive that line from one account — there is no second point from which it becomes visible.
Around 25% of revenue goes past the restaurant, and 95% of that loss is switched-off menu items — not a closed restaurant (3%) and not cancellations (2%). The typical picture is 40–70 items off at once, with individual dishes stuck in the stop-list for over 2,000 hours.
Reviews are bimodal: 51% five-stars against 28% one-stars, with only 3% fours — the rating is made by the extremes. Around 80% of the appeals we file with Grab end with the review removed. Someone doing it for the first time knows neither the wording nor the deadlines.
Keep reading
Two ways to go from here
Both work. The first costs nothing and does not require us.
The method and the norms — open and free
The five stages we run on every account, published in full. Alongside them, market norms from 96 restaurants so you have something to compare your numbers against.
An audit of your listing
Paste your restaurant’s Grab link and the report comes back in a couple of minutes: menu and search, photo coverage, reviews, prices against the neighbours. Free, no strings; after that it is 10% of delivery revenue with no upfront payment.
Diagnose my listingor message us directly →