marketplace onboarding on the supply side, from download to first order
getting a driver to install your partner app is the easy part. getting them through documents, approval, training and a first delivery is where a delivery marketplace wins or loses its supply.
by shritupdated 9 min read

tl;dr (too long, didn't read)
on a delivery marketplace, the sellers are your drivers. supply onboarding is everything between "installed the partner app" and "completed a first order": documents, verification, training, kit, going online, and the first job itself.
new drivers quit at the same few spots: document upload, the approval wait, training, and the first time they go online and don't know what to press. a typical path is 12 to 17 steps and 72 to 96 hours, by one vendor's count.
count a driver only when they finish a first delivery, the way Shadowfax does. then fix the steps that are confusing, in the driver's own language, on the screen where they're stuck, and be honest about the ones no amount of explaining will fix.
the driver who almost made it
picture a new rider installing your partner app on a tuesday evening. sign-up takes two minutes with an OTP. so far so good.
then the app asks for a licence photo, an RC, a bank account and a selfie. one photo gets rejected and the message doesn't say why. the approval screen says the same thing for two days. there's a training video in English. there's a field asking for a code they were supposed to get from somebody. on thursday they open a different app, one their friend already uses.
in your dashboard they count as a signup. in your order volume they count as nothing. the Play Store is full of people stuck at exactly these points. a Snabbit review from march 2026 is someone "asking for onvording agent code". a Borzo review from september says the "same approval message appears again and again". a MagicFleet reviewer is stuck on "manual training order not reviewed yet". none of these people wanted to leave. they were waiting for the app to tell them what to do next.
why the supply side is the expensive side
most writing about marketplace seller onboarding is about shops listing products. in delivery and logistics the sellers are people with bikes, and the stakes are higher, because every order needs one of them to show up.
the big listed companies say this in their filings. Eternal, the company behind Zomato, tracks "delivery partner recruitment and onboarding cost" as its own line, and reported 5.52 lakh monthly active delivery partners in FY26. Swiggy's DRHP talks about "quicker onboarding of delivery partners" as a way to "reduce delivery partner attrition rates". when onboarding shows up in an annual report, it's a cost someone is being asked about.
the cost per driver is going up. one vendor, Caller Digital, puts onboarding a rider at ₹1,600 to 2,400 by mid 2026, up from ₹800 to 1,400 in 2024 (a vendor claim, also repeated by RichAutomate). the same vendors say half a new cohort stops logging in by somewhere between day 88 and day 110. and the pool keeps churning. NITI Aayog projects India's gig workforce growing from 7.7 million in 2020-21 to 23.5 million by 2029-30, while Mint reports typical tenure has fallen to three or four months. you're going to onboard a lot of drivers, over and over, so every leak in the funnel gets paid for again with each cohort.
where new drivers give up
almost every partner app runs the same sequence, whether it's food, quick commerce, parcels or trucks. the labels change. the shape barely does.
| stage | what the driver does | what usually goes wrong |
|---|---|---|
| 1. sign up | phone number, OTP, city, vehicle type | a referral or agent code they don't have |
| 2. documents | licence, RC, PAN, bank, selfie | rejected photos with no reason given |
| 3. verification | wait for approval | a status screen that never changes |
| 4. training | watch videos, pass a quiz, sometimes a test order | videos in the wrong language, a test nobody reviews |
| 5. kit and slots | collect a bag or tshirt, pick a slot, maybe pay a fee | fees and deposits that surprise them |
| 6. first order | go online, accept, pick up, deliver, collect OTP | not knowing which button does what, on a live order |
Caller Digital, which sells calling bots to these companies, counts 12 to 17 steps from download to first delivery, taking 72 to 96 hours in a typical case. that's a vendor's number, but it matches what you see if you install a few of these apps yourself. three days is a long time to hold the attention of someone who needs to start earning this week.
look at what the bad steps have in common. some ask the driver for something they don't have, like a code or a deposit. some leave them waiting with no idea how long. and some, the training and the first order, are just confusing, because the app explains itself in words the driver doesn't read comfortably. those three kinds of problem need different fixes.
count a driver when they deliver
"registered partners" is a flattering number. it goes up every time someone types a phone number. Shadowfax, in its UDRHP, only counts a partner once they've "completed at least one delivery". that's the right finish line for supply onboarding, and it's the one we'd use. we wrote more about it in activation rate for driver apps.
once that's the definition, the useful numbers are the conversions between stages. how many signups upload all documents, how many get approved, how many finish training, how many go online, how many complete a first order, and how long each of those takes. break it down by city and by language and one stage will stand out as the expensive one.
what's a language problem and what isn't
it's tempting to assume new drivers struggle because they aren't used to phones. the data says otherwise. a government study by VVGNLI across four metros found 92% of delivery workers had secondary to graduate level schooling, and IDinsight found 95% owned a smartphone before joining. these people use phones all day.
what they don't get is an app that speaks their language. in a 2023 Borzo company survey reported by CNBC TV18, 84% of couriers said they use a local language while delivering and 82% wanted a local language inside the app. a Zypp review from october 2026 is just "no option to choose telugu as language". most partner apps are written in English by people in an office, then handed to drivers who think in Hindi, Marathi or Telugu.
be honest about the rest, though. no explanation fixes a verification queue that takes four days, a joining fee the driver can't pay, a deposit they don't trust, a team lead who hasn't approved their slot, or a zone with no orders. if your funnel leaks at one of those, fix the operation first. voice helps with the steps where the driver could have done it themselves if somebody had shown them how.
what the bigger players did
the companies with the most drivers have tried several things, and their choices are worth a look. Shadowfax built an in-house tool called Frodo, described in its UDRHP. Porter added regional language audio to its app. a former Zomato product lead has talked about adding vernacular audio walkthroughs in six languages inside an offer card, piloting it in three cities for two weeks against the old flow, and hitting targets before a national rollout. that one was about incentives, and no exact lift was shared, but the pattern is clear. the big players moved explanation into the app, spoken, in the driver's language.
the other common answer is people on phones. one large platform used 300 to 500 outsourced callers to chase partners who "had not taken their first order", according to a vendor story from Bolna. another vendor, Hunar, claims one platform spent ₹120 to 130 crore a year on telecallers at 8 to 10% conversion. calling bots are cheaper, about ₹28 (Bolna, plus telephony) to ₹47 (Caller Digital) for a five minute call. but a call happens away from the screen. the driver has to listen, hang up, then find the button on their own.
by our estimate, a five minute guided script inside the app is about 4,000 characters, which costs roughly ₹4 to 12 per driver at Sarvam's text to speech prices. that's an estimate, not a quote, but it's a different order of cost, and it happens while the driver's thumb is on the right screen.
show them on the screen they're stuck on
that's the problem we built heytappr around. it's a voice guide that lives inside your Android partner app. when a new driver is stuck, it highlights where to tap and talks them through it in Hindi first, with regional languages after that, right on the screen they're on, until they finish a first order. every step is shown on screen as well as spoken, so it still works on mute in a noisy market.
it only speaks the guides your team wrote and approved, so it never invents an answer about payouts or penalties. your ops team edits guides from a dashboard and they go live without an app release. if a release moves a button, heytappr flags the broken step. and you get a funnel from signup to first job that shows the exact step where drivers quit. it ships as an Android SDK: a Gradle dependency and one line in Application.onCreate.
measure the gap, then close it
none of this is exotic. most of it comes down to asking for less, telling drivers what's happening while they wait, and being there in their language when they get lost. if you want something to hold your own flow up against, this is the checklist we'd use:
- sign-up needs a phone number and city. any agent or referral code is optional or explained.
- every rejected document says exactly what was wrong and how to retake it.
- the approval screen says what's being checked and roughly how long it takes.
- training is short, spoken in the driver's language, and doesn't loop.
- fees and deposits are shown in plain rupees before the driver commits.
- the first order has someone, or something, walking the driver through each tap.
- you count a driver as onboarded only after a completed first delivery.
- you track drop-off per stage, per city and per language, every week.
Kiki started a delivery business in a strange town with a borrowed broom and a spot at a bakery counter. your new drivers have it harder in some ways. the least the app can do is speak to them in a language they think in.
see it in your product
fifteen minutes: heytappr answering a real question by walking a user through a real interface, out loud.
questions people ask
What is supply-side onboarding on a delivery marketplace?
Taking a new driver from installing the partner app to completing a first order: sign-up, document upload, verification, training, kit and slots, going online, and the first delivery itself.
Where do new delivery drivers drop off during onboarding?
Most often at document upload, during the approval wait, in training, and on the first order when they go online and don't know which button to press. Some drop-off comes from fees, deposits or slow verification, which better explanations won't fix.
How long does driver onboarding take?
One vendor, Caller Digital, counts 12 to 17 steps and 72 to 96 hours from download to first delivery in a typical case. The goal is to shorten time to first order, not just time to registration.
When should a delivery app count a driver as onboarded?
After their first completed delivery. Shadowfax uses this definition in its UDRHP, counting a partner only once they have completed at least one delivery.
How can a delivery app onboard drivers without a big calling team?
Fix the operational blockers first, then guide drivers inside the app in their own language on the screen where they are stuck. HeyTappr does this for Android partner apps with spoken, on-screen steps your team approves.
go deeper
keep reading
- how to write guide steps a new driver can follow by ear
- a spotlight tour in jetpack compose, on a driver signup screen
- why your onboarding overlay broke on tuesday
- why new delivery drivers quit before their first order
- in-app guidance on android, for the apps drivers use
spotted something out of date? tell us and it gets fixed.