Faster than paper, or nobody uses it
Most Indian hotels still check guests in with a photocopier and a paper register. I joined DigiStay before it was incorporated and, with the founders, built an OCR check-in that front-desk staff learn in about 10 minutes. It scaled to 50+ hotels.
At a glance
Check-in meant photocopying an ID and copying it into a register by hand: 10–15 minutes a guest, for tired guests and for staff who aren't tech-savvy and change often.
Our real competitor was the paper register. If the digital flow wasn't faster than paper, no feature list would get it used.
OCR check-in: photograph the ID, type a phone number and room rate, and everything else fills itself. Then guided capture, once blurry photos broke it.
Under 2 minutes per check-in, learnable in ~10 minutes, live in 50+ hotels.
A photocopier and a register: the check-in stack in most Indian hotels.
The guest hands over an ID. Someone photocopies it, then copies the name, address and ID details into a register by hand. It's slow for the guest, tedious for staff, and it was the standard in the majority of hotels.
DigiStay set out to digitise how Indian hotels run: check-in, guest experience, hotel management. I joined as a product consultant before the company was even incorporated and worked directly with the founders to build and launch the MVP.
The rival wasn't another app. It was the register.
Staff who aren't tech-savvy
Front desks run on people who never signed up for software, and turnover is high. Whatever we built had to be learnable in minutes, again and again.
Guests at the end of a journey
They're tired and want their room. Every extra minute at the desk is felt.
Incumbents nobody adopts
The platforms I analysed were feature-heavy, dated and slow to learn. The gap wasn't capability. It was usability.
- Take ID
- Photocopy
- Copy name, address, ID no.
- Phone, room, rate
- File the copy
- Photo of ID
- OCR autofills
- Type phone + rate
- Checked in
Speed wasn't a feature. It was the adoption plan.
“The digital process must be faster than paper, or nobody will use it.”
That rule cut scope hard. Staff shouldn't type anything a camera could read. The only fields left for a human were the ones no ID contains: the guest's phone number and the room rate.
Click the ID. Type two fields. Everything else fills itself.
Most guests already carry Aadhaar, a driving licence or a passport. We trained a custom OCR model to pull the name, address and ID type from a photo of it. Staff take the photo, type a phone number and room rate, and the check-in is done in under 2 minutes.
I sat with hotel staff on-ground while they used it. They learned it in about 10 minutes, which mattered more than any feature, given how often front-desk teams change.
NameAarav Sharma
ID typeAadhaar
Address14 Lake View Rd, Indiranagar, Bengaluru
ID no.XXXX XXXX 4821
Guest's ID · fictional sample
At launch, the model wasn't the problem. The photos were.
The first version struggled. Uploaded images were blurry, poorly lit or cut off at the edges, and no model reads a name from half a card.
We rebuilt the capture step around real-time guidance: the screen highlights when the ID sits correctly in the frame, and a green tick confirms it's good to go. The fix wasn't a smarter model. It was helping people take a photo the model could read.
Point the camera at the ID
Waiting for a photo…
- In focus
- Well lit
- Whole card in frame
Growth could wait a sprint. Trust couldn't.
We'd just launched and were pushing hard: acquire hotels fast, build billing, integrate a channel manager. Meanwhile, the hotels we already had were hitting real problems: slow image upload, a camera that didn't work, names extracted wrongly.
Talking to those hotels convinced me we had to step off the roadmap. I argued to the founders that product-market fit on the core feature came before expansion: if current users aren't adopting what we built, adding more won't help. It meant slowing aggressive growth targets, and I pushed for it anyway. We fixed the core first, then went back to new features.
Grow
- Acquire hotels fast
- Build billing
- Channel-manager integration
Fix
- Slow image upload
- Camera not working
- Names extracted wrongly
CallFix the core, earn the trust, then expand.
Staff told us where things belonged. We moved them there.
One room, one tab
Staff ignored food ordering because it lived in a separate tab from room details. At billing time they wanted everything under the room. We merged the two: +15% engagement.
Direct bookings for repeat guests
Roughly 15–20% of guests were repeat visitors. We offered them direct booking through the platform at a much lower commission than OTAs: +12% direct bookings for pilot hotels.
- 0:00Food-order call missed
A guest orders; the desk doesn't pick up.
- Right awayWhatsApp to staff
Confirm the order.
- +5 minReminder
No confirmation within 5 minutes.
- ThenPhone call to staff
Still nothing? Ring them.
- ThenGuest told it's cancelled
Automatically, so nobody waits for food that isn't coming.
Measured at the desk, not in a demo.
- Check-in time
- Staff time to check a guest in, from ID in hand to a complete record. Paper: 10–15 minutes. Digital: under 2.
- Where
- On-ground, with front-desk staff in real hotels. We refined the flow until it reliably beat paper.
- Adoption
- Whether existing hotels actually used it. The problems they raised went ahead of new features.
- Flow changes
- A redesigned check-in flow (UI, button placement) ran in 5 hotels first; we compared results across them before rolling out.
Why at the deskA flow that wins in a demo can still lose at a busy front desk. Timing it on-ground, with the staff who would use it, was the only test of “faster than paper” that counted.
Under two minutes at the desk, 50+ hotels on the platform.
What could speed have cost us?
Extraction accuracy
Fast is worthless if the name is wrong. Wrong extractions were among the issues existing hotels raised, and they went ahead of new features.
Capture quality
Blurry, dark, cropped photos were the first failure. Guidance became part of the flow, not a tip in a manual.
Real adoption, not sign-ups
Sign-ups aren't usage. Talking to existing hotels surfaced the upload, camera and extraction problems, and that is why they went ahead of growth.
“The best product solutions aren't the most technically complex ones, but the ones that genuinely make users' lives easier.”
The OCR model got the attention. The capture frame and the two-field form did the work.