What to Build for Hotels and Travel Operators in India

For: Owner or COO of a mid-size hotel group or inbound travel operator in India — running 3–15 properties or 500+ tour packages a year — who is spending on three or four disconnected SaaS tools and still losing margin to manual reconciliation, OTA commissions, and no-shows they could have predicted

If you run 3–15 hotels or a mid-sized inbound travel operation in India, build custom software for exactly three things in this order: an OTA-to-GST reconciliation engine, a direct-booking channel with a real cancellation-risk model, and a consolidated guest and traveller profile that survives across properties and package types. Everything else — PMS, channel manager, accounting, CRM email sends — stays as SaaS. The reason is unglamorous: the money you are losing is not sitting inside any one tool, it is sitting in the seams between them, and those seams are shaped by Indian GST rules that no global SaaS vendor prices in.

This post is for owners and COOs who are already spending on Cloudbeds or Hotelogix or eZee, plus a channel manager, plus Tally or Zoho Books, plus something for CRM — and still cannot answer the question "what did we actually net from MakeMyTrip last month, after commission, after GST mismatch, after no-shows?" in under a week. If that is you, keep reading.

The three problems that are actually costing you money

1. The gap between OTA-reported rate and GST-reconciled net

Everyone talks about OTA commissions. The published number for MakeMyTrip/Goibibo is 18–25% once you include promotional participation, per Jhattse Business's 2026 breakdown. Once you add preferred-partner fees, payment processing, and promo discount participation, the real all-in cost lands closer to 24–28%.

That number is bad. But it is not the number that kills you. What kills you is that a single OTA booking generates at least three independent GST streams — the hotel's accommodation tax, the OTA's commission tax, and the OTA's convenience fee tax — and none of them follow the same reconciliation path. A single folio can carry up to four GST rates (12% or 18% on the room depending on tariff, 5% or 18% on F&B, 18% on banquet and laundry), but most PMS exports flatten all of it into one tax line.

The consequence: you find out about the mismatch quarterly, when your CA flags GSTR-1 against actual bank settlements. By then the disputed booking is 60–90 days old, the OTA account manager has rotated, and your leverage is gone.

What to build: a reconciliation engine that pulls three feeds daily — PMS bookings (with folio-level tax splits preserved), OTA extranet reports (MMT, Booking.com, Agoda, Cleartrip), and bank settlement statements — and matches them at the reservation-ID level. Every mismatch above a threshold becomes a ticket with an SLA against the OTA. GST splits stay intact all the way through, so your GSTR-1 filing pulls straight from reconciled data instead of a manual re-key.

What changes: month-end close for a 5-property group typically compresses from 3–5 staff days to a few hours of exception review. More importantly, your OTA dispute window shrinks from 90 days to 7. That is where the recovered margin actually shows up.

What it takes: this is a 10–14 week build for a group with 3–5 OTA integrations and one PMS. The cost driver is not the matching logic — that part is straightforward — it is the OTA extranet integrations, because MMT, Booking.com, and Agoda each expose data differently and change formats without notice. Every additional OTA is roughly 2–3 weeks of integration and ongoing maintenance. If your PMS does not expose folio-level tax splits via API (many do not), add a data-normalisation layer, which is another 3–4 weeks.

2. Direct bookings with a cancellation-risk model

OTA bookings have roughly 3× the no-show rate of direct bookings, with OTA cancellation rates in Asia running 24–42% versus 23% for direct, per D-EDGE's 2024 report. Every OTA booking is a lower-quality booking than a direct one, and you are paying 18–25% for the privilege.

The shift is achievable. A 60-room Udaipur hotel that moved MakeMyTrip inventory allocation from 40% to 20% and grew direct bookings from 25% to 45% saved ₹8.4 lakh annually in commissions. That is one property. Multiply across a group.

What to build: a booking engine on your own domain that (a) surfaces best-available-rate parity with a clear direct-booking incentive (loyalty points, room upgrade, late checkout — pick one, not three), (b) captures WhatsApp opt-in at booking, and (c) runs a lightweight risk model on each incoming reservation to flag high-cancellation-probability bookings for a proactive confirmation call 48 hours before check-in. The model does not need to be exotic — booking lead time, payment method, guest history, seasonality, source campaign, and room type get you most of the way.

What it is bad at: a custom booking engine will not out-market MakeMyTrip. Nobody discovers your hotel on your website. This lever only works if you already have brand pull — repeat guests, corporate accounts, wedding referrals — or a category advantage (heritage property, wellness, boutique). If you are a generic business hotel in a metro fighting on price, keep the OTA volume and focus on reconciliation instead.

What it takes: 8–12 weeks for the booking engine and WhatsApp confirmation flow, another 6–8 weeks for the risk model once you have 6–12 months of clean booking + no-show data to train on. If you do not have that data, build the booking engine first and the model later. Do not let a vendor sell you an "AI cancellation predictor" without your own data behind it.

3. Consolidated guest and traveller profile

Your 45-year-old repeat guest who stays at your Jaipur property in December and your Goa property in April is two different people in your systems. Your inbound tour operator has a Munich-based agent who has sent 30 travellers over three years across four package types, and nobody in your ops team can pull that history in under 20 minutes.

This matters for two reasons that show up on the P&L: upsell conversion (a repeat guest with a known F&B preference converts 3–4× better on pre-arrival upsell than a cold one) and cancellation risk (a guest with three prior stays and zero cancellations is a fundamentally different risk than a first-time OTA booker, but your PMS treats them identically).

What to build: a guest data platform — really just a well-modelled Postgres database with an ID-resolution layer — that ingests from PMS, booking engine, WhatsApp, tour package system, and F&B POS, and resolves identities using phone number, email, and passport/ID number. Expose it to your front-office and reservations teams as a simple guest lookup, and to your marketing automation tool (keep this as SaaS — Zoho, WebEngage, whatever) as a clean segment feed.

What it takes: 6–10 weeks for the ingestion and ID-resolution layer if your source systems have decent APIs. The real work is data hygiene — deduplicating years of dirty guest records — and that is people-time, not engineering-time. Budget for a data analyst for 4–6 weeks on top of the build.

What tour operators specifically should add

If you are running 500+ packages a year — inbound, outbound, or domestic — three additional things matter:

Dynamic package pricing with supplier-cost sync. Your margins on a Kerala backwaters package move every time a houseboat operator or Kochi hotel revises rates. Most operators re-price quarterly and lose 4–8 points of margin between revisions. A pricing engine that pulls supplier rates weekly and re-suggests package prices is a 6–8 week build, and it pays back inside a season.

TCS handling on the booking flow. The Union Budget 2026–27 reduced TCS on overseas tour packages to a flat 2%, down from the previous 5%/20% two-slab regime. If your booking system still has the old logic hardcoded, you are either over-collecting (and refunding, which annoys customers) or under-collecting (and eating it). Fix this in your booking flow, not in a spreadsheet at month-end. Talk to your CA before finalising the logic — the implementation details matter.

Voucher and DMC reconciliation. Same shape as OTA reconciliation for hotels, but with DMCs and ground handlers instead. Same build, different data sources. If you are also running a hotel group, share the reconciliation engine — the matching logic is identical.

What to keep as SaaS (and stop trying to replace)

• PMS — Hotelogix, eZee, Cloudbeds all work. Do not build this. The mid-scale segment is projected to grow from USD 3.75B in 2023 to USD 6.3B by 2030, and cloud PMS vendors are investing accordingly. You will not out-build them.
• Channel manager — SiteMinder, STAAH, RateGain. Same logic. Buy.
• Accounting — Tally or Zoho Books. Your CA already knows them.
• Email/WhatsApp marketing — WebEngage, MoEngage, WATI. Feed them from your custom guest data layer, but do not build the sender.
• Payment gateway — Razorpay, Cashfree. Never build.

Sequencing: what to do in which quarter

Order matters here, because each build makes the next one easier.

Quarter 1 — Reconciliation engine. Highest ROI, lowest dependency on other systems. Pays for itself inside two OTA settlement cycles. Also forces you to clean up your PMS tax-split configuration, which you will need for everything else.

Quarter 2 — Guest data platform. Now that your PMS and OTA feeds are structured, ingest them into a proper guest layer. Do the deduplication work.

Quarter 3 — Direct booking engine with WhatsApp confirmation. You now have clean guest data to power personalisation and a reconciliation engine to prove the commission savings on your board deck.

Quarter 4 — Cancellation risk model and dynamic pricing. Both need the previous three quarters of clean data to work. Attempting them earlier is the most common mid-market mistake — you end up with a model trained on garbage.

An honest note: only about 1 in 3 multi-property mid-scale brands have adopted cloud-based platforms so far. If you are still on an on-premise PMS or spreadsheets, sequence Quarter 0 before all of this: move to a cloud PMS. Do not build custom software on top of a system you cannot get data out of.

The cost and timeline shape

Total build across the four quarters, for a group of 3–5 properties with 2–4 OTA channels, sits in the range of a mid-six-figure INR investment for the reconciliation engine alone, scaling with the number of OTAs and PMS complexity. The full four-quarter roadmap is a multi-year software asset, not a project. What moves the number most:

• How many OTAs and DMCs you integrate — each adds 2–3 weeks and ongoing maintenance
• Whether your PMS has a real API — the difference between a 10-week and a 16-week reconciliation build
• Data cleanliness in your existing systems — dirty guest records add analyst time, not engineering time, but they add real weeks
• Whether you want mobile apps for front-office staff or just web dashboards — apps roughly double the front-end scope

The three specifics that turn any of this into an actual quote: (1) your current PMS and whether it exposes folio-level tax data, (2) the exact list of OTA and DMC channels you sell through, and (3) whether you already have 12+ months of clean booking history to train models on. If you want to walk through those against your own operation, that is what a scoping conversation with a team like CodeNicely's India practice — or any competent build partner — is for.

What not to build (yet)

• An in-house channel manager. The maintenance cost of keeping up with OTA API changes is a full-time team. Buy.
• A "revenue management AI". Vendors will sell you one. Unless you have three years of clean occupancy, rate, and demand data, you are buying a black box that outputs their opinion, not yours.
• A guest-facing mobile app. Guests do not want another app. They want WhatsApp. Build the WhatsApp flow first, and only consider an app if you have a loyalty programme with real engaged repeat volume.
• Blockchain-anything. No.

Frequently Asked Questions

Should a 4-property hotel group replace its PMS with something custom?

Almost never. Cloud PMS vendors like Hotelogix, eZee, and Cloudbeds are investing at a scale a mid-size group cannot match, and your CA and front-office staff already know the workflows. Build custom on top of the PMS — reconciliation, guest data, direct booking — not underneath it. The exception is if you are on an on-premise system with no API, in which case migrate to a cloud PMS first and revisit custom builds after.

How much does it actually cost to build a custom booking system for hotels in India?

The range is wide because the drivers vary so much: number of OTA integrations, PMS API quality, whether you need multi-property inventory logic, WhatsApp and payment flows, and whether the cancellation risk model is in scope. A basic direct booking engine with payment integration and a single PMS sync is a smaller build than a multi-property engine with dynamic pricing and risk scoring. The three questions that turn the range into a quote: which PMS, how many properties, and whether the risk model is Phase 1 or Phase 2.

Is it worth building software to handle GST reconciliation, or can my CA do it in Excel?

Your CA can do it in Excel — the question is whether the delay costs more than the software. If you are finding OTA mismatches 60–90 days after the booking, you have lost the negotiating window with the OTA. Structured reconciliation compresses month-end close from 3–5 days to a few hours and, more importantly, shrinks the dispute window to about a week. For a group doing meaningful OTA volume, the recovered commission usually funds the build inside a year. This is a matter for your CA to weigh in on for your specific books — get their view before committing.

How long before we see ROI from a direct-booking engine?

Depends entirely on whether you have brand pull. A heritage property or boutique with existing repeat guests and corporate accounts can shift 15–25 points of OTA share to direct within 12 months, per the Udaipur case referenced above. A generic business hotel competing on price in a metro will struggle to shift more than 5–8 points and may never fully pay back the build. Look at your current direct-booking share and repeat-guest ratio before deciding — if repeat guests are already 30%+ of your business, this works.

What changed for tour operators with the Budget 2026–27 TCS update?

The TCS rate on overseas tour packages was reduced to a flat 2%, replacing the earlier 5% (below ₹10 lakh) and 20% (above ₹10 lakh) slabs. If your booking system still has the old two-slab logic hardcoded, fix it — and have your CA confirm the exact treatment for your package structures, because implementation details around LRS thresholds and package composition matter.