Booking.com price monitoring: budget for 100 hotels across 30 dates
Monitoring 100 hotels across 30 check-in dates creates 3,000 hotel-date observations per run. Whether that means 3,000 requests depends on the data channel, batching and room requirements.
Last updated Nov 16, 2026
Specify a comparable hotel price
A community request for Booking.com hotel ratesasks for prices, availability and room types, while describing difficulties getting data access. That motivates our example; it does not establish today's protection stack.
Imagine a hotel operator tracking 100 competitors each day for the next 30 check-in dates. Start with one night, two adults, one room, one currency, one booker country and one platform. Define whether the comparison uses the lowest available offer or a particular room type, cancellation policy and meal plan. A refundable room with breakfast is not equivalent to a non-refundable room-only offer. Record total stay price and charge inclusion, not just an ambiguous nightly headline.
Check access before designing the pipeline
Booking.com's Demand API prerequisitesrequire a Managed Affiliate Partner relationship, Partner Centre access and credentials. Access is not established simply by opening an ordinary customer account. Assess whether the intended use is available under the relevant agreement before treating this as your data source.
The Demand API v3.3 availability endpoint accepts accommodation IDs with check-in, check-out, booker and guest context, and documents up to 50 accommodations in one call. It can return prices and availability with additional charge information requested separately through its options. A specialist hotel-data supplier may offer another delivery contract; a generic fetching API leaves parsing and offer comparison to you.
Our proposed pipeline is stable hotel IDs → rolling date and guest combinations → availability observations → room and rate-plan normalization → comparable-price history. Distinguish no matching availability from failed collection. A missing result is not evidence that the whole property is sold out, and a hotel-price monitor does not establish occupancy or revenue.
Count combinations, then count requests
Observations per run = hotels × check-in dates × stay lengths × guest profiles
Monthly observations = observations per run × runs per day × days| Scenario | Lengths × guest profiles | Runs / day | Monthly observations |
|---|---|---|---|
| Baseline | 1 × 1 | 1 | 90,000 |
| More comparable searches | 3 × 2 | 1 | 540,000 |
| Four daily updates | 3 × 2 | 4 | 2,160,000 |
Scroll tables sideways to see all columns.
In a one-hotel-per-fetch implementation, the baseline means 90,000 successful fetches per month, before retries or extra room requests. In an eligible API integration, if each documented 50-hotel batch can share the same request context, the same baseline could use 2 batches × 30 check-in dates × 30 runs = 1,800 calls. That is a conditional batching calculation, not a measured delivery rate or a promise about your agreement.
Do not automatically multiply by every room type: a response may already include multiple products. Count extra fetches only when the chosen route needs them. The 30-day rolling horizon is rechecked each day, so observations deliberately repeat a hotel-date pair at different times. They are not 90,000 distinct hotels or dates.
Separate access problems from comparison errors
For a website-based implementation, unresolved questions include whether dates and occupancy persist across navigation, whether required prices need rendering, and how challenges, throttling and session expiry are detected. We have not tested these conditions or identified a protection vendor. For an authorized API, inspect method quotas and request validation instead. In both cases, keep country, currency, platform and offer conditions consistent before comparing prices.
Use bounded retry handling for collection errors and retain their status. Do not retry a valid no-availability observation until it becomes a price: that would bias the dataset and waste resources. Require freshness timestamps from a third-party data feed, and check whether its price includes the requested taxes and fees before declaring two offers comparable.
Estimate resources against a completion window
Under the one-hotel-per-fetch baseline, one daily run needs 3,000 successful fetches. With invented assumptions of 90% fetch success and 10 seconds per attempt, a one-hour completion window needs about 10 continuously occupied workers: ceil(3,000 ÷ 0.90 × 10 ÷ 3,600). Use 20 as a comparison allowance for headroom, then revise it against the chosen route's actual throughput and limits.
The baseline produces 100,000 attempts/month at 90% success. At an assumed 300 KB each, that is 30 GB in decimal units; if every attempt occupied a browser for 10 seconds, it would use about 277.8 browser-hours before idle time or overhead. Browser time is an alternative scenario, not a demonstrated requirement. For a success-billed API, the billable page count may remain 90,000 even though retries consume resources internally.
Turn the workload into a plan shortlist
| Route | Baseline planning unit | What remains to verify |
|---|---|---|
| Demand API, eligible access | 1,800 batched calls under the stated assumptions | Agreement, quotas, permitted use and any charges |
| Specialist hotel data | 90,000 hotel-date observations | Room coverage, freshness, batching and billed unit |
| One hotel per managed fetch | 90,000 successful pages | Rendering, context support, domain tariffs and parser |
These links compare generic infrastructure, not verified prices for this target. They assume HTTP-only fetching, unknown protection, 20 concurrent requests, monthly billing, no required exit country and no budget cap. Page size, success rate and browser duration use the calculator's disclosed defaults. Change these inputs when you have evidence. The form does not transmit a target-specific billing category.
The links model only the one-hotel-per-fetch branch. Do not put the batched API's 1,800 calls into the generic calculator and describe the result as Booking.com's API price. ScrapeSelect does not price the partner agreement or certify target support. Keep a plan only if the provider confirms the required fields, context, tariff and completion window; then compare its monthly commitment, allowance and overage with the same workload.
Start procurement with the 90,000-observation specification, not a claim that 100 hotels means 100 requests per day. If budget is constrained, reduce update frequency or the number of stay and guest combinations before sacrificing comparability.
Continue with Amazon frequency and plan selection or business-directory yield.