A hotel rate and availability sync pipeline between an internal pricing engine and a Channex-compatible channel manager — the data backbone I own, plus the dashboard I helped build on top of it.
HANZ keeps a hotel chain's room rates and availability in sync — flowing between its internal pricing engine and a Channex-compatible channel manager that pushes to the booking platforms.
I own the heart of it: a five-stage ETL pipeline in Laravel 8 and PostgreSQL that extracts, transforms, and reconciles pricing in bulk — chunked, idempotent, and audited at every stage. On top of it, I contributed to the React inventory dashboard the hotel team uses day to day.
It runs on an hourly full-sync and a five-minute webhook selective-sync, with a collision guard so the two never fight. The hard part of a pipeline like this isn't moving data — it's moving a lot of it, correctly, without stepping on itself.
Here is how rates actually move. A Laravel pipeline extracts pricing in chunks, explodes date ranges, computes the final numbers in PostgreSQL itself, and pushes back to the channel manager — every stage chunked and logged. The React dashboard reads the same data for the hotel team.
14-day rolling rate/availability grid · drag-to-select · bulk-update modal
5-stage flow · chunked extract · CarbonPeriod explosion · bulk upsert via mavinoo/laravel-batch
6,000-row chunk upserts · VIEW-based pricing math (last-minute discount)
internal rate plans + property data
pushes rates out to the booking platforms
A pricing pipeline has one job: be right, at volume, every time. These four are the engineering that earns that trust.
The pipeline upserts in 6,000-row chunks via mavinoo/laravel-batch and pushes back in 10-row chunks, sized to the channel manager's limits. Extraction is chunked too — 50 properties and 15 rate plans per call — and CarbonPeriod explodes every date range so a price exists for every night.
Final rates — including last-minute discounts — are computed by a PostgreSQL VIEW rather than in PHP. The pipeline reads numbers that are already correct; the business rule has exactly one home, and it's the one closest to the data.
An hourly cron runs a full reconciliation; a five-minute webhook handles selective updates. A collision guard keeps the two from touching the same property at once — so a fast webhook never races a slow full-sync into an inconsistent state.
The pipeline writes to three audit-log tables and posts a Telegram alert per stage — so when a sync goes wrong, the failure is already located and timestamped before anyone goes looking for it.
Reported from the HANZ codebases at NEWDEV — pipeline stage, chunk-size, and dashboard figures taken from the Laravel and React projects. Snapshot taken May 2026.