Everyone claims reliability. Here is ours, in numbers.
These figures come from delivery manifests and the row counts of files we actually shipped, not from a marketing estimate. Where we can show you the schema, we show you the schema.
Aggregated from SFTP transfer manifests and parquet metadata across live pipelines. Clients are described by sector rather than named — we publish a client's name only once they've agreed to it in writing.
Pipelines currently running.
Cadence, volume, format, and error count for each. The error column is the one worth reading.
| Sector | Dataset | Files | Cadence | Span | Format | Errors |
|---|---|---|---|---|---|---|
| Event ticketingDaily event snapshots plus ad-hoc listing pulls with full fee breakdown. | Event + listing pricing | 475 | Daily | 18 Jan – 15 Aug 2026 | Parquet → SFTP | 0 |
| Restaurant deliveryStore-level menu and pricing coverage across a national delivery platform. | Menu pricing, ratings & reviews | 79 | Weekly | Rolling | Parquet → SFTP | 0 |
| Electronic componentsManufacturer part lists in, resolved availability and pricing out. | Part search + availability | 12 | Monthly | Apr – Aug 2026 | TSV in, ZIP out | 0 |
| Quick-service restaurantsPer-store pricing across a national franchise footprint. | Store-level menu pricing | 6 | Scheduled | Rolling | Parquet → SFTP | 0 |
This is what actually lands.
Not a mocked-up example, the real field list from a delivered listings file, including the full fee breakdown that makes the difference between a headline price and what a buyer really pays.
- File
- listings_20260511.parquet
- Rows in this file
- 1,866,027
- Compressed size
- 1.6 MB
- Fields
- 18
1.87 million rows in 1.6 MB, because columnar formats compress well and we deliver in one. The same data as CSV would be roughly forty times the size.
| 01 | marketplace | string |
| 02 | event_id | int64 |
| 03 | listing_id | string |
| 04 | section_id | string |
| 05 | section_name | string |
| 06 | row_name | string |
| 07 | seat_numbers | string |
| 08 | ticket_quantity | int32 |
| 09 | value_score | double |
| 10 | quality_score | double |
| 11 | display_price_pre_checkout | decimal |
| 12 | all_in_price_pre_checkout | decimal |
| 13 | display_price_checkout | decimal |
| 14 | buyer_fee_checkout | decimal |
| 15 | other_fee_checkout | decimal |
| 16 | sales_tax_checkout | decimal |
| 17 | all_in_price_checkout | decimal |
| 18 | cache_time | timestamp |
The work most vendors quietly decline.
“Every source will work” is an easy promise to make. Here is the specific scope we assess, including the two lines we won't cross, which matter just as much.
JavaScript-rendered pages, changing layouts, pagination, and high-volume public listings managed as production pipelines.
Structured public endpoints and feeds integrated where their access conditions permit the requested use.
Publicly visible commercial data normalized across web and app formats when access is permitted and no credentials are required.
Prices and availability that differ by location, collected with a documented market context rather than mixed into one dataset.
Millions of records per day on a fixed daily window, sustained over months rather than benchmarked once.
We don't defeat authentication to reach data that isn't public, whatever the project is worth.
If a dataset would contain personal data, there has to be a basis for collecting it. We raise this in scoping, not after delivery.
Top Rated on Upwork
Upwork's Top Rated tier, held on a verified profile with 100% job success across 37 contracts. Awarded by Upwork on completed work, not self-reported.
Check the profile yourself
Directory listings rather than quality ratings — we submitted to these ourselves. They're here because the pages are public and you can open them, like everything else on this one.
Ask us to prove it on your sources.
The most useful test isn't our numbers. It's whether we can collect the sites you care about. Send the list and we'll extract a real sample from them, free, before you commit to anything.
contactus@pyronets.comSend us three sources. We'll send back real records.
No retainer, no obligation. You'll get a feasibility read on each source, a proposed schema, and a genuine sample extracted from your own targets.
No retainer required to find out whether your sources are feasible.