Methodology & limitations

Everything published here is computed from our own daily capture. This page describes exactly how, and where the data should not be trusted.

How the data is collected

Once a day we read the public product listings published by each retailer — the same catalogue data their own storefronts serve to shoppers — plus a licensed retail product API. We record price, list price and stock state for every filament and printer listing, then store the result.

We log an observation only when something changes: a listing appears, its price moves, its stock state flips, or it disappears. A separate ledger records which retailers were successfully checked on each run, so a gap in the record can always be distinguished from a retailer we simply could not reach that day. Across 1,190 retailer-day checks so far, 51 failed and are recorded as failures rather than as missing products.

How cost per kilogram is calculated

Spool sizes vary, so all filament prices are normalised to cost per kilogram. Net spool weight is read from the product listing itself. Where a listing does not state an unambiguous weight, we exclude it rather than assume it is a 1 kg spool — 15,432 of 17,448 tracked listings currently carry a usable weight.

We deliberately do not use the shipping weight retailers publish, because it includes the spool and packaging: a 1 kg spool commonly ships at 1.3–1.5 kg. Sample coils are excluded from per-kilogram figures.

Known limitations

• Listed prices only. We record the price shown on a public listing. We cannot see cart-level volume discounts, trade or wholesale pricing, or negotiated rates, so per-kilogram figures overstate what a high-volume buyer actually pays.
• Not every retailer. We track 21 retailers. Several large sellers publish no usable public catalogue and are absent. This is not a census of the market.
• Storefronts that share a catalogue are counted once. Where two storefronts list the same products at the same prices and move them together, we treat them as a single retailer, because counting both would overstate how many independent price signals exist.
• Outage lengths are under-counted. Outages still ongoing at the end of the window have no measured length, which biases the median downward.
• Short history. Filament tracking began 2026-08-13; printers on 2026-09-06. Seasonal effects and major sale events are not yet represented, so nothing here should be read as a long-run trend.
• Shipping and tax are excluded throughout.

What we are not

We do not sell filament or printers, and we are not affiliated with any manufacturer. Figures are descriptive observations of published listings, not claims about any retailer's conduct, and not purchasing advice.

Questions or corrections: engineer.boudreaux@gmail.com