---
title: "Failed prints, purge and supports: costing the filament and hours that never ship"
description: Price failed prints by dividing the cost of every attempt by the good parts. Measure your own failure rate, and split colour-change purge across a plate.
updated: 2026-11-21
canonical: https://printfarmerp.com/blog/failed-prints-reprint-cost/
---

# Failed prints, purge and supports: costing the filament and hours that never ship

By the PrintFarmer team. Published 21 November 2026. Updated 21 November 2026.
https://printfarmerp.com/blog/failed-prints-reprint-cost/

## Short answer

Account for failed prints by adding up the material and machine time of every attempt, including the ones that failed, and dividing by the number of good parts. Measure your own failure rate for a month instead of borrowing a 10 percent rule of thumb, weight each failure by how far it got, and price colour-change purge per plate, split across the copies on it.

## Key takeaways

- Cost per good part is the cost of all attempts divided by the parts that shipped; a flat percentage is a shortcut for the same thing.
- Failures per attempt are divided out (cost ÷ (1 − rate)); failures per good part are multiplied in (cost × (1 + rate)). Mixing them up misprices the job.
- A failure at 90 percent of a four-hour print costs about 45 times one at 2 percent, so weight failures by hours run, not by count.
- Colour-change purge depends on the number of filament changes per layer, not the number of parts, so copies printed together on one plate share it.
- Record wastage separately from usage, or a printer eating a spool a week hides inside the cost of good parts.

To account for failed prints, add up the material and machine time of every attempt, including the attempts that failed, and divide by the number of good parts that shipped. A failure that stopped at layer three costs almost nothing; one that stopped an hour from the end costs nearly a whole part. So measure how far each failure got, not just how many there were, and price from your own numbers rather than a borrowed percentage.

This article covers the three kinds of filament that never ship, how to measure your own failure rate in a month, two formulas, a worked example in rupees, and how to split colour-change purge across copies on a plate. It assumes you already have a base cost per part; if not, start with [how to price a 3D print in India](https://printfarmerp.com/blog/price-3d-prints-india/).

## What filament never reaches the customer?

Three kinds, and they behave differently in a price:

1. **Failed prints.** Spaghetti, a part that let go of the bed, a spool that ran out, a power cut. Random, and costed as a rate.
2. **Supports, brims and rafts.** Planned waste. The slicer already counts these grams in its estimate for the plate, so they belong in the base cost of the part, not in a failure allowance.
3. **Colour-change purge.** On a single-nozzle printer with an AMS or similar changer, each switch of filament pushes the old colour out through the nozzle before printing resumes. Bambu's wiki calls this flushing, and by default the nozzle then primes on a prime tower, a sacrificial piece printed alongside the parts.[[2]](https://wiki.bambulab.com/en/software/bambu-studio/multi-color-printing)[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change) Planned too, but it depends on the plate, not the part.

Only the first needs a failure rate. The other two come out of the slicer, if you read them correctly.

## How do I measure my failure rate?

Keep a log for one month, one row per plate that came off a printer. Four columns are enough:

| Plate (date, printer, material) | Parts good / failed | Hours and grams lost | Cause |
|---|---|---|---|
| 03 Nov, P1S-2, PLA black | 6 / 0 | 0 h, 0 g | |
| 03 Nov, P1S-3, PETG grey | 0 / 4 | 2.5 of 5.0 h, 110 g | Lifted at corner |
| 04 Nov, A1-1, PLA white | 11 / 1 | 0.2 h, 6 g | One part knocked over |

These rows are an illustration of the format, not data. Weigh the failed print to get grams lost; the [filament by weight article](https://printfarmerp.com/blog/track-filament-by-weight/) covers doing that against a known tare.

At the end of the month work out two numbers:

- **Count rate:** failed parts divided by parts attempted.
- **Time-weighted rate:** hours lost divided by hours of good output.

The second is the one to price from, because it reflects how far each failure got. Note that it is measured against good output, not attempts, so it goes into the multiply form of formula one below, not the divide form. The cause column is what tells you whether a fix (a cleaner bed, a drier spool) would be cheaper than the allowance.

Why not borrow a figure? Snapmaker's cost guide uses a 10 percent example, and the raspberry.tips calculator uses 10 percent as a stated assumption while noting that published figures vary between 1 and 20 percent.[[8]](https://www.snapmaker.com/blog/how-to-calculate-your-3d-printing-costs/)[[9]](https://raspberry.tips/en/3d-printing-cost-calculator) A range that wide is a reason to measure, not a number to use. Vendor-run comparisons of printer failure rates exist too, but they compare specific machines under test conditions and do not tell you about your bench.

## Formula one: a flat failure allowance

For everyday work where failures are random and similar in size:

**Cost per good part = base cost ÷ (1 − failure rate per attempt)**

Snapmaker's guide uses this form, with a 10 percent example in which a $10 base cost becomes $11.11.[[8]](https://www.snapmaker.com/blog/how-to-calculate-your-3d-printing-costs/) With the ₹126 base cost from the worked example below, 126 ÷ 0.9 = ₹140.00.

Not every calculator does it this way. The raspberry.tips calculator adds the rate as a percentage of cost, which is base cost × (1 + rate).[[9]](https://raspberry.tips/en/3d-printing-cost-calculator) Both are right, for different rates:

- If the rate is **failures per attempt** (4 failed out of 54 started), divide by (1 − rate).
- If the rate is **failures per good part** (4 failed per 50 shipped), multiply by (1 + rate).

With 4 failures in 54 attempts, 126 ÷ (1 − 4/54) = 126 × 54/50 = ₹136.08, and 126 × (1 + 4/50) = 126 × 1.08 = ₹136.08. Same answer. Use a per-attempt rate in the multiply form and you under-price slightly; the gap widens as the rate grows.

## Formula two: cost per good part, for batches and risky geometry

For a batch, a tall part, a new material or anything you have not printed before, cost the batch directly:

**Cost per good part = (cost of good attempts + cost of failed attempts) ÷ good parts**

where each failed attempt is costed for the fraction it completed. This is also how to check a quote after the job, and it is what feeds your flat rate for next month.

## Worked example: a batch of 50 where 4 fail

Assumptions, stated so you can replace them with yours:

- Part: 60 g of PLA, 4 hours per part, one part per plate.
- Filament at ₹1,200 per kg, so ₹1.20 per gram.
- Machine rate of ₹13.50 per hour (depreciation, power and maintenance for a Bambu Lab P1S), as worked out in the [pricing article](https://printfarmerp.com/blog/price-3d-prints-india/).
- Filament used by a failed print is taken as proportional to how far it got. Weighing it is better.

Base cost of one attempt: 60 × 1.20 = ₹72.00 material, plus 4 × 13.50 = ₹54.00 machine time, so **₹126.00**.

The four failures:

| Failure | How far it got | Material | Machine time | Cost |
|---|---|---|---|---|
| Lifted at layer three | 2% | ₹1.44 | ₹1.08 | ₹2.52 |
| Spaghetti | 25% | ₹18.00 | ₹13.50 | ₹31.50 |
| Spool ran out | 50% | ₹36.00 | ₹27.00 | ₹63.00 |
| Layer shift near the top | 90% | ₹64.80 | ₹48.60 | ₹113.40 |
| **Total** | **1.67 attempts** | **₹120.24** | **₹90.18** | **₹210.42** |

All four were reprinted successfully, so 54 plates were started and 50 good parts shipped.

- Good attempts: 50 × 126 = ₹6,300.00
- Failed attempts: ₹210.42
- Total: ₹6,510.42
- **Cost per good part: 6,510.42 ÷ 50 = ₹130.21**

The time-weighted rate gives the same figure. The failures lost 1.67 × 4 = 6.68 printer hours against 200 hours of good output, a rate of 3.34 percent per good part, and 126 × 1.0334 = ₹130.21.

Compare the shortcuts:

| Method | Cost per good part |
|---|---|
| Ignore failures | ₹126.00 |
| Weighted by how far each got (actual) | ₹130.21 |
| Count rate, 4 of 54, divided out | ₹136.08 |
| Borrowed 10 percent, divided out | ₹140.00 |

Counting every failure as a whole attempt overstates this batch by about ₹6 a part, because two of the four failed early. On a different batch where most failures come late, the count rate would understate. That is the case for logging hours lost.

The ₹130.21 is cost, not price. Your margin goes on top, as in the [pricing article](https://printfarmerp.com/blog/price-3d-prints-india/).

## A late failure costs more than an early one: pricing by risk

In the example, the layer-shift failure cost ₹113.40 and the layer-three failure ₹2.52, 45 times as much. The raspberry.tips calculator makes the same point: a print that fails partway costs less than a whole part, and failures that come at the very end push the real surcharge above the flat rate.[[9]](https://raspberry.tips/en/3d-printing-cost-calculator)

So price risk by the job, not one rate for everything:

- **Long single-part prints** (eight hours or more) carry more exposure per failure. Use a higher allowance or split the part.
- **Tall, thin geometry and new materials** fail more and later. Quote them with formula two and a first-article print.
- **Small parts many to a plate** lose little per failure, but a plate that lets go loses all of them. Count failures by part, not by plate.

## How much does colour-change purge cost?

### How much filament does one change use?

Bambu Studio calculates a flushing volume for every from/to pair of filaments, based on colour and material.[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change) Dark to light needs more than light to dark.[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change) In Bambu Studio's source code, the fallback value in that table is 280 cubic millimetres per pair, though the auto-calculated values in your project will differ.[[3]](https://github.com/bambulab/BambuStudio/blob/master/src/libslic3r/PrintConfig.cpp) Converted with manufacturer densities:

| Material | Density (data sheet) | 280 mm³ in grams |
|---|---|---|
| Bambu PLA Basic | 1.24 g/cm³[[5]](https://cdn.shopify.com/s/files/1/0380/5149/files/Bambu_PLA_Basic_Technical_Data_Sheet.pdf) | 0.28 × 1.24 = 0.35 g |
| Bambu PETG Basic | 1.25 g/cm³[[6]](https://cdn.shopify.com/s/files/1/0380/5149/files/Bambu_PETG_Basic_Technical_Data_Sheet.pdf) | 0.28 × 1.25 = 0.35 g |
| Bambu ABS | 1.05 g/cm³[[7]](https://www.additive-x.com/shop/mpattachments/file/viewonline/id/566/product_id/2315) | 0.28 × 1.05 = 0.29 g |

Bambu's own wiki gives a real figure: an 11 g four-colour cube at 0.2 mm layers made 153 filament changes and 83 g of wasted filament, about 0.54 g per change and more than seven times the weight of the cube.[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change) Raising the layer height to 0.28 mm cut it to 111 changes and 61 g.[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change) Do not price from either number; read the filament change count and flushed grams from your own sliced plate.

### Why purge is per plate, not per part

The wiki explains that on any layer, the number of filament changes is almost always one less than the number of filaments on that layer.[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change) It does not grow with the number of objects. In the same wiki example, duplicating the cube to nine copies on one plate left the filament changes and the waste unchanged.[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change) With the default "By layer" print sequence, six copies of a two-colour nameplate on one plate make roughly the same number of changes as one copy.[[3]](https://github.com/bambulab/BambuStudio/blob/master/src/libslic3r/PrintConfig.cpp) "By object" sequence prints each copy separately and loses that sharing.

Example, with assumed slicer figures: a plate makes 16 colour changes and wastes 12 g in flush and prime tower. At ₹1.20 per gram, that is ₹14.40 a plate.

- One copy on the plate: ₹14.40 of purge per part.
- Six copies on the plate: 14.40 ÷ 6 = ₹2.40 per part.

So quote multicolour work by plate: purge cost divided by the copies on that plate. A single multicolour piece should carry its full purge, and a customer ordering one should see that in the price.

Bambu Studio and OrcaSlicer both have options that put purge to use: "Flush into objects' support" (on by default, and it needs the prime tower enabled), "Flush into objects' infill" (off by default; can show through light or translucent walls) and "Flush into this object" (off by default), which prints purge into a sacrificial or functional object.[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change)[[3]](https://github.com/bambulab/BambuStudio/blob/master/src/libslic3r/PrintConfig.cpp)[[4]](https://github.com/SoftFever/OrcaSlicer/blob/main/src/libslic3r/PrintConfig.cpp) In Bambu Studio, the flush multiplier (1.00 by default) scales the whole table, and Bambu's wiki suggests trying 0.8 or 0.9 with a test print.[[3]](https://github.com/bambulab/BambuStudio/blob/master/src/libslic3r/PrintConfig.cpp)[[1]](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change)

## Who pays for a failed print or a reprint?

A common split, which is general practice and not legal advice:

- **Shop failures** (adhesion, tangles, run-out, power, a printer fault): yours. They are what the allowance pays for, and the customer should never see them as a line item.
- **File problems you could not reasonably catch** (wrong dimensions, walls too thin to print as designed): the customer pays for a reprint, if the quote said so before you started.
- **Changes after approval** (a new colour, a revised file): a new job, priced as one.

Write it into the quote in plain words, for example: "Reprints caused by print faults are on us. Reprints needed because of an error in the supplied file, or a change after approval, are charged at the quoted rate per part." The [quoting article](https://printfarmerp.com/blog/quote-custom-3d-print-orders/) covers what else to write down.

## Why do wastage and usage belong in separate columns?

If failed grams are booked as usage, the cost of good parts looks higher and nobody can tell why. Kept apart, wastage shows which printer, material or job is losing filament, and stock counts still reconcile because both leave the spool.

## What a failure does to the schedule

The layer-shift failure above cost 3.6 printer hours as well as ₹113.40. A reprint goes back into the queue ahead of or behind other work, and a delivery date promised with no slack moves. Give large batches a buffer of printer hours from your time-weighted rate, and replan the reprint the day the failure is found. The [print queue article](https://printfarmerp.com/blog/print-queue-multiple-printers/) covers keeping orders, parts and plates straight across printers.

## Where PrintFarmer fits

In PrintFarmer, a person records each plate's result as good and failed counts per part. A plate with any failures moves to reprint planning, so the failed parts go back into the queue instead of becoming a short shipment. Filament leaves the stock ledger as two separate movements, usage and wastage. The grams are prefilled from each part's grams per unit and the counts recorded, so a failed part is estimated at its full weight; a person confirms the figure or replaces it with the weighed one. The production report gives the failure counts your count rate comes from.

It does not detect failures. There is no printer connection or camera monitoring; a person at the machine records what happened. It records counts, not how far a failed print got, so the hours lost for a time-weighted rate still come from your own log. It does not read flush volumes from the slicer either; you enter the grams. [Tell us you're interested](https://printfarmerp.com/#interest). It is not launched yet.

## Frequently asked questions

### What failure rate should a 3D print farm assume?

Measure your own for a month. Pricing guides often use 10 percent as a placeholder, and one calculator notes published figures range from 1 to 20 percent, which is too wide to price from. Your rate depends on your printers, materials and the geometry you take on.

### How do I calculate the cost per good part when some prints fail?

Add the material and machine cost of every attempt, counting a failed attempt for the fraction of the print it completed, then divide by the number of good parts. For a single part, that is the base cost divided by one minus the failure rate per attempt.

### Should the customer pay for a failed print?

Not for failures that are the shop's own, such as adhesion, a tangled spool or a power cut; those belong in your price as an allowance. A reprint caused by a file the customer supplied, or a change after they approved it, is reasonable to charge if the quote said so in advance.

### How much filament does an AMS colour change waste?

It depends on the colour pair, the material and the nozzle, and Bambu Studio calculates it per pair. In Bambu's own example, an 11 g four-colour cube at 0.2 mm layers made 153 filament changes and 83 g of waste, about 0.54 g per change.

### What is the difference between filament usage and filament wastage?

Usage is filament that went into parts a customer received. Wastage is filament that did not: failed prints, purge, prime towers and test pieces. Keeping them apart shows what a job really cost and which printer is losing material.

## Sources

1. [Reduce Waste during Filament Change (Bambu Lab Wiki)](https://wiki.bambulab.com/en/software/bambu-studio/reduce-wasting-during-filament-change). Accessed 21 November 2026.
2. [Multi-color printing (Bambu Lab Wiki)](https://wiki.bambulab.com/en/software/bambu-studio/multi-color-printing). Accessed 21 November 2026.
3. [Bambu Studio source, PrintConfig.cpp (flushing volume matrix, flush options, print sequence defaults)](https://github.com/bambulab/BambuStudio/blob/master/src/libslic3r/PrintConfig.cpp). Accessed 21 November 2026.
4. [OrcaSlicer source, PrintConfig.cpp (flush into infill, support and object options)](https://github.com/SoftFever/OrcaSlicer/blob/main/src/libslic3r/PrintConfig.cpp). Accessed 21 November 2026.
5. [Bambu PLA Basic Technical Data Sheet (density 1.24 g/cm³, ISO 1183)](https://cdn.shopify.com/s/files/1/0380/5149/files/Bambu_PLA_Basic_Technical_Data_Sheet.pdf). Accessed 21 November 2026.
6. [Bambu PETG Basic Technical Data Sheet (density 1.25 g/cm³, ISO 1183)](https://cdn.shopify.com/s/files/1/0380/5149/files/Bambu_PETG_Basic_Technical_Data_Sheet.pdf). Accessed 21 November 2026.
7. [Bambu ABS Technical Data Sheet V2.0, reseller-hosted copy (density 1.05 g/cm³, ISO 1183)](https://www.additive-x.com/shop/mpattachments/file/viewonline/id/566/product_id/2315). Accessed 21 November 2026.
8. [How to Calculate Your 3D Printing Costs (Snapmaker)](https://www.snapmaker.com/blog/how-to-calculate-your-3d-printing-costs/). Accessed 21 November 2026.
9. [3D printing cost calculator (raspberry.tips)](https://raspberry.tips/en/3d-printing-cost-calculator). Accessed 21 November 2026.
