pith. sign in

arxiv: 1907.07673 · v1 · pith:OOMRJ6N7new · submitted 2019-07-17 · 💻 cs.LG · stat.ML

Network Based Pricing for 3D Printing Services in Two-Sided Manufacturing-as-a-Service Marketplace

Deepak Pahwa , Binil Starly This is my paper

Pith reviewed 2026-05-24 20:21 UTC · model grok-4.3

classification 💻 cs.LG stat.ML
keywords network based pricing3D printing servicesmachine learningmanufacturing-as-a-servicetwo-sided marketplaceprice classificationsupplier profilingdata mining
0
0 comments X

The pith

Machine learning on scraped marketplace data classifies 3D printing suppliers into one of seven price categories at 65 percent accuracy for US listings.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper demonstrates a data-mining approach that trains machine learning models on supplier profiles scraped from existing 3D printing marketplaces to predict competitive price ranges. Features such as supplier experience, capabilities, customer reviews, and scale of operations are used to assign listings to price categories. The resulting classifiers reach 65 percent accuracy on US suppliers and 59 percent on Europe suppliers, compared with a 25 percent random baseline. This performance suggests that network-derived estimates can replace ad-hoc pricing decisions by service bureaus. The work positions the method as a practical tool that could be embedded directly inside two-sided manufacturing marketplaces.

Core claim

A data mining approach with machine learning methods is used to estimate a price range based on the profile characteristics of 3D printing service suppliers. The model considers factors such as supplier experience, supplier capabilities, customer reviews and ratings from past orders, and scale of operations among others to estimate a price range for suppliers' services. Data was gathered from existing marketplace websites, which was then used to train and test the model. The model demonstrates an accuracy of 65% for US based suppliers and 59% for Europe based suppliers to classify a supplier's 3D Printer listing in one of the seven price categories. The improvement over baseline accuracy of

What carries the argument

Machine learning classifier trained on scraped marketplace data that maps supplier profile features to one of seven price categories.

If this is right

  • Conventional activity-based costing methods are inefficient for strategically pricing 3D printing services in a connected marketplace.
  • Prices can be determined through data mining methods rather than arbitrary or subjective decisions.
  • Tools built on the classifier can be integrated into online marketplaces to assist independent service bureaus in setting rates.
  • Machine learning based methods are promising for network based pricing across manufacturing marketplaces.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same scraped-data approach could be tested on other on-demand manufacturing services that operate through two-sided platforms.
  • Retraining the classifier on larger or multi-platform datasets might raise accuracy above the reported 65 percent level.
  • Service bureaus could run the model periodically on their own listings to detect when their current rates deviate from network norms.

Load-bearing premise

Historical marketplace data scraped from existing websites is representative, unbiased, and sufficient to train a classifier whose output can be used to set competitive prices for new or unseen suppliers.

What would settle it

Apply the trained classifier to a fresh set of suppliers never seen during training and measure whether the fraction of listings placed in the correct price category remains near 65 percent or falls close to the 25 percent baseline.

read the original abstract

This paper presents approaches to determine a network based pricing for 3D printing services in the context of a two-sided manufacturing-as-a-service marketplace. The intent is to provide cost analytics to enable service bureaus to better compete in the market by moving away from setting ad-hoc and subjective prices. A data mining approach with machine learning methods is used to estimate a price range based on the profile characteristics of 3D printing service suppliers. The model considers factors such as supplier experience, supplier capabilities, customer reviews and ratings from past orders, and scale of operations among others to estimate a price range for suppliers' services. Data was gathered from existing marketplace websites, which was then used to train and test the model. The model demonstrates an accuracy of 65% for US based suppliers and 59% for Europe based suppliers to classify a supplier's 3D Printer listing in one of the seven price categories. The improvement over baseline accuracy of 25% demonstrates that machine learning based methods are promising for network based pricing in manufacturing marketplaces. Conventional methodologies for pricing services through activity based costing are inefficient in strategically pricing 3D printing service offering in a connected marketplace. As opposed to arbitrarily determining prices, this work proposes an approach to determine prices through data mining methods to estimate competitive prices. Such tools can be built into online marketplaces to help independent service bureaus to determine service price rates.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper proposes using machine learning on scraped marketplace data to classify 3D printing suppliers into one of seven price categories based on features such as experience, capabilities, reviews, and scale of operations. It reports 65% accuracy for US suppliers and 59% for European suppliers, improving over a 25% baseline, and argues this demonstrates the promise of data-driven network-based pricing over ad-hoc methods in two-sided manufacturing-as-a-service marketplaces.

Significance. If the approach were shown to generalize beyond the training distribution, it could offer a practical tool for service bureaus to set competitive prices in online marketplaces. The core idea of replacing activity-based costing with learned mappings from supplier profiles is potentially useful, but the manuscript provides no evidence of robustness, external validation, or causal utility for pricing decisions.

major comments (3)
  1. [Abstract] Abstract: The headline accuracies (65% US, 59% Europe) and baseline comparison are presented with no information on dataset size, number of samples per region, feature definitions or preprocessing, model architecture or hyperparameters, cross-validation procedure, or class distribution, rendering the central performance claim impossible to evaluate.
  2. [Abstract] Abstract and data description: Price-category labels are derived directly from the same scraped marketplace listings used for training; without external hold-out sets, temporal validation, or live testing, the reported accuracies largely measure in-sample correlation rather than predictive utility for unseen suppliers.
  3. [Abstract] Abstract: The claim that the model can 'estimate a price range for suppliers' services' and 'determine competitive prices' rests on the untested assumption that historical scraped listings are representative of new or unseen suppliers; no analysis of selection bias, survivorship bias, or endogeneity (price affecting visibility) is supplied.
minor comments (1)
  1. [Abstract] Abstract: The seven price categories are referenced but never defined (e.g., whether they are quantiles, fixed dollar ranges, or platform-specific bins).

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback. We agree that the abstract must be expanded with methodological details to allow evaluation of the reported accuracies. We will revise accordingly. We also agree that the evaluation is limited to patterns within the scraped data and that claims about estimating prices for new suppliers require stronger qualification; we will add explicit discussion of these limitations. The core demonstration of feature informativeness over a random baseline remains valid as an initial exploration, though we acknowledge the absence of external validation or bias analysis.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The headline accuracies (65% US, 59% Europe) and baseline comparison are presented with no information on dataset size, number of samples per region, feature definitions or preprocessing, model architecture or hyperparameters, cross-validation procedure, or class distribution, rendering the central performance claim impossible to evaluate.

    Authors: We agree the abstract omits these details. The full manuscript describes data scraping from marketplace sites and use of supplier profile features, but the abstract will be revised to report approximate sample sizes per region, the seven price bins, the feature set (experience, capabilities, reviews, scale), preprocessing, the classifier type, cross-validation method, and class balance. This addresses evaluability without altering the reported numbers. revision: yes

  2. Referee: [Abstract] Abstract and data description: Price-category labels are derived directly from the same scraped marketplace listings used for training; without external hold-out sets, temporal validation, or live testing, the reported accuracies largely measure in-sample correlation rather than predictive utility for unseen suppliers.

    Authors: This observation is correct. The labels and features come from the same listings, and performance is measured via cross-validation on that data. The accuracies therefore capture in-sample associations rather than out-of-distribution prediction. We will revise the abstract and add a limitations paragraph clarifying that the results demonstrate informative features within observed listings but do not constitute validated prediction for entirely new suppliers. No external hold-out or temporal split is available in the current dataset. revision: partial

  3. Referee: [Abstract] Abstract: The claim that the model can 'estimate a price range for suppliers' services' and 'determine competitive prices' rests on the untested assumption that historical scraped listings are representative of new or unseen suppliers; no analysis of selection bias, survivorship bias, or endogeneity (price affecting visibility) is supplied.

    Authors: We acknowledge the assumption is untested. The manuscript contains no explicit analysis of selection, survivorship, or endogeneity biases. We will add a limitations section discussing these issues and their potential impact on generalizability to new suppliers. The work is positioned as an initial data-driven demonstration rather than a production pricing tool; stronger causal or prospective validation would require additional data collection beyond the present study. revision: partial

Circularity Check

1 steps flagged

Price-category classifier accuracy reduces to fitting observed marketplace prices by construction

specific steps
  1. fitted input called prediction [Abstract]
    "The model demonstrates an accuracy of 65% for US based suppliers and 59% for Europe based suppliers to classify a supplier's 3D Printer listing in one of the seven price categories. The improvement over baseline accuracy of 25% demonstrates that machine learning based methods are promising for network based pricing in manufacturing marketplaces."

    Price-category labels are obtained by binning the very same scraped listing prices used as training targets. The classifier is therefore fitted to recover the empirical price distribution; the accuracy number is the in-sample fit quality, not an independent prediction of what prices new suppliers should set.

full rationale

The paper trains and evaluates a supervised classifier whose target labels (seven price bins) are computed directly from the same scraped marketplace listings that supply the input features. Reported accuracies (65%/59%) therefore measure how well the chosen features recover the observed price distribution rather than providing an independent test of a pricing rule for unseen suppliers. This matches the fitted-input-called-prediction pattern; no external hold-out, causal identification, or parameter-free validation is described. No self-citation chains or ansatz smuggling appear in the provided text.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The claim rests on the domain assumption that scraped marketplace listings form an unbiased training distribution and on the standard machine-learning assumption that i.i.d. supervised classification will generalize to future listings.

axioms (1)
  • domain assumption Marketplace data scraped from existing websites is representative of supplier pricing behavior and can be used to train generalizable price classifiers
    Invoked when the authors state that data gathered from marketplace websites was used to train and test the model.

pith-pipeline@v0.9.0 · 5782 in / 1303 out tokens · 31290 ms · 2026-05-24T20:21:57.505516+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

  1. [1]

    and Wang, Y., 2018

    Chan, S.L., Lu, Y. and Wang, Y., 2018. Data -driven cost estimation for additive manufacturing in cybermanufacturing. Journal of manufacturing systems, 46, pp.115-126

    work page 2018

  2. [2]

    and Lin, C.J., 2011

    Chang, C.C. and Lin, C.J., 2011. LIBSVM: a library for support vector machines. ACM transactions on intelligent systems and technology (TIST), 2(3), p.27

    work page 2011

  3. [3]

    and Xie, K., 2017

    Chen, Y. and Xie, K., 2017. Consumer valuation of Airbnb listings: a hedonic pricing approach. International journal of contemporary hospitality management, 29(9), pp.2405-2424

    work page 2017

  4. [4]

    and Rodriguez, N., 2012

    Duran, O., Maciel, J. and Rodriguez, N., 2012. Comparisons between two types of neural networks for manufacturing cost estimation of piping elements. Expert Systems with Applications, 39(9), pp.7788-7795

    work page 2012

  5. [5]

    “Reverse auction mechanism design for the acquisition of prototyping servic es in a manufacturing -as-a-service marketplace, Journal of Manufacturing Systems, 48(Part C), pp

    Pahwa, D., Starly B ., Cohen P., 2018. “Reverse auction mechanism design for the acquisition of prototyping servic es in a manufacturing -as-a-service marketplace, Journal of Manufacturing Systems, 48(Part C), pp. 134-143

    work page 2018

  6. [6]

    and Goodwill, A., 2018

    Gibbs, C., Guttentag, D., Gretzel, U., Morton, J. and Goodwill, A., 2018. Pricing in the sharing economy: a hedonic pricing model applied to Airbnb li stings. Journal of Travel & Tourism Marketing, 35(1), pp.46-56

    work page 2018

  7. [7]

    How much is your spare room worth?

    Hill, D., 2015. How much is your spare room worth?. IEEE Spectrum, 52(9), pp.32-58

    work page 2015

  8. [8]

    and Dicknes, P., 2003

    Hopkinson, N. and Dicknes, P., 2003. Analysis of rapid manufacturing—using layer manufacturing processes for production. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 217(1), pp.31-39. Inside Airbnb Dataset available at: http://insideairbnb.com/get-the-data.html

    work page 2003

  9. [9]

    and Koch, R., 2012, August

    Lindemann, C., Jahnke, U., Moi, M. and Koch, R., 2012, August. Analyzing product lifecycle costs for a better understanding of cost drivers in additive manufacturing. In 23th Annual International Solid Freeform Fabrication Symposium–An Additive Manufacturing Conference. Austin Texas USA 6th-8th August

    work page 2012

  10. [10]

    and Striukova, L., 2016

    Rayna, T. and Striukova, L., 2016. A taxonomy of online 3D printing platforms. In 3D Printing (pp. 153-166). TMC Asser Press, The Hague

    work page 2016

  11. [11]

    and Wegener, K., 2013

    Rickenbacher, L., Spierings, A. and Wegener, K., 2013. An integrated cost- model for selective laser melting (SLM). Rapid Prototyping Journal, 19(3), pp.208-214

    work page 2013

  12. [12]

    and Hague, R., 2006

    Ruffo, M., Tuck, C. and Hague, R., 2006. Cost estimation for rapid manufacturing -laser sintering production for low to medium volumes. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 220(9), pp.1417-1427. Schröder, M., Falk, B. and Schmitt, R., 2015. Evaluation of cost structures of additive manufacturi...

    work page 2006

  13. [13]

    and Sangani, K., 2015

    Tang, E. and Sangani, K., 2015. Neighborhood and price prediction for San Francisco Airbnb listings

    work page 2015