ORDEEN · GUIDES · 6 MIN READ
Ordeen vs Sawtooth Software: Lighthouse Studio and Discover compared
Ordeen
Compare Ordeen with Sawtooth Lighthouse Studio and Discover on methods, Bayesian estimation, simulation, security and pricing.
Book a 20-min demo
Sawtooth Software is a long-established conjoint vendor with a large, experienced user base. It sells two products: Lighthouse Studio, a Windows desktop application for advanced choice modelling, and Discover, a browser-based tool for simpler studies.
Ordeen Choice Modelling is a cloud platform that runs six pricing and choice methods, including adaptive CBC (ACBC) and menu-based choice (MBC), in one browser workspace with a respondent panel built in. Here is how it compares with both Sawtooth products
Side-by-side on every dimension
Annual software price: Ordeen £35,000/yr Annual or £10,000/qtr (5 users in each tier); Sawtooth ~$4,500–$15,000/yr per specialist seat on an annual lock-in.
Panel: Ordeen roughly £3–£4 per complete through the integrated Prodege panel in non-expiring packs (£4 at 500, £3.50 at 1,000, £3 at 2,500); Sawtooth bring-your-own or brokered separately at $4–$12+.
Methods: Ordeen ships all six (CBC, ACBC, MaxDiff, MBC, Van Westendorp, Gabor-Granger) plus nested-logit and latent-class, fit through one estimation pipeline; Sawtooth CBC, ACBC, MaxDiff, MBC, with higher-tier modules priced separately.
Estimation engine: Ordeen GPU-accelerated Hierarchical Bayes with convergence diagnostics (R-hat, ESS, divergences) on every result; any fit that fails the convergence check is withheld rather than shown; Sawtooth CPU Gibbs sampling with a Metropolis-Hastings step and user-run multi-chain comparisons.
Conditional pricing: Ordeen's engine keeps the experimental design orthogonal across tiers in log-price space and un-centers the draws back to raw currency across WTP, simulator, elasticity and TURF; Sawtooth supports CP but is prone to collinearity and convergence issues without manual coding.
Willingness to pay: Ordeen computes Jensen-free aggregate-first WTP at every posterior draw (Bayesian Krinsky-Robb) with Harrell-Davis credible intervals and refuses figures outright when the price slope changes sign across draws; Sawtooth reports point ratios or raw simulation draws, prone to ratio blowups.
Segmentation: Ordeen combines k-means++ on a signal-weighted feature representation with full Bayesian NUTS latent class and Gemini-guardrailed segment naming; Sawtooth offers CCEA / latent class (point-estimate EM) and HB K-means run manually by the analyst.
Simulator and TURF: Ordeen's simulator reports posterior-expectation shares with BBC credible intervals, sub-second elasticity sweeps, paired-CRN butterfly deltas and a unified competitive TURF; Sawtooth's desktop simulator often requires Excel export or a paid web-hosting add-on.
AI and fieldwork QA: Ordeen includes a brief-to-market LLM wizard, World Bank quota matching, the Peitz & Olsen warm-up game and AI open-end fraud cleaning; Sawtooth is 100% manual setup with no built-in panel or AI brief ingestion.
At a glance
Sawtooth is the better choice if:
Your team already has years of Lighthouse Studio studies, files and training, and needs continuity with past work.
You need surveys in several languages, the randomized first choice simulation rule, or a standalone Excel simulator to hand to clients.
Your procurement policy requires SOC 2 Type 2 today. Sawtooth holds SOC 2 Type 2 (not ISO 27001). Ordeen has had two independent security reviews (April and October 2026); SOC 2 and ISO 27001 certification is in progress.
Ordeen is the better choice if:
You want CBC, ACBC, MaxDiff, MBC, Van Westendorp and Gabor-Granger in one browser platform, with no desktop install and no separate modules. Discover covers CBC and MaxDiff; Lighthouse Studio is desktop software and sells MBC as a separate product.
You want a respondent panel (Prodege) inside the tool, automatic checks on open-ended answers, and a first draft of the competitor market prepared from your brief.
You want multi-chain Hamiltonian Monte Carlo estimation with a convergence check on every result, Shapley attribute importance, and interactions added automatically.
You want flexible billing (£35,000/year annual or £10,000/quarter), 5 users included, with the integrated respondent panel priced at £3.00 to £4.00 per complete rather than a desktop licence that doesn't cover panel costs.
1. Methods
Sawtooth spreads its methods across two products:
Discover (cloud): CBC, with nested attributes for conditional pricing, and MaxDiff.
Lighthouse Studio (desktop): CBC with conditional pricing tables, ACBC and MaxDiff; MBC as a separate product; question templates for Van Westendorp and Gabor-Granger.
Ordeen (cloud) runs all six methods in one workspace: CBC; ACBC (used automatically when a CBC design has more than seven attributes); MaxDiff (anchored by default); MBC; Van Westendorp, with bootstrap intervals and the Newton-Miller-Smith purchase-intent extension; and Gabor-Granger, with an adaptive price ladder and calibrated demand. CBC and ACBC support conditional pricing.
2. Bayesian estimation: two different samplers
Both products estimate individual preferences with Hierarchical Bayes, but with different algorithms:
Sawtooth CBC/HB uses Gibbs sampling with a Metropolis-Hastings step. To get a multi-chain convergence statistic (R-hat), an analyst runs the estimation several times with different random seeds and compares them.
Ordeen uses the No-U-Turn Sampler (NUTS, a form of Hamiltonian Monte Carlo) in NumPyro/JAX, multi-chain on GPU for CBC and ACBC, with weakly-informative priors chosen so the posterior is driven by the data.
Every Ordeen result reports r̂, ESS and divergence counts directly in the output; any fit that fails the convergence check is withheld rather than shown. We have not published a numerical comparison with Lighthouse Studio.
3. Attribute importance and interactions
Sawtooth's own Discover documentation notes the main weakness of range-based importance: "a price attribute ranging from $100 to $200 will appear less important than one ranging from $100 to $300." Ordeen measures importance with a Shapley decomposition of each attribute's contribution to predicted choice instead.
For interactions, Ordeen adds attribute pairs to the model automatically where the data supports them, and reports each with a 95% credible interval. Within segments, an empirical check with proper false-discovery control and small-cell corrections filters spurious interactions before anything reaches the output.
4. Simulation, willingness to pay and sharing
Sawtooth's desktop Choice Simulator offers logit, first choice and randomized first choice rules, standard errors on shares, and an Excel simulator export. Its WTP bootstrap (300 samples) is off by default. Discover has a logit simulator with optional 95% intervals, Segment Finder (latent class) and TURF.
Ordeen's simulator gives every share a 95% credible interval from the posterior draws, an elasticity sweep across the price range, a two-attribute heatmap and brand-portfolio TURF. Willingness to pay has a 95% interval calculated at every posterior draw, and is refused outright rather than shown when the price effect changes sign across draws. Shareable links are available for stakeholders to review the results; the shared report shows the findings but not the interactive simulator. Exports are CSV, Excel (XLSX), JSON and Parquet.
5. Pricing
Sawtooth sells annual subscriptions and quotes on request. A third-party listing (Capterra) gives about $10,900 a year per licence for Lighthouse Studio's advanced suite; respondents are bought separately.
Ordeen is sold by subscription in two flexible ways: £35,000/year on the standard annual plan, or £10,000 per quarter billed quarterly — 5 users included in each, with extra 5-user packs available at the tier rate. Respondent packs from the integrated Prodege panel are priced inside the platform at £3.00 to £4.00 per complete and never expire.
Related reading: What is Hierarchical Bayes? · CBC vs ACBC · Attribute importance: range vs Shapley
Sources
Sawtooth Software: CBC/HB technical documentation, Lighthouse Studio manual (Choice Simulator, conditional pricing, WTP), Discover help (CBC utilities, simulator), sawtoothsoftware.com/help (checked October 2026).
Sawtooth Software trust/security page (checked October 2026).
Capterra listing for Sawtooth Software (third-party, checked October 2026).
Sawtooth Software is a long-established conjoint vendor with a large, experienced user base. It sells two products: Lighthouse Studio, a Windows desktop application for advanced choice modelling, and Discover, a browser-based tool for simpler studies.
Ordeen Choice Modelling is a cloud platform that runs six pricing and choice methods, including adaptive CBC (ACBC) and menu-based choice (MBC), in one browser workspace with a respondent panel built in. Here is how it compares with both Sawtooth products
Side-by-side on every dimension
Annual software price: Ordeen £35,000/yr Annual or £10,000/qtr (5 users in each tier); Sawtooth ~$4,500–$15,000/yr per specialist seat on an annual lock-in.
Panel: Ordeen roughly £3–£4 per complete through the integrated Prodege panel in non-expiring packs (£4 at 500, £3.50 at 1,000, £3 at 2,500); Sawtooth bring-your-own or brokered separately at $4–$12+.
Methods: Ordeen ships all six (CBC, ACBC, MaxDiff, MBC, Van Westendorp, Gabor-Granger) plus nested-logit and latent-class, fit through one estimation pipeline; Sawtooth CBC, ACBC, MaxDiff, MBC, with higher-tier modules priced separately.
Estimation engine: Ordeen GPU-accelerated Hierarchical Bayes with convergence diagnostics (R-hat, ESS, divergences) on every result; any fit that fails the convergence check is withheld rather than shown; Sawtooth CPU Gibbs sampling with a Metropolis-Hastings step and user-run multi-chain comparisons.
Conditional pricing: Ordeen's engine keeps the experimental design orthogonal across tiers in log-price space and un-centers the draws back to raw currency across WTP, simulator, elasticity and TURF; Sawtooth supports CP but is prone to collinearity and convergence issues without manual coding.
Willingness to pay: Ordeen computes Jensen-free aggregate-first WTP at every posterior draw (Bayesian Krinsky-Robb) with Harrell-Davis credible intervals and refuses figures outright when the price slope changes sign across draws; Sawtooth reports point ratios or raw simulation draws, prone to ratio blowups.
Segmentation: Ordeen combines k-means++ on a signal-weighted feature representation with full Bayesian NUTS latent class and Gemini-guardrailed segment naming; Sawtooth offers CCEA / latent class (point-estimate EM) and HB K-means run manually by the analyst.
Simulator and TURF: Ordeen's simulator reports posterior-expectation shares with BBC credible intervals, sub-second elasticity sweeps, paired-CRN butterfly deltas and a unified competitive TURF; Sawtooth's desktop simulator often requires Excel export or a paid web-hosting add-on.
AI and fieldwork QA: Ordeen includes a brief-to-market LLM wizard, World Bank quota matching, the Peitz & Olsen warm-up game and AI open-end fraud cleaning; Sawtooth is 100% manual setup with no built-in panel or AI brief ingestion.
At a glance
Sawtooth is the better choice if:
Your team already has years of Lighthouse Studio studies, files and training, and needs continuity with past work.
You need surveys in several languages, the randomized first choice simulation rule, or a standalone Excel simulator to hand to clients.
Your procurement policy requires SOC 2 Type 2 today. Sawtooth holds SOC 2 Type 2 (not ISO 27001). Ordeen has had two independent security reviews (April and October 2026); SOC 2 and ISO 27001 certification is in progress.
Ordeen is the better choice if:
You want CBC, ACBC, MaxDiff, MBC, Van Westendorp and Gabor-Granger in one browser platform, with no desktop install and no separate modules. Discover covers CBC and MaxDiff; Lighthouse Studio is desktop software and sells MBC as a separate product.
You want a respondent panel (Prodege) inside the tool, automatic checks on open-ended answers, and a first draft of the competitor market prepared from your brief.
You want multi-chain Hamiltonian Monte Carlo estimation with a convergence check on every result, Shapley attribute importance, and interactions added automatically.
You want flexible billing (£35,000/year annual or £10,000/quarter), 5 users included, with the integrated respondent panel priced at £3.00 to £4.00 per complete rather than a desktop licence that doesn't cover panel costs.
1. Methods
Sawtooth spreads its methods across two products:
Discover (cloud): CBC, with nested attributes for conditional pricing, and MaxDiff.
Lighthouse Studio (desktop): CBC with conditional pricing tables, ACBC and MaxDiff; MBC as a separate product; question templates for Van Westendorp and Gabor-Granger.
Ordeen (cloud) runs all six methods in one workspace: CBC; ACBC (used automatically when a CBC design has more than seven attributes); MaxDiff (anchored by default); MBC; Van Westendorp, with bootstrap intervals and the Newton-Miller-Smith purchase-intent extension; and Gabor-Granger, with an adaptive price ladder and calibrated demand. CBC and ACBC support conditional pricing.
2. Bayesian estimation: two different samplers
Both products estimate individual preferences with Hierarchical Bayes, but with different algorithms:
Sawtooth CBC/HB uses Gibbs sampling with a Metropolis-Hastings step. To get a multi-chain convergence statistic (R-hat), an analyst runs the estimation several times with different random seeds and compares them.
Ordeen uses the No-U-Turn Sampler (NUTS, a form of Hamiltonian Monte Carlo) in NumPyro/JAX, multi-chain on GPU for CBC and ACBC, with weakly-informative priors chosen so the posterior is driven by the data.
Every Ordeen result reports r̂, ESS and divergence counts directly in the output; any fit that fails the convergence check is withheld rather than shown. We have not published a numerical comparison with Lighthouse Studio.
3. Attribute importance and interactions
Sawtooth's own Discover documentation notes the main weakness of range-based importance: "a price attribute ranging from $100 to $200 will appear less important than one ranging from $100 to $300." Ordeen measures importance with a Shapley decomposition of each attribute's contribution to predicted choice instead.
For interactions, Ordeen adds attribute pairs to the model automatically where the data supports them, and reports each with a 95% credible interval. Within segments, an empirical check with proper false-discovery control and small-cell corrections filters spurious interactions before anything reaches the output.
4. Simulation, willingness to pay and sharing
Sawtooth's desktop Choice Simulator offers logit, first choice and randomized first choice rules, standard errors on shares, and an Excel simulator export. Its WTP bootstrap (300 samples) is off by default. Discover has a logit simulator with optional 95% intervals, Segment Finder (latent class) and TURF.
Ordeen's simulator gives every share a 95% credible interval from the posterior draws, an elasticity sweep across the price range, a two-attribute heatmap and brand-portfolio TURF. Willingness to pay has a 95% interval calculated at every posterior draw, and is refused outright rather than shown when the price effect changes sign across draws. Shareable links are available for stakeholders to review the results; the shared report shows the findings but not the interactive simulator. Exports are CSV, Excel (XLSX), JSON and Parquet.
5. Pricing
Sawtooth sells annual subscriptions and quotes on request. A third-party listing (Capterra) gives about $10,900 a year per licence for Lighthouse Studio's advanced suite; respondents are bought separately.
Ordeen is sold by subscription in two flexible ways: £35,000/year on the standard annual plan, or £10,000 per quarter billed quarterly — 5 users included in each, with extra 5-user packs available at the tier rate. Respondent packs from the integrated Prodege panel are priced inside the platform at £3.00 to £4.00 per complete and never expire.
Related reading: What is Hierarchical Bayes? · CBC vs ACBC · Attribute importance: range vs Shapley
Sources
Sawtooth Software: CBC/HB technical documentation, Lighthouse Studio manual (Choice Simulator, conditional pricing, WTP), Discover help (CBC utilities, simulator), sawtoothsoftware.com/help (checked October 2026).
Sawtooth Software trust/security page (checked October 2026).
Capterra listing for Sawtooth Software (third-party, checked October 2026).