Basic Conjoint Analysis

Choice Analysis Full Version →

Understand how categorical product attributes drive consumer choice. Upload choice-based conjoint (CBC) data, estimate part-worth utilities, simulate market shares, and find optimal product configurations — all in pure JavaScript, right in your browser.

Privacy: All estimation runs entirely in your browser. Your data never leaves your device.

1 ✓

Upload Data

2 ✓

Configure

3 ✓

Estimate & Analyze

4 ✓

Simulate

OVERVIEW & WHAT THIS TOOL SOLVES

When you ask customers in a standard survey which product features they want, they rate every feature "high importance" and demand the lowest price—giving you no idea what to prioritize. Basic Choice-Based Conjoint (CBC) forces realistic trade-offs by presenting side-by-side product profiles and estimating individual-level part-worth utilities for categorical attributes right in your browser, so you can see which attributes truly drive choice and simulate market shares before launch.

🎯 Questions It Answers

Which product attribute (e.g., price tier, brand, feature level, or warranty) has the largest impact on customer choice?
Which specific level within each attribute adds the most preference utility over your chosen baseline reference option?
If we launch our configuration against competitors, what market share will each capture — and how many customers will simply walk away ("Would Not Buy Any")?

📂 What Data You Need

Choice-Task CSV: Long-format table with Respondent ID, Task ID, Alternative ID, categorical attribute columns, and a 0/1 Chosen indicator.
Optional "None" Alternative: Supports both forced-choice studies and tasks with a None ("Would not buy any") outside option.
Need Continuous Curves / WTP / K-Means? Switch to the Full Conjoint Tool for continuous/quadratic price, dollar WTP, segmentation, and profit simulation.

💡 What You'll Discover

Attribute Importance Ranking: Side-by-side comparison of Mean Individual Importance vs. Average-Utility Importance.
Part-Worth Utilities & Walk-Away Hurdle: Individual and sample-average preference utilities plus the ASC_None buy/no-buy threshold.
Interactive Market Simulator: Live multinomial logit share predictions showing both product switching and total category expansion/walk-away rates.

🔍 Key Nuances & Basic vs. Full Conjoint Comparison

Why the "None / Would Not Buy" Option Matters: Without an outside option, every simulated scenario forces 100% of respondents to buy something—even if every product on the shelf is overpriced. Including a None alternative estimates each respondent's walk-away utility threshold (ASC_None), letting you separate competitor share-stealing from category abandonment.
How Attribute Importance Is Calculated: Mean individual importance computes each respondent's utility range (max minus min level) per attribute and normalizes to 100% before averaging across the sample (matching the Full Conjoint Tool). We also show Importance of Average Utilities so you can spot when opposing tastes cancel out on average.
Reference Level Coding: Each attribute designates one level as the baseline reference (utility = 0), which you can change in Step 2. Numeric and currency levels are automatically sorted in natural ascending order.
Exact Parity with Full Conjoint: When given the same categorical specification, reference levels, and L2 regularization (λ = 0.1), this tool produces the exact same MNL utilities, importances, and simulated market shares as the Full Conjoint Tool.

📐 Under the Hood: Model Equations & Variable Definitions

1. Part-Worth Utility, Outside Option (ASC_None), Multinomial Logit Choice Probability & L2-Regularized Estimation

\[ U_{jt} = \beta_{\text{None}} \cdot \mathbb{1}[j = \text{None}] + \sum_{a \in \text{attributes}} \sum_{\ell \neq \text{ref}(a)} \beta_{a,\ell} \cdot \mathbb{1}[\text{level}_{jt}(a) = \ell], \qquad P(\text{choose } j \mid t) = \frac{\exp(U_{jt})}{\sum_{k} \exp(U_{kt})} \] \[ \max_{\boldsymbol{\beta}} \;\; \sum_{t} \sum_{j} y_{jt} \cdot \log P(\text{choose } j \mid t) \;-\; \lambda \lVert \boldsymbol{\beta}_{\text{attr}} \rVert^2 \]

💬 Read it in plain English: A product profile's total utility (U_jt) is the sum of the preference scores (β_a,ℓ) for its active attribute levels relative to each attribute's reference baseline (0), while the "None / Would Not Buy" option has utility β_None (ASC_None). The softmax formula converts utilities into choice probabilities, and L2 ridge regularization (λ) stabilizes attribute part-worths while leaving the walk-away intercept unpenalized.

U_jt

Total Profile Utility OUTPUT ⓘ

Overall attractiveness score of product alternative j in choice task t.

β_a,ℓ

Part-Worth Utility Coefficient ESTIMATED ⓘ

Preference weight assigned to level ℓ of attribute a relative to that attribute's reference level (0).

1[level(a)=ℓ], y_jt

Attribute Indicator & Observed Choice YOUR DATA ⓘ

Indicator equal to 1 when alternative j displays level ℓ, and observed choice y_jt ∈ {0, 1}.

P(choose j), λ

Predicted Choice Share & L2 Penalty INFERENCE ⓘ

Predicted probability of selecting profile j and L2 shrinkage parameter λ preventing extreme individual estimates.

Load a marketing use case:

Select a preset to auto-load realistic CBC study data, or upload your own CSV. The download button lets you inspect or modify the dataset in Excel before re-uploading.

Upload long-format CBC data where each row represents one alternative in one choice task for one respondent. Required columns: respondent_id, task_id, alternative_id, chosen (0/1), plus categorical attribute columns.

Drag & drop CBC file

CSV with respondent_id, task_id, alternative_id, chosen, and attribute columns

No file uploaded.

INPUTS & SETTINGS

VISUAL OUTPUT

Mean individual importance summarizes each person's utility ranges before averaging. Importance of average utilities measures the consensus profile; opposing preferences can cancel. Both depend on the tested levels and ranges.

Attribute	Mean individual (%)	Average utilities (%)

Higher bars indicate larger relative fitted utility ranges under the named aggregation method.

How to read this chart

Each pair of bars compares the mean individual importance with the importance of average utilities. Each series totals 100% when utility ranges are nonzero. Opposing tastes can make the two series disagree.

Height = relative fitted utility range under each aggregation method, not variance explained.
Tested levels matter: changing their ranges can change importance without changing respondents' preferences.
No universal cutoff makes an attribute negligible. Compare both summaries and check differences between respondents.

💡

Marketing Insight

A small importance value applies only to the tested ranges. Compare both summaries before judging priorities; a low average-utility result can hide strong disagreement.

Average utility for each attribute level relative to its baseline reference level (ref = 0). Compare within attributes to see which levels win.

How to read this chart

Each horizontal bar shows the average part-worth utility for one attribute level, grouped by attribute and color-coded.

Baseline-Referenced (Ref = 0): Your chosen reference level for each attribute sits at 0.00; positive bars add utility above that baseline and negative bars subtract utility.
Outside Option ("None" / ASC_None) as the Walk-Away Hurdle: In Baseline-Referenced mode, a product with all baseline (ref) levels has a total utility of 0.00. To beat walking away for the average respondent, a product configuration's part-worths must sum to more than ASC_None (U_product > ASC_None). If ASC_None is negative (e.g., -0.21), the all-baseline product (0.00) already clears the average hurdle, whereas configurations whose net utility drops below -0.21 will lose the average shopper to "would not buy." If ASC_None is positive, the all-baseline product falls short on its own and needs positive feature upgrades to cross the hurdle.
Zero-Centered (Mean = 0): Shifts each attribute's levels so their mean is 0.00 (useful for seeing above-average vs. below-average levels without changing the utility spread or market shares).
±1 SD Whiskers: Toggle on the heterogeneity whiskers to see where respondents strongly disagree (e.g., polarized tastes on spice, ad load, or walk-away threshold).

📊

Interpretation Guide: Scoring Configs Against the Walk-Away Hurdle

To evaluate any product configuration, sum the part-worth utilities of its chosen levels across all attributes (starting from 0.00 for the baseline levels). When an outside option is included, compare that sum against ASC_None: clearing ASC_None means the average utility favors buying over walking away (in a logit model, U_product = ASC_None is the 50/50 indifference point, and individual respondents vary around the mean by ±1 SD).

Select attribute to inspect:

Distribution of respondent part-worth utilities for the selected attribute.

How to read this chart

While the full Part-Worth bar chart above summarizes averages across all attributes at once, this single-attribute view opens up the full distribution across every respondent in your sample.

Violin + Respondent Dots: The shaded violin curve shows where respondents cluster (one peak = consensus; two peaks = polarized segments). The inner box marks the 25th–75th percentile (IQR) with the median and mean, and each jittered dot is one respondent.
Respondent Trajectories: Connects each respondent's utilities across the levels of this attribute, colored by which level is that respondent's personal #1 favorite. Parallel slopes mean respondents agree on the ranking; crisscrossing "X" lines reveal opposing taste segments.
Zero-Centered (Mean = 0) vs. Baseline-Referenced (Ref = 0): In Zero-Centered mode, each respondent's utilities within this attribute are centered around 0.00 so every level (including the reference level) displays its own full distribution. Switch to Baseline-Referenced to anchor the reference level at 0.00 and see exact utility gains or losses relative to that baseline.
% Personal #1 Level (First-Choice Share): In the summary table below, compare Mean Utility against % Personal #1 Level. A polarized level can have a negative average utility yet still be the #1 favorite for a passionate 30–40% segment!

Level	Mean Utility	Median (IQR)	Std. Dev.	% > Baseline (Ref)	% Personal #1 Level
Select an attribute after estimation to view respondent-level distribution metrics.

MODEL RESULTS

Total respondents estimated: –

Mean tasks per respondent: –

Mean pseudo-R² (McFadden): –

Outside Option ("None"): –

Estimation time: –

Interpretation Aid

Pseudo-R²: Measure of model fit per respondent. Values 0.2–0.4 are typical for CBC studies; higher values indicate stronger predictive power.

Outside Option ("None"): Shows the sample mean of ASC_None (the utility of walking away vs. a product with all baseline reference levels) and the observed survey opt-out rate.

Tasks per respondent: More tasks (10–20) produce more stable coefficient estimates.

APA-Style Report

Estimate the model to see the APA-style statistical report.

Managerial Interpretation

Business-focused interpretation will appear here after estimation.

Aggregated Part-Worth Utilities

How to read colors & symbols: ★ Best Level Most-preferred level in attribute +0.380 Positive mean utility (preferred above baseline ref = 0.000) −0.510 Negative mean utility (less preferred than baseline ref = 0.000) Note on statistical significance vs. heterogeneity: Green and red colors indicate preference direction (sign) relative to baseline, not a frequentist p-value threshold. Because this model estimates a separate ridge-penalized utility vector for every respondent, Std. Dev. measures real taste heterogeneity across people rather than sampling standard error.

Attribute	Level	Mean Utility	Std. Dev.	Min	Max
Estimate the model to view aggregated part-worth utilities.

MARKET SIMULATION

Scenario Builder

Define competing product profiles using categorical attribute dropdowns. Label each profile as Our Product or Competitor to track both brand share and category walk-away ("Would Not Buy Any") rates.

Total addressable market size (customers)

Used to convert share percentages into expected buyer and walk-away customer counts.

Simulation Results

Total Category Buyers

–

Buy at least one offered product

Walk-Away ("Would Not Buy")

–

Opt out of the category

Our Product Share

–

Combined share of "Our Product" cards

Leading Option

–

Highest predicted share

Predicted preference share based on individual-level multinomial logit choice probabilities averaged across all respondents.

How to read this chart

Bars represent predicted preference share and sum to 100% across all active options (including "None / Would Not Buy" when enabled).
Blue = Our Products, Amber/Orange = Competitors, Slate Grey = None / Would Not Buy Any (walk-away).
Preference Share (Logit) vs. First-Choice Winners: Preference share uses smooth logit probabilities per respondent; the table also shows how many respondents rank each option as their strict #1 highest-utility choice.

💡

Marketing Insight: Cannibalization vs. Category Expansion

When you raise price or remove a popular feature, watch where the lost share goes: does it shift to a Competitor (brand switching) or into "None / Would Not Buy" (customers walking away from the market altogether)?

Option / Product	Role	Preference Share (%)	Expected Customers	First-Choice Winners (n / %)

Basic Conjoint Analysis

👨‍🏫 Professor Mode: Guided Learning Experience

OVERVIEW & WHAT THIS TOOL SOLVES

🎯 Questions It Answers

📂 What Data You Need

💡 What You'll Discover

DATA SOURCE

Use a Case Study

Upload Your Data

INPUTS & SETTINGS

Column Mapping

Step 2: Configure Attributes, Reference Levels & "None" Option

Step 3: Estimate Individual Utilities ?
Regularized MNL
Estimates individual-level preferences using L2-penalized maximum likelihood. Matches the mathematical formulation of the Full Conjoint Tool.

VISUAL OUTPUT

Attribute Importance: Two Summaries

Part-Worth Utilities

Respondent Preference Heterogeneity (By Attribute)

MODEL RESULTS

APA-Style Report

Managerial Interpretation

Aggregated Part-Worth Utilities

MARKET SIMULATION

Scenario Builder

Simulation Results

PRODUCT OPTIMIZATION

Brute-Force Configuration Search

Select attributes to vary

Top 10 Configurations

👨‍🏫 Professor Mode: Guided Learning Experience

OVERVIEW & WHAT THIS TOOL SOLVES

🎯 Questions It Answers

📂 What Data You Need

💡 What You'll Discover

DATA SOURCE

Use a Case Study

Upload Your Data

INPUTS & SETTINGS

Column Mapping

Step 2: Configure Attributes, Reference Levels & "None" Option

Step 3: Estimate Individual Utilities ? Regularized MNL Estimates individual-level preferences using L2-penalized maximum likelihood. Matches the mathematical formulation of the Full Conjoint Tool.

VISUAL OUTPUT

Attribute Importance: Two Summaries

Part-Worth Utilities

Respondent Preference Heterogeneity (By Attribute)

MODEL RESULTS

APA-Style Report

Managerial Interpretation

Aggregated Part-Worth Utilities

MARKET SIMULATION

Scenario Builder

Simulation Results

PRODUCT OPTIMIZATION

Brute-Force Configuration Search

Select attributes to vary

Top 10 Configurations

Step 3: Estimate Individual Utilities ?
Regularized MNL
Estimates individual-level preferences using L2-penalized maximum likelihood. Matches the mathematical formulation of the Full Conjoint Tool.