Skari

Data Analysis

From statistics and modeling to finance and marketing analytics — all in one place.

StatisticsFrom hypothesis testing to time series and clustering — a wide range of statistical methods.SEMType the research topic — the model, the survey and a report in thesis order follow.Auto ModelingTrains and compares several machine-learning models automatically, and explains why they predicted what they did.

Specialized Analysis

Quality

Survey

Go beyond standard surveys with Conjoint, AHP, and Price Sensitivity — plus CSAT, NPS, and more.

Explore Survey

ConjointEstimates preference share and market share by attribute combination.AHP PriorityDerives priority ranking for each item through pairwise comparison.Price Sensitivity (PSM)Analyzes the optimal price point and resistance range from responses.

Other Methods

Gabor-Granger Kano Model Driver Analysis MaxDiff TURF Usability (SUS)NPS Brand Tracking View All Methods

Choice-based conjoint and trade-off decisions

CBC Conjoint

Understand every customer choice.

Present realistic product alternatives and measure how customers trade features, brands, and prices. Then compare new product concepts before you launch.

Attribute ImportancePart-worth UtilitiesMarket SimulatorBootstrap StabilityIndividual UtilitiesSensitivity Analysis

Try Free See PSM Pricing

OLS

Pooled + Individual

BTL

Market Simulator

400×

Bootstrap Resamples

R²

Model Fit

CONJOINT METHODS

Choose the right conjoint method

From classic CBC to adaptive tasks, menu building, and MaxDiff prioritization — each method extracts preference in a different way.

CBCMost popular

Choice-Based Conjoint

Respondents choose one option from a set of full profiles. Closest to real purchase behavior.

ACBCHigh precision

Adaptive CBC

Task difficulty adapts to each respondent's preferences, increasing precision with fewer questions.

MBCBundling

Menu-Based Conjoint

Respondents build their own bundle from a menu of options — ideal for configurable products.

MaxDiffFeature prioritization

Maximum Difference Scaling

Respondents pick best and worst items from small sets — excellent for prioritizing long feature lists.

Ratings do not reveal trade-offs

01Customers rate every feature as important, so the results do not reveal a real priority.

02You know which feature is preferred, but not what customers would give up to get it.

03Price decisions are being made without knowing how customers trade price against value.

04There is no reliable way to compare product concepts before launch.

CBC shows the trade-offs behind each choice and how preference changes across product concepts.

FROM SETUP TO FIELDWORK

Build the attributes. See the exact task respondents get.

No separate preview step — every attribute and level you add updates the task on the right immediately.

createsurvey · cbc-conjoint

CBC Conjoint Analysis

ConjointCustom tasks · min. 200 respondents

Attributes (min. 2)

3 attributes

Each attribute is one characteristic that varies across product bundles (e.g. "Price"). Each level below it is a value that characteristic can take (e.g. "$10", "$20"). Add at least 2 attributes, each with at least 2 levels.

1BrandDescription (optional)×

Levels

1Brand A×

2Brand B×

3Brand C×

+ Add level

2PriceDescription (optional)×

Levels

1$10×

2$20×

+ Add level

3QualityDescription (optional)×

Levels

1Standard×

2Premium×

+ Add level

Apply to survey

Live

Live preview

Task 1 of 8

Which option do you prefer?

Option 1

BRAND

Brand B

PRICE

$30

QUALITY

Standard

Option 2

BRAND

Brand A

PRICE

$20

QUALITY

Premium

Option 3

BRAND

Brand C

PRICE

$10

QUALITY

Premium

None of these

‹ Previous

Next ›

Left: the attribute builder every CBC study starts from. Right: exactly what respondents see — same labels, same order, updated as you type.

FROM RESPONSES TO RESULTS

Choice utilities and product scenarios

Every figure below comes from one real example study — a laptop choice survey of 30 respondents, fit through the same engine your own data runs on.

Choice task design

Enter attributes and levels; balanced choice tasks are generated automatically, varied per respondent and checked for design efficiency.

Choice task 4 / 123 attributes · 9 levels

Brand A₩1,000,000512GB

Chosen

Brand B₩1,500,0001TB

Choose

Brand C₩2,000,000256GB

Choose

4 / 12

Attribute importance

Decompose the choice into percentage contributions to compare the relative importance of each attribute.

Relative importanceTotal 100%

Brand

43.8%

Price

42.5%

Storage

13.7%

Interpretation

Brand drives 43.8% of the choice — the single strongest factor.

Part-worth utilities and WTP

How each level ranks inside its attribute — and what that gap is worth in money.

Part-worth · BrandImportance 43.8%

Brand A (ref)

Brand B

−2.0

Brand C

−3.9

1TB (ref)

512GB

+1.2

WTP · Brand C−₩1,026,186

Product and price simulation

Enter concepts and the logit model returns preference share, so you can compare price options before launch.

Share of preference3 concepts

Brand B · ₩1,000,000 · 256GB71%

Brand A · ₩2,000,000 · 512GB28%

Brand C · ₩1,500,000 · 1TB1%

* Logit-based share estimate assuming a market of only the entered concepts.

Preference segments

Latent class analysis identifies customer groups with distinct preference structures.

Segment insightLatent Class

36.7%

Segment 1

Driven mainly by price (36%); prefers ₩1,000,000.

All attributes ›

50%

Segment 2

Driven mainly by brand (36.2%); prefers Brand A.

All attributes ›

13.3%

Segment 3

Driven mainly by price (48.3%); prefers ₩1,500,000.

All attributes ›

Model fit and interpretation

Holdout hit rate and McFadden R² state the reliability up front, then the report generates the practical implications as sentences.

Model fit & conclusionsConditional Logit (MLE)

Respondents30

Observations720

Parameters6

McFadden R²0.515

Hit rate78.3%

1Brand drives 43.8% of the choice — the strongest factor.

2Brand A is preferred over Brand C by 3.9 points.

3Price (42.5%) is the second factor.

4The top concept wins 73.8x the share of the last.

A COMPLETE REPORT

See the full report from summary to conclusion.

The six cards above are fragments of this. Scroll the real output end to end, right here on the page.

CBC example — laptop choice · 30 simulated respondents

Executive Summary

30 respondents · 3 attributes

This CBC study estimated choice data from 30 respondents using a Conditional Logit (MLE) model, with a holdout hit rate of 78.3%. Among the 3 attributes, Brand (importance 43.8%) most heavily drives choice, and preference drops sharply to -3.9 at the "Brand C" level. The optimal combination (Brand A · ₩1,000,000 · 512GB) shows a 71% share of preference in the market simulation, up to 73.8× higher than competing alternatives.

Study Design & Model Fit

Sample, model, and the holdout hit rate — check these before reading anything else below.

Study Design & Model FitMcFadden R² 0.515

Metric	Value
Respondents	30
Observations	720
Attribute composition	Brand · Price · Storage capacity
Estimator	Conditional Logit (MLE) (converged)
Alternatives per task	3 (None 0%)
Parameters	6
McFadden R²	0.515
Hit rate vs. chance	33.3% → 78.3%

Relative Attribute Importance

The share each attribute holds in the overall choice decision.

Relative importancesums to 100%

Interpretation

Brand is the single largest factor, accounting for 43.8% of the choice decision.

Part-Worth Utility by Level

Relative preference among levels within each attribute. Right of center (blue) = preferred over the attribute average; left (orange) = less preferred.

Part-worth utilities3 attributes

Brand

importance 43.8%

Brand A (ref)

Brand B

−2*

Brand C

−3.9*

Price

importance 42.5%

₩1,000,000 (ref)

₩1,500,000

−2.1*

₩2,000,000

−3.7*

Storage capacity

importance 13.7%

1TB (ref)

512GB

+1.2*

256GB

+0.4

Willingness to Pay (WTP)

Each level converted to a monetary value relative to the reference level, derived from the price coefficient.

WTP · by level₩1,026,186

Brand

Brand B

Avoid ₩534,649

Brand C

Avoid ₩1,026,186

Storage capacity

512GB

Prefer ₩315,661

256GB

Prefer ₩110,534

Optimal Concept & Market Simulation

The combination with the highest total part-worth, and (when a scenario was run) the simulated share of preference among the entered concepts.

Optimal concept & share simulation71%

OPTIMAL CONCEPT

BrandBrand A

Price₩1,000,000

Storage capacity512GB

+1.2Total utility index (highest among given levels)

Share of Preference Simulation

Brand B · ₩1,000,000 · 256GB71%

Brand A · ₩2,000,000 · 512GB28%

Brand C · ₩1,500,000 · 1TB1%

※ Logit-rule share estimate assuming a market of only the entered concepts.

Segment Insights

Consumer segments identified by post-hoc latent class analysis. Click a segment to see its full attribute breakdown.

Segment insightLatent Class

36.7%

Segment 1

Driven mainly by Price (36%), preferring "₩1,000,000".

View all attributes

50%

Segment 2

Driven mainly by Brand (36.2%), preferring "Brand A".

View all attributes

13.3%

Segment 3

Driven mainly by Price (48.3%), preferring "₩1,500,000".

View all attributes

Conclusions & Recommendations

Action points generated directly from the analysis results above.

Conclusions & recommendations7 points

Brand is the single largest factor, accounting for 43.8% of the choice decision.

Within Brand, "Brand A" is preferred over "Brand C" by 3.9 utility points — the widest swing in the model.

"Brand C" (Brand) carries the largest price impact in the model — respondents prefer it less than the reference level ("Brand A") by ₩1,026,186.

Price (42.5%) is the second driver — its swing is smaller than Brand but still shifts choice meaningfully.

The optimal combination (Brand A · ₩1,000,000 · 512GB) reaches a total utility index of +1.2.

In the simulated market, the leading concept captures 73.8× the preference share of the weakest one.

The largest segment (50% of respondents) is driven mainly by Brand, preferring "Brand A".

Methodology — Part-worths are zero-centered within each attribute; importance is the normalized range of each attribute's utilities. Market simulation uses the logit rule for share of preference.

Scroll to explore the full report

IN THE REPORT

The value behind each choice

From attribute importance and level utilities to product simulation, every result is organized for product and pricing decisions.

Attribute importance

Compare how strongly each attribute contributes to customer choice.

Part-worth utilities

See which level of each attribute adds or reduces preference.

Spider / radar view

View the balance of attribute importance at a glance.

Bootstrap stability

Check whether the importance ranking remains stable across repeated samples.

Coefficient table

Review utility estimates with standard errors, p-values, and confidence intervals.

Significance flags

Identify which estimated differences are statistically meaningful.

Market simulator

Build product scenarios and estimate their relative preference share.

Multi-product comparison

Compare several product concepts side by side as their specifications change.

Individual utilities

Explore how preferences vary across respondents and customer groups.

Sensitivity analysis

Find which product change produces the largest movement in preference.

Design around real choices

Compare product concepts, understand trade-offs, and test how preference changes before launch.

Get Started Free

Executive Summary

30 respondents · 3 attributes

Study Design & Model Fit

Sample, model, and the holdout hit rate — check these before reading anything else below.

Study Design & Model FitMcFadden R² 0.515

Metric	Value
Respondents	30
Observations	720
Attribute composition	Brand · Price · Storage capacity
Estimator	Conditional Logit (MLE) (converged)
Alternatives per task	3 (None 0%)
Parameters	6
McFadden R²	0.515
Hit rate vs. chance	33.3% → 78.3%