ORDEEN · Choice Modelling · 3 MIN READ

What is MaxDiff? Best-Worst Scaling for Prioritisation

MaxDiff (best-worst scaling) is a survey technique for prioritising a long list of items. Cleaner differentiation than rating scales.

Book a 20-min demo What is MaxDiff? Best-Worst Scaling for Prioritisation

MaxDiff (short for Maximum Difference Scaling, also known as best-worst scaling) is a survey research technique for forcing respondents to prioritise a long list of items — features, messages, needs, concerns — against each other. It produces a cleaner, more differentiated ranking than traditional rating scales.

Ordeen runs MaxDiff studies with hierarchical Bayesian item scores, multi-chain convergence checks (r̂, ESS, divergences) and 95% posterior credible intervals on every item rank. See how it fits in the full platform in our methodology.

How it works

Rather than ask respondents to rate every item individually (which produces compressed, undifferentiated scores), MaxDiff shows small sets (typically 4–5 items) and asks respondents to pick the best and worst. By rotating items across many sets, every respondent effectively ranks every item against every other item.

Each best-worst answer is decomposed into a rank-ordered logit:

$$P(\text{best}=b, \text{worst}=w \mid S) = \frac{e^{u_b}}{\sum_{i \in S} e^{u_i}} \cdot \frac{e^{-u_w}}{\sum_{i \in S \setminus \{b\}} e^{-u_i}}$$

Item utilities are estimated hierarchically (so each respondent's priorities are preserved). Each respondent's utilities are then centred to sum to zero across items, which fixes the otherwise arbitrary additive constant without changing any choice probability.

Why MaxDiff beats rating scales

  • No scale-use bias — can't be gamed by respondents who rate everything high or low

  • Forced trade-offs — reveals genuine priorities rather than enthusiasm

  • Comparable across cultures — avoids differences in rating-scale usage patterns across geographies

  • Discriminates between items — even items that would all score 8/10 on a scale differ clearly on MaxDiff

Anchored MaxDiff

Standard MaxDiff only tells you relative importance — you know Feature A is more important than B, but not whether B is “important” in absolute terms. Anchored MaxDiff adds a per-respondent threshold so the scale becomes “important vs not important” rather than purely relative. This is what makes money-denominated or classification results possible. Ordeen's MaxDiff is anchored by default, and its results include reach and TURF for the item list.

When to use MaxDiff

  • Prioritising a long list of product features for a roadmap

  • Testing messaging claims to find the most persuasive

  • Needs-states research — which customer problems matter most

  • Benefit prioritisation — which benefits should headline a product

  • Brand attribute importance — what matters most about a brand

Items on the list should be at the same conceptual level (don't mix “low price” with “feature X”) and ideally number 10–30 (fewer makes the exercise trivial, more fatigues respondents).

MaxDiff vs Conjoint

Conjoint tests trade-offs between attribute levels within a product (e.g. Brand A vs Brand B, £10 vs £15). MaxDiff tests trade-offs between items at the same level (e.g. Feature A vs Feature B vs Feature C). Use conjoint for pricing and attribute-level decisions, MaxDiff for prioritising items on a flat list.

Related terms

Conjoint analysis · Hierarchical Bayes · Full methodology

MaxDiff (short for Maximum Difference Scaling, also known as best-worst scaling) is a survey research technique for forcing respondents to prioritise a long list of items — features, messages, needs, concerns — against each other. It produces a cleaner, more differentiated ranking than traditional rating scales.

Ordeen runs MaxDiff studies with hierarchical Bayesian item scores, multi-chain convergence checks (r̂, ESS, divergences) and 95% posterior credible intervals on every item rank. See how it fits in the full platform in our methodology.

How it works

Rather than ask respondents to rate every item individually (which produces compressed, undifferentiated scores), MaxDiff shows small sets (typically 4–5 items) and asks respondents to pick the best and worst. By rotating items across many sets, every respondent effectively ranks every item against every other item.

Each best-worst answer is decomposed into a rank-ordered logit:

$$P(\text{best}=b, \text{worst}=w \mid S) = \frac{e^{u_b}}{\sum_{i \in S} e^{u_i}} \cdot \frac{e^{-u_w}}{\sum_{i \in S \setminus \{b\}} e^{-u_i}}$$

Item utilities are estimated hierarchically (so each respondent's priorities are preserved). Each respondent's utilities are then centred to sum to zero across items, which fixes the otherwise arbitrary additive constant without changing any choice probability.

Why MaxDiff beats rating scales

  • No scale-use bias — can't be gamed by respondents who rate everything high or low

  • Forced trade-offs — reveals genuine priorities rather than enthusiasm

  • Comparable across cultures — avoids differences in rating-scale usage patterns across geographies

  • Discriminates between items — even items that would all score 8/10 on a scale differ clearly on MaxDiff

Anchored MaxDiff

Standard MaxDiff only tells you relative importance — you know Feature A is more important than B, but not whether B is “important” in absolute terms. Anchored MaxDiff adds a per-respondent threshold so the scale becomes “important vs not important” rather than purely relative. This is what makes money-denominated or classification results possible. Ordeen's MaxDiff is anchored by default, and its results include reach and TURF for the item list.

When to use MaxDiff

  • Prioritising a long list of product features for a roadmap

  • Testing messaging claims to find the most persuasive

  • Needs-states research — which customer problems matter most

  • Benefit prioritisation — which benefits should headline a product

  • Brand attribute importance — what matters most about a brand

Items on the list should be at the same conceptual level (don't mix “low price” with “feature X”) and ideally number 10–30 (fewer makes the exercise trivial, more fatigues respondents).

MaxDiff vs Conjoint

Conjoint tests trade-offs between attribute levels within a product (e.g. Brand A vs Brand B, £10 vs £15). MaxDiff tests trade-offs between items at the same level (e.g. Feature A vs Feature B vs Feature C). Use conjoint for pricing and attribute-level decisions, MaxDiff for prioritising items on a flat list.

Related terms

Conjoint analysis · Hierarchical Bayes · Full methodology

Related guides

Contract Analysis

Jurisdiction-Grounded AI Contract Review & Risk

Why generic chatbots hallucinate legal contract risk, and how Ordeen grounds contract review, portfolio risk scoring and renewal drafting in governing law.

Read the guide
Forecasting

Intermittent Demand Forecasting: Croston, TSB & MinT

How to forecast lumpy, sparse SKU demand and reconcile hierarchies: Croston's method, SBA, TSB, MinT reconciliation and out-of-sample AI signal validation.

Read the guide
Price Monitoring

Retail Checkout Price Monitoring vs Sticker Scrapers

Why basic retail price scrapers miss real checkout undercutting: how delivery thresholds, variant SKU matching and screenshot evidence protect brand RRP.

Read the guide