Score by criterion across options, not by option across criteria — column-first prevents halo effects from inflating favorites

Why This Is a Rule

The halo effect causes a strong impression on one dimension to inflate scores on unrelated dimensions. When you score one option across all criteria (row-first), a high score on the first criterion creates a positive impression that unconsciously inflates subsequent criterion scores for that same option. "This apartment has an amazing kitchen" (score: 9) → "The neighborhood is pretty good too" (inflated from what would have been a 6 to an 8). The kitchen score creates a halo that brightens everything else about that apartment.

Scoring one criterion at a time across all options (column-first) defeats the halo effect by forcing apples-to-apples comparison. When evaluating "neighborhood quality" across all apartments simultaneously, you're comparing neighborhoods to neighborhoods — not comparing an apartment that already impressed you to one that didn't. The comparison context is the criterion, not the option, so no single option's strength can inflate its scores elsewhere.

This is the same principle used in structured interviews: score all candidates on question 1 before moving to question 2, rather than scoring one candidate on all questions and then the next. The format prevents earlier answers from creating a halo that biases later scoring.

When This Fires

• Every time you're scoring options in a decision matrix (Use a weighted decision matrix when options exceed 3 and criteria exceed 4 — working memory cannot hold all dimensions at once)
• When you notice yourself thinking "this option just feels better overall" during scoring — that's the halo effect
• During group evaluations where one person's enthusiasm for an option might bias the group's scoring
• When you want the matrix to reveal new information rather than confirm existing impressions

Common Failure Mode

Row-first scoring feels more natural: "Let me evaluate Apartment A on all dimensions, then Apartment B." This feels thorough but produces inflated scores for whichever option makes the best first impression. You score Apartment A across all criteria while still under the spell of its best feature, then score Apartment B across all criteria while comparing everything to Apartment A's halo.

The Protocol

(1) After assigning weights (Assign criterion weights before scoring options — knowing scores first lets you unconsciously rig the weights), score criterion by criterion, not option by option. (2) For criterion 1: evaluate all options on this criterion only. Which option is best on this dimension? Score them relative to each other. (3) Move to criterion 2. Forget how each option scored on criterion 1. Evaluate fresh on this new dimension. (4) Continue through all criteria. (5) Only after all criteria are scored → look at the totals. This is the first time you see each option's overall performance. (6) The column-first discipline feels slower, but it produces more accurate scores because each criterion is evaluated in a comparison context where the halo effect can't operate.