Comparison Grid

Fix the criteria and weights before anyone has a favorite, check that you are comparing like with like, then say what the comparison changes.

Practice, not research: a comparison grid is a simplified form of multi-criteria decision analysis, which has formal foundations (Keeney & Raiffa, 1993; Keeney, 2002); a simple grid is only as good as its criteria and weights. The like-for-like check is craft logic, and the 1 to 5 scale is a rule of thumb.

Use it when:

• You need to compare two to three companies, suppliers, tools or ways of working on the same terms.
• A team argues about which option is better, and each side uses different criteria.
• You want to benchmark yourself against competitors or the alternatives your customers use.

Skip it when:

• The comparison must end in one recommendation with costs, risks and success metrics: use 2.11 Options & Recommendations.
• You are comparing markets or countries for entry: use 7.11 Market Screening Table.

A comparison grid is the practitioner's simple form of multi-criteria decision analysis, a field with rigorous formal foundations (Keeney and Raiffa, 1993). The simplification is fair craft, but it puts all the weight on the two choices the formal theory handles most carefully: which criteria you compare on, and how much each counts. Keeney (2002) describes common mistakes in making value trade-offs, and a quick grid invites several of them. Three that the library watches for, as its own reading rather than a list from that paper: weights set in the abstract instead of against how far the options actually differ, overlapping criteria that count one virtue twice, and scores adjusted until the expected winner wins.

A separate piece of craft logic covers comparability. A grid can only conclude about like against like. Differences in size, market, maturity or time period do not forbid a comparison, but they limit what it can honestly claim. So Step 2 sets criteria and weights before any scoring, Step 3 turns differences into written caveats, and Step 5 asks what the comparison changes.

Say what the comparison is for and which entities you compare. Two or three is enough; a fourth usually adds noise (rule of thumb).

Strong answer: "Purpose: understand why small cafes switch away from us, to set our new cafe offer. Entities: our bakery, the rival bakery across town, the supermarket bake-off counter." Not: "Compare bakeries."

Purpose (the decision or question it feeds):

Entity 1:

Entity 2:

Entity 3:

Illustrative example: the business, the roles and all numbers are made up.

The Gothenburg bakery from 2.1 wants to know what former cafe customers buy instead (an open question from 2.10).

Step 1: Our bakery, the rival bakery across town, the supermarket bake-off counter.

Step 2: The owner and a driver set four criteria before looking at any data: smallest order (weight 5), price per loaf (4), delivery before 7 a.m. (3), range of breads (2).

Step 3: The supermarket does not deliver, so its delivery score is left out of the comparison and noted as a caveat. The rival's prices are from last autumn, also noted.

Step 4: Smallest order: 2, 3, 5. Price: 3, 3, 4. Delivery: 5, 4, not compared. Range: 5, 3, 2. Our bakery leads on delivery and range, and loses where the cafes said it hurts.

Step 5: Most important difference: order size. Shared by all: fresh bread every morning. Insight: cafes do not leave for a better bakery, they leave for smaller amounts. Action: the half-crate test in 2.11.