Repertory Grid Technique: A Complete Guide for Customer Research

Bottom line: The repertory grid technique is a structured interviewing method that surfaces the personal constructs — the bipolar dimensions of meaning — that people use to evaluate products, brands, and experiences. Instead of asking customers what they think and receiving rehearsed, top-of-mind answers, you present sets of items and ask how two are alike yet different from a third. The output is a rigorous, semi-quantitative map of how your market actually perceives your category. With a platform like Koji, you can run repertory-grid-style elicitation at a scale that was previously impossible with manual one-on-one sessions.

Most customer research fails in a subtle way: it captures the vocabulary the researcher brought to the conversation, not the vocabulary living in the customer's head. When you hand someone a satisfaction survey, you have already decided which attributes matter. The repertory grid flips this — it lets respondents reveal the criteria they care about, in their words, before you ever impose a framework.

Where the Method Comes From

The repertory grid was developed by clinical psychologist George Kelly as the practical instrument of his Personal Construct Theory, published in The Psychology of Personal Constructs (1955). Kelly argued that every person acts like a scientist, constantly building and testing a personal system of contrasts to make sense of the world. Two people can look at the same product and evaluate it against entirely different dimensions — one sees "premium vs. cheap," another sees "trustworthy vs. sketchy." The grid was designed to extract those dimensions without the researcher supplying them.

Since the 1960s the technique has migrated from clinical psychology into market research, UX, brand strategy, knowledge elicitation, and product management, precisely because it is one of the few methods that is both qualitative in depth and quantitative in structure.

The Three Building Blocks

Every repertory grid has three components:

1. Elements — the things being compared. These are concrete and drawn from the respondent's real experience: competing products, brands they have used, features, or scenarios (e.g., "the last time I onboarded a new tool"). Six to ten elements is typical.
2. Constructs — the bipolar dimensions of judgment the respondent uses to tell elements apart. A construct always has two poles: easy to set up ↔ needs an engineer, feels premium ↔ feels disposable. Constructs are elicited, never supplied.
3. Ratings — each element is rated against each construct, usually on a 1–5 or 1–7 scale. This produces the grid: elements as columns, constructs as rows, ratings in the cells.

Triadic Elicitation, Step by Step

The engine of the method is triadic elicitation. You show the respondent three elements at a time and ask a single, disciplined question:

"Think about these three. In what important way are two of them alike, and different from the third?"

Suppose the elements are three project-management tools the respondent has used. They might say, "Two of these keep everything in one place, but the third one made me juggle five tabs." You have just elicited a construct: consolidated ↔ fragmented. You record both poles, then move to the next triad. Each new combination pulls out a fresh dimension until the respondent stops generating new distinctions — the point of construct saturation.

The reason triads work is cognitive: comparison is far easier and less leading than introspection. Asking "what matters to you in a tool?" produces generic answers ("it should be easy to use"). Asking someone to explain why two specific tools group together forces them to articulate a real, discriminating criterion.

Laddering Up and Down

Once a construct surfaces, you deepen it with laddering:

• Laddering up asks why that construct matters ("Why is keeping everything in one place important to you?") and climbs toward values and goals — the customer's deeper motivation.
• Laddering down asks how you would recognize it ("What specifically makes a tool feel fragmented?") and descends toward concrete, observable attributes you can actually design for.

Laddering turns a flat list of constructs into a means-end chain that connects product features to the outcomes customers are hiring your product to achieve. (See the laddering technique guide for the full protocol.)

Building and Rating the Grid

After eliciting constructs, ask the respondent to rate every element on every construct's scale. A cell might read: Tool A = 2 on "consolidated ↔ fragmented" (closer to fragmented). The completed grid is a numeric fingerprint of that person's mental model — which products they see as similar, which constructs travel together, and where your product sits relative to competitors.

How to Analyze a Repertory Grid

Repertory grids support several levels of analysis:

• Content analysis of constructs — cluster the raw construct labels across respondents into themes. This alone often reveals evaluation criteria you were not measuring.
• Visual inspection — read the grid for patterns: which elements cluster, which constructs are correlated.
• Cluster and principal-component analysis — for larger samples, statistical routines group similar elements and similar constructs, producing a perceptual map of the category. This is where the repertory grid overlaps with perceptual mapping.

The distinctive payoff: you learn both what dimensions matter and where each brand sits on them — from the same instrument.

When to Use the Repertory Grid

Reach for this method when you need to:

• Understand how customers actually categorize and compare options in your market, before writing a survey or positioning statement.
• Build a differentiated brand or product positioning grounded in customer language.
• De-bias a concept test by eliciting evaluation criteria instead of assuming them.
• Elicit expert or power-user mental models for complex or technical products.

It is less suited to simple satisfaction tracking or when you already know precisely which attributes to measure — a structured survey is faster there.

Repertory Grid vs. Surveys and Focus Groups

A traditional survey asks customers to rate attributes you chose — so it can only confirm or deny hypotheses you already had. A focus group surfaces open language but is dominated by the loudest voice and is nearly impossible to quantify. The repertory grid is the rare method that captures respondent-generated language and yields structured, comparable ratings. Its historical drawback was cost: each grid required a trained interviewer and 45–90 minutes of one-on-one time, capping most studies at a handful of participants.

Running Repertory Grids at Scale with Koji

This is where an AI-native platform changes the economics. Koji's AI interviewer conducts the triadic elicitation conversationally — presenting element triads, asking the "alike vs. different" question, capturing both poles of each construct, and automatically laddering with follow-up questions to reach the underlying value, exactly as a skilled human moderator would. Because there is no moderator to schedule and no calendar to coordinate, you can field the same rigorous protocol with 50 or 500 respondents instead of 8.

Koji's six structured question types map cleanly onto the grid workflow:

• open_ended questions capture the elicited construct poles in the customer's own words, with AI follow-up probing for depth.
• ranking questions let respondents order elements by preference, an efficient companion to triadic sorting.
• scale questions capture the 1–5 or 1–7 element ratings that populate every grid cell as clean, chartable data.
• single_choice and multiple_choice questions confirm which constructs a respondent recognizes across the set.
• yes_no questions gate branches (e.g., "Have you used this product?") so each respondent only rates elements they actually know.

Because every rating is captured as structured data, Koji's automatic analysis clusters constructs across the whole sample and produces the perceptual map in a real-time report — collapsing what used to be weeks of manual transcription and coding into hours. Voice or text, no moderator, 24/7.

Worked Example: Positioning a New Analytics Tool

A product team wants to position a new analytics product. They load six elements (their tool plus five competitors the user has touched) and let Koji run triadic elicitation across 120 target buyers. Constructs that emerge include set-up-in-minutes ↔ needs-onboarding, answers-my-question ↔ dumps-raw-data, and trustworthy-numbers ↔ have-to-double-check. Ratings reveal that incumbents cluster tightly on "powerful but needs onboarding," leaving "answers-my-question in minutes" almost unoccupied. The team has just found their positioning — in customer language, backed by ratings from 120 people, not a hunch from a whiteboard.

Common Mistakes to Avoid

• Supplying constructs. The moment you offer "would you say it is easy or hard to use?" you have contaminated the elicitation. Let the respondent generate the poles.
• Abstract elements. Elements must be concrete things the respondent has genuinely experienced, or ratings become guesses.
• Skipping laddering. Without laddering you get surface attributes, not the values that actually drive choice.
• Too many elements. Beyond ~10 elements, triadic combinations explode and respondents fatigue. Koji's async format mitigates fatigue, but keep element sets tight.

Related Resources

• Structured Questions Guide — the six question types that turn grid ratings into chartable data
• Laddering Technique Guide — climb from constructs to underlying customer values
• Projective Techniques — other methods for surfacing what customers cannot say directly
• Perceptual Mapping Guide — turn construct ratings into a category positioning map
• In-Depth Interview Methodology — the broader craft of one-on-one qualitative research
• Coding Qualitative Data — cluster elicited constructs into themes across your sample