Worked Example: The GEICO Skewness Loop

This example shows one disciplined approach to end-to-end consumer research and the reasoning behind each step.

The method described here was developed from experience running consumer research at scale. The vehicle is the GEICO brief used in training; the principle applies to any brand and category.

The objective

You have a GEICO research brief. Before running anything, get clear on what you are trying to learn. In this example: what is stopping target segments from choosing or staying with GEICO, and what could change that?

Step 1: Run a survey

Start broad. Run a quantitative survey across your target segments to get a read on where the brand stands and how the category is playing out. Use the brand track for competitive benchmarking, or a custom survey if you have specific hypothesis questions.

Look at the distributions. Most of the findings will be expected. The interesting part is what is not expected.

Tools: Run Quantitative Research

Step 2: Find the high-skewness questions

In the survey results, look for questions where answers cluster heavily at one end of the scale: high skewness. These are the findings that are not just directional but emotionally charged. Something about that question or the category it touches is strongly felt by the audience.

High skewness is a signal, not an answer.

Tools: Read & Interpret Results

Step 3: Ask the AI twins “why”

Take the high-skewness questions into the Converse workspace. Ask the AI twins across multiple segments why they responded that way. Frame the question around the objective: “Here is what your segment said in a survey. What is the underlying reason?”

The twins will give you the qualitative reasoning that the survey could not surface: the emotional drivers, the unspoken assumptions, the barriers that never appear in a bar chart.

Collect the strongest insights: the “nuggets.” These are the specific, usable facts about what is driving the skewed result.

Tools: Converse with a Twin, Run Qualitative Research

Step 4: Put the nuggets through brainstorming

You now have a set of problems or tensions: why something is not working, what the consumer wants that they are not getting, what barrier exists. Take these into a brainstorming session.

Use one of the structured brainstorming methods available in the platform (Six Thinking Hats, Round Robin, or others) to generate possible solutions. The structured methods are familiar to corporate audiences and produce outputs that are easier to prioritise.

From the brainstorming output, identify the top three solutions that seem most viable and most likely to move the needle.

Tools: Run Qualitative Research → Innovation Research / Brainstorming, Methods

Step 5: Run a validation survey

Take the top three solutions back to the respondents. Run a second survey asking: “If we did X, would that change your behaviour?” Measure preference and intent for each option.

This survey answers whether the proposed solutions actually solve the problem, at scale, with statistical weight behind the answer.

Tools: Run Quantitative Research

Step 6: Qual follow-up for negative responses

Some respondents will reject the proposed solutions. That rejection is data. Run a qualitative session, a focused interview or small group, asking those segments “why not?”

Their answers will either reveal a flaw in the solution (something you can fix) or an uncontrollable constraint (a market condition or personal circumstance that cannot be changed). If it is controllable, revise and retest. If it is uncontrollable, document it and move on.

Tools: Run Qualitative Research, Converse with a Twin

The loop in summary

Understand the objective
  → Survey (what is happening?)
    → Find high-skewness questions (where is the signal?)
      → Ask twins why (what is driving it?)
        → Brainstorm solutions (what could we do?)
          → Validate with survey (would this work?)
            → Qual on negatives (why not, and can we fix it?)

This is not a one-way process. Any step may send you back to a previous one. A brainstorming output may reveal a hypothesis worth testing before you validate. A validation survey may surface a new high-skewness finding that opens a new research thread.

Stay connected to the objective at every step. Every piece of research you run should answer a question that matters for the brief, not just produce data.

What comes next

Build a Portfolio: set up the research context before running any of this
Create Segments — the 4 Methods: build the segments you need
Read & Interpret Results: read survey output at each stage
Methods: the full catalog of research methods used at each step