Matrix

See the decisions your marketing needs to support.

This edition maps participants against decision tasks to help identify questions, evidence needs and gaps worth investigating. The worked example is fictional. Completing a Matrix does not establish that its assumptions are correct.

A customer can want your product and still lack a basis for choosing it.

Engineering may need to establish technical fit. A business owner may need to understand the operating costs. An agent may have narrowed the options before either person began their evaluation.

Matrix helps you map participants against decision tasks and identify where your marketing could help.

Use it to investigate one buying situation, make uncertainties visible and choose a focused next step. It is a working diagnostic within Adaptive Marketing; completing the map does not establish that its assumptions are correct.

Start with one buying situation.

Choose a customer context, an offer and a decision you need to understand.

“Companies buying our software” is too broad. A more useful starting point is:

A growing software company is deciding whether to buy our support assistant, build one using an AI platform or continue with its current process.

Before building the map, record:

  • Customer outcome: What is the customer trying to accomplish?
  • Alternatives: What could they buy, build, adapt or leave unchanged?
  • Proposed value: Why might your offer be worth choosing?
  • Current uncertainty: What do you need to learn before changing your marketing?

If you are unsure whether the customer’s needs or alternatives have changed, begin with the market-change exercise. Matrix examines the buying decision within that context.

Try “What has changed in your market?” →

Map participants against decision tasks.

Matrix has two dimensions.

Participants are the people, teams or agents that shape the choice.

Decision tasks are the judgments and actions involved in discovering, evaluating, choosing and adopting an option.

Useful starting columns are:

Matrix decision tasks and what each covers
Task What it covers
Discover options Establish the need and find possible approaches.
Assess fit Examine requirements, constraints and suitability.
Validate evidence Check the basis for important claims and assumptions.
Authorize Establish who may approve or veto a choice, and within what limits; distinguish recommendations from authority.
Act Carry out the next step, implement the choice or assume responsibility for its use.

These are not fixed stages in a funnel. Tasks may overlap, repeat or happen in a different order. Adapt the columns when the situation requires it.

In each relevant cell, write the participant’s question, judgment or responsibility. Record actual authority separately from influence. In the Authorize column, distinguish recommend, approve and veto, and state the scope of each. A recommendation is not permission to spend; approval of technical fit is not necessarily approval of the purchase.

In the Act column, distinguish carrying out a task from executing within delegated limits. Record who granted that authority and its boundaries—for example, an agent permitted to place an order only after human approval and within an agreed spending limit. Do not infer authority from a participant’s involvement.

Use unknown where involvement has not been established. Use no involvement identified where you have investigated but found none, and retain the basis for that conclusion. Do not let an empty cell imply certainty.

Work through an example.

The following support-assistant purchase is fictional. It illustrates the exercise; it is not a customer case or evidence of results.

The support lead wants to reduce repetitive work while passing difficult cases to a human. Engineering needs to assess integration and consider whether an internal solution would be suitable. A business owner must decide whether the expected benefit warrants the commitment.

For this example, assume the customer has confirmed that the support lead uses an AI assistant to research options. It has no purchasing authority.

Fictional example: participants against decision tasks in a support-assistant purchase
Participant Discover options Assess fit Validate evidence Authorize Act
Support lead Which approaches could reduce repetitive work? Does the option suit our workflow and escalation needs? How does it handle representative questions? Recommend an option; no spending approval assumed Own the support workflow and supply evaluation cases
Engineering lead Buy, build or adapt existing tools? Does it fit our systems and operating constraints? Can we verify integration, escalation and failure handling? Approve technical fit within the agreed remit Implement and maintain the customer-side integration
Business owner Is this problem a priority for investment? Is the proposed approach a worthwhile use of resources? Are benefit, cost and effort assumptions credible? Approve spending Assign business ownership and review outcomes
Research assistant Find options under the support lead’s instructions Compare documented requirements Surface sources and missing information for human checking No purchasing authority Return findings to the support lead

The table makes several responsibilities visible behind the phrase “buy a support assistant.”

It also places the agent within a delegated task. The support lead remains responsible for deciding what to ask and what to verify.

A real investigation might identify security, procurement, legal or other participants. Add them when evidence shows their relevance; do not assume every purchase has the same structure.

Connect a cell to evidence.

Choose a consequential cell and examine what supports your understanding.

Suppose, within the fictional scenario, a buyer interview and a review of current materials suggest that engineering cannot assess integration and escalation responsibilities.

Fictional example: connecting one Matrix cell to evidence
Question Example finding
What must engineering establish? Whether the assistant can work with the existing support system and pass appropriate cases to a human
What is currently available? A feature page and demonstration
What remains unresolved? Supported interfaces, configuration ownership, implementation work and behaviour when the assistant cannot answer
What supports this diagnosis? The interview and documentation review within this fictional example
What remains unknown? Whether the concern recurs among similar buyers and whether documentation alone would resolve it

This is more actionable than “engineering needs more content.”

It identifies a decision, an uncertainty and a basis for further investigation. It also preserves the limits of what has been learned.

Give sources short labels and connect them to the relevant cell. For this fictional example: Engineering lead—Validate evidence: interview E1 and documentation review D1; not yet confirmed with other buyers. Keep the source details and collection dates in your supporting notes. Evidence for this cell does not establish the other responsibilities or assumptions in the grid.

Record the connections between participants.

The grid cannot capture every relationship. Add notes about important handoffs, dependencies and points where decisions are revisited.

In the example:

Engineering informs the business case. The cost assessment depends partly on implementation and maintenance effort.

The support lead informs technical evaluation. Engineering needs representative questions and escalation requirements to assess fit.

Agent-assisted research informs the shortlist. Missing or misinterpreted information may require manual investigation or affect which products receive consideration.

These connections have consequences for marketing. A technical guide, an evaluation task and a cost worksheet should use consistent assumptions about the offer and the customer’s responsibilities.

Where the internal alternative matters, use the buy-versus-build worksheet to examine both options on consistent terms.

Use the buy-versus-build comparison worksheet →

Choose one gap worth addressing.

Prioritize a gap that matters to the decision and is supported by enough evidence to justify the next step.

Before choosing, ask:

  • Could this change the decision? Identify the consequence of leaving the gap unresolved.
  • What supports it? Separate observations from assumptions and note where evidence is limited.
  • What is the smallest useful way to investigate it? Choose a proportionate next step.

That step may be an investigation rather than a new asset.

For the support-assistant example, a possible response is a focused integration guide and evaluation task covering:

  • Supported interfaces and configuration requirements.
  • Ownership of escalation rules and source documentation.
  • How to examine behaviour with representative support cases.
  • Known limitations and questions requiring further investigation.

Check whether the product can meet the requirement before treating it as a communication problem. Better documentation cannot resolve a capability the offer does not provide.

Assign an owner to the next step and record why it deserves attention.

Define what improvement would look like.

Before introducing the response, establish what happens today and what you expect to change.

Fictional example: defining what improvement would look like
Test element Example
Problem Engineering cannot assess integration and escalation responsibilities from existing materials.
Proposed change Provide a focused guide and evaluation task.
Expected improvement Evaluators can explain the requirements and reach a better-informed fit decision.
Starting point Record recurring questions, unresolved issues and assistance needed in comparable evaluations.
Evidence to examine What evaluators understand, which questions remain and how they explain their conclusions.
Costs and risks Preparation and support effort; possible confusion or unnecessary evaluation work.
Review point Agree when to review and which evaluations to include before introducing the change.
Next decision Retain, revise or stop the response; investigate a product gap; or test another explanation.

Fewer questions do not automatically mean better understanding. A faster decision does not automatically mean a better one. An evaluator may correctly conclude that the offer is unsuitable.

Ask what happened and why. Consider whether the evaluations are comparable and what other changes could explain the result. Limited evidence can inform a next step without proving a general rule.

Try it on your own business.

Copy these tables into a document or spreadsheet to complete the exercise. The tables on this page are not editable.

First, complete the context.

Worksheet: context fields for your buying situation
Field Your notes
Buying situation Customer context, offer and decision
Customer outcome What the customer needs to accomplish
Alternatives Other suppliers, internal development, existing tools or no change
Proposed value Why your offer could be worth choosing
Uncertainty What you need to learn before deciding what to change

Next, build a small participant-by-task grid. Use the starting tasks below or adapt them.

Worksheet: blank participant-by-task grid
Participant and responsibility Discover options Assess fit Validate evidence Authorize Act

Write questions or responsibilities in the relevant cells. Mark unknowns explicitly. For an agent, record who delegates the task and what authority has been granted.

Keep supporting notes alongside the grid:

Worksheet: supporting records to keep alongside the grid
Record What to capture
Sources Label each source, record when it was collected and reference it in the relevant cells; note what remains unconfirmed
Assumptions What you currently infer but have not established
Handoffs Where one participant relies on another’s findings or action
Priority gap One consequential uncertainty or unmet evidence need
Next step and owner The investigation or intervention, and who will carry it out
Outcome and baseline What should improve and what happens today
Review point When you will assess the evidence and decide what follows

Stop the first pass when you have identified one consequential gap, recorded the evidence supporting it and its limits, and agreed a next step, owner and review point. You do not need to fill every cell. If the evidence is too thin to justify a change, make gathering it your next step.

Keep the map small enough to use. Add detail when it changes a decision.

Connect Matrix to IMPACT.

Matrix supports the Identify and Map stages of IMPACT.

Its findings inform what evidence to Prove, which explanations need to Align and what activities to Coordinate. Test produces observations that may change the map.

The wider market remains relevant. If customers can build more themselves, revisit the alternatives. If they delegate more work to agents, revisit responsibilities and evidence needs. If a competitor changes its offer, reconsider the value you need to demonstrate.

Matrix helps organize that investigation. It does not replace customer conversations, market research or competitive judgment.

Improve the map—and the method.

A Matrix is a set of claims about how a buying decision works.

Correct it when a customer contradicts it. Add a missing participant when evidence warrants it. Remove detail that creates work without improving understanding.

The framework itself is also under development. Its usefulness depends on whether it helps you identify a meaningful gap and choose a proportionate response.

The output is a clearer decision about your marketing: what to investigate, what to change and how to learn whether it helped.

Explore IMPACT →
Read the Adaptive Marketing approach →

The agent on the buying committee

Revision history

Edition 1.0 — 15 November 2026
Initial diagnostic and exercise, including guidance on uncertainty, delegated authority and connections between participants.

Shared revision principle

We revise the method when evidence or practical experience warrants a change. For meaningful revisions, we explain what changed, why it changed and what readers using an earlier edition should reconsider. Earlier editions will remain available for reference.