In brief
A useful commercial hypothesis connects an observed signal to an intervention, evaluation criteria, and a follow-up decision. The fictional example follows a decision brief after a demonstration and distinguishes observed usefulness from proven revenue impact.
A campaign report can show activity without explaining the reasoning behind it. To make the Revenue Harness cycle concrete, I will follow a commercial hypothesis through an entire decision process.
The scenario is fictional. It does not describe an EPIC customer, an experiment we conducted, or a proven result. Its purpose is to show which records and decisions I would require before calling an operation a learning process.
The signal: proposals stall after the demo
A software company notices that some opportunities stall after a demonstration. Sales attributes the problem to pricing. Marketing suspects that buyers still do not understand the product.
The agent's first task is to assemble authorized evidence: CRM records, meeting notes, and loss reasons. It separates the buyer's statements from the seller's interpretation. Empty fields remain empty instead of becoming generated explanations.
In this scenario, review reveals another possibility: buyers are asking for help explaining the proposal internally. The team considers whether the difficulty lies in preparing a decision rather than understanding the interface.
That is a signal worth investigating. It does not establish that price has stopped mattering.
The hypothesis: help the internal discussion
The team writes a bounded hypothesis: for opportunities with a confirmed problem and contact, a decision brief explaining impact, implementation effort, and offer limitations may help buyers conduct their internal evaluation.
The intervention would be to prepare that brief after the demonstration. The primary outcome would be progression to a previously defined commercial stage. Clarification requests, preparation time, and rework would help the team understand the process.
Revenue would still be tracked, but a closed deal would not immediately be attributed to the brief. Other factors could influence the outcome.
| Record | Definition for this example |
|---|---|
| Audience | Opportunities meeting agreed eligibility criteria |
| Intervention | A decision brief provided after the demonstration |
| Comparison | The usual process, where a valid comparison is feasible |
| Safeguards | No invented financial returns, implementation promises, or customer references |
| Follow-up decision | Keep, revise, or stop the intervention |
The design: agree how to observe before acting
A comparison needs to account for differences between opportunities. If only the most promising deals receive the brief, higher progression will not isolate its effect.
Where volume and conditions permit, the team can design random assignment between the intervention and the usual process, supported by someone with experimentation expertise. Eligibility criteria, metrics, and the analysis point should be agreed in advance. Where that is not feasible, the work remains exploratory observation with explicit limits on causal conclusions.
Microsoft's Experimentation Platform provides a public reference for organizing work around hypotheses and impact evaluation. The commercial workflow here is illustrative. It does not reproduce a Microsoft experiment or promise comparable results.
Authorization: review the customer-facing output
The commercial owner approves the brief's template and sources. The agent can organize information and prepare drafts. A person checks claims involving pricing, timing, implementation, and customer commitments before sending.
If a source conflicts with the current offer, the task stops at that point. The draft flags the uncertainty rather than hiding it behind a generic sentence.
The operation preserves which version was sent, by whom, and for which opportunity. Changes to the template do not erase the conditions of the previous round. That history matters when the team later interprets the results.
The outcome: separate observation from conclusion
Suppose review of the fictional pilot finds buyers using the brief in internal meetings, but the team cannot distinguish its effect on commercial progression. That is a possible and useful outcome: an indication of usefulness, without demonstrated revenue impact.
The appropriate decision might be to improve the material, address recurring questions, and plan another observation. Announcing that the agent increased sales would go beyond the evidence.
If the document creates confusion or promises that require correction, the team can stop the test. A well-supported stop also improves the next decision. It prevents a flawed intervention from becoming standard practice merely because it was automated.
Learning: record what changes
The final record should explain what was observed, what remains uncertain, and which action was chosen. It should link back to the documents and original hypothesis. The next agent receives the conditions under which the format may help, rather than an instruction to repeat it indiscriminately.
This is the cycle proposed in Context is king. How does it become revenue?: context informs action, and review updates context.
To start in your business, choose a commercial decision small enough to follow through. What evidence would make you change the next action, even if the outcome did not support the original hypothesis?
Questions and answers
Is this an EPIC customer case?
No. It is a fictional scenario explaining the records and decisions in a Revenue Harness cycle. It does not represent a customer, an experiment conducted, or a proven result.
Does an opportunity progressing prove the intervention worked?
No. Other factors may explain progression. Evaluation needs to consider differences between opportunities and the limits of the design used to observe outcomes.
What should a team record when results are inconclusive?
Record observations, uncertainty, the chosen next action, and supporting sources. Inconclusive results can guide a revision or another observation without supporting a revenue claim.