MMarketing Against The Grain
← All frameworks
Strategy

The Actionability Gate

Test every analysis for recommendations that trigger concrete action

Difficulty
Easy
Time to result
~days to results
Steps
5
Confidence
94%

The Actionability Gate evaluates analysis by working backward from execution. First identify the decision or business outcome at stake, then ask a question whose answer could materially affect that outcome. After the analysis is complete, require clear recommendations and test whether each recommendation leads to a specific action. If the result merely describes the data, refine the question or analysis until the decision path becomes visible. AI makes calculations, transformations, and visualizations dramatically faster, but that speed increases the importance of choosing meaningful questions. The framework therefore treats judgment—not data processing—as the scarce capability and uses actionability as the final quality test.

Origin

Extracted from Marketing Against The Grain, where Kieran Flanagan described how AI removes much of the analytical workload and makes question selection and actionability the critical remaining skills.

Core principles

  • 01Analysis is useful only when it changes a decision
  • 02Good questions connect evidence to action
  • 03Automation shifts the bottleneck from calculation to judgment
  • 04Recommendations must be specific enough to execute

How to run it

  1. 1

    Name the decision

    State the decision, change, or resource allocation the analysis is intended to support.

    Pro tip Phrase it as a choice between concrete alternatives.

    Watch out Do not begin with a dataset merely because it is available.

  2. 2

    Form an actionable question

    Ask a question whose answer would alter that decision or produce a meaningful next step.

    Pro tip Include the audience, outcome, and comparison in the question.

    Watch out Broad prompts often produce descriptive but unusable findings.

  3. 3

    Run the analysis

    Use AI or conventional analytical tools to inspect, transform, compare, and visualize the relevant evidence.

    Pro tip Ask the system to explain both its method and uncertainty.

    Watch out Fast output is not proof that the input data or method is sound.

  4. 4

    Demand recommendations

    Ask what the findings imply and request a short set of evidence-linked recommendations.

    Pro tip Require each recommendation to cite the finding that supports it.

    Watch out Generic best practices may not actually follow from the analysis.

  5. 5

    Convert recommendations into actions

    Assign a concrete action, owner, and success measure to every recommendation worth retaining.

    Pro tip Prefer a small experiment when the evidence is directional.

    Watch out If no action follows, revisit the original question rather than decorating the result.

In the wild

Choosing cross-platform podcast topics

The hosts compared podcast performance across RSS and YouTube. Discovering weak correlation exposed a strategic problem: they needed topics that could appeal to both audiences rather than optimizing blindly for either platform.

The analysis produced a concrete search for topics in the overlap between RSS and YouTube audiences.

Evaluating a campaign report

A marketing team asks which campaign segments should receive more budget, not merely which segments generated the most clicks. The analysis connects conversion quality and acquisition cost to a budget reallocation experiment.

The team receives a testable allocation decision instead of a descriptive dashboard.

Common mistakes

Analyzing without a decision

Starting with an open-ended request to find insights often produces observations that have no operational consequence.

Accepting generic recommendations

An AI may append standard advice that is only loosely connected to the evidence. Require an explicit link between each recommendation and the analysis.

Treating speed as value

Completing an analysis quickly is valuable only if the result improves a real decision or action.

Is it for you?

Best for

It is best for marketers, analysts, and leaders using AI or human researchers to support business decisions.

Not ideal for

It is not ideal for exploratory research whose purpose is discovery rather than immediate action.

From the transcript

Whenever I see someone do an analysis, I always ask them, "Okay, like if you do this analysis, are you gonna have clear recommendations and…

Kieran Flanagan · 04:30

And now, the real skill is learning how to use that power, like how to ask it to go do things that are actually meaningful,…

Kieran Flanagan · 04:30

From the episode

ChatGPT’s Code Interpreter: Your $20 Personal Data Analyst (#141)