Content ICP and Grader
Define your ideal reader, then score every draft against that audience.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 7
- Confidence
- 92%
The Content ICP and Grader adapts ideal customer profiling to editorial strategy. Begin with accounts already reaching the market, examine the people who engage with them, and segment those people into clusters based on shared roles, interests, problems, and behaviors. Select the cluster most aligned with the creator's objective, such as awareness or demand generation, and turn it into an ideal reader profile. That profile then becomes the input for a content grader: a consistent scorecard that tests whether a draft addresses the reader's priorities, uses appropriate language, creates attention, and supports the intended goal. The mechanism creates a feedback loop from audience evidence to content design and from draft evaluation to revision.
Origin
Extracted from Marketing Against the Grain, where Kieran Flanagan proposed using Grok's access to X data to build an ideal customer profile for content and then grade drafts against it.
Core principles
- 01Content needs an ideal audience just as products need ideal customers.
- 02Observed engagement is stronger evidence than assumed audience preferences.
- 03Audience clusters should capture distinct needs and behaviors.
- 04A scoring model should evaluate content against a defined reader profile.
- 05Repeated grading helps creators refine positioning and execution.
How to run it
- 1
Define the content objective
Choose the primary result the content should produce, such as awareness, engagement, or qualified demand. Use that objective to determine which audience clusters matter most.
Pro tip Select one primary objective so the grader does not reward conflicting outcomes.
Watch out A vague objective produces an equally vague content ICP.
- 2
Choose reference accounts
Identify competitors and creators whose audiences overlap with the people you want to reach. Use several accounts to reduce dependence on one creator's unusual following.
Pro tip Include both direct competitors and adjacent category educators.
Watch out Do not assume every follower of a large account belongs in your target market.
- 3
Analyze engaged followers
Collect available engagement signals and examine the profiles of people who reply, repost, or otherwise interact. Focus on demonstrated engagement rather than passive follower counts.
Pro tip Weight repeated engagement more heavily than a single interaction.
Watch out Platform data may omit signals such as likes or bookmarks.
- 4
Cluster the audience
Group engaged people by shared role, needs, interests, behaviors, or desired outcomes. Give each cluster a concise description that distinguishes it from the others.
Pro tip Keep clusters actionable enough that a writer could tailor a draft to each one.
Watch out Overlapping or overly broad clusters will weaken the final profile.
- 5
Build the content ICP
Select the cluster most aligned with the content objective and document its pains, motivations, language, preferred formats, and reasons for engaging. Treat this as the ideal reader for future work.
Pro tip Include observable evidence for every major audience claim.
Watch out Do not turn unsupported demographic guesses into targeting requirements.
- 6
Create the grader
Translate the content ICP into scoring criteria covering relevance, specificity, language, attention, credibility, and fit with the desired action. Define what strong and weak performance looks like for each criterion.
Pro tip Use anchored scoring descriptions instead of unexplained numeric ratings.
Watch out A grader without explicit criteria will merely reproduce subjective taste.
- 7
Score, revise, and learn
Grade a draft, revise its weakest dimensions, and publish the improved version. Feed actual performance data back into the ICP and scoring criteria over time.
Pro tip Track predicted scores alongside real engagement to calibrate the grader.
Watch out Do not treat the AI-generated score as proof that content will perform.
In the wild
A B2B creator analyzes engagement around respected marketing operators and finds a cluster of fractional CMOs struggling to manage multiple clients. The resulting content ICP emphasizes time pressure, client credibility, and reusable systems. A newsletter draft scores poorly on specificity, so the creator replaces general productivity advice with a concrete client-handoff workflow before publishing.
→ The final draft speaks directly to a defined professional audience and has a measurable hypothesis behind it.
A software founder studies people replying to several infrastructure creators. The audience separates into practitioners, engineering leaders, and investors. The founder chooses engineering leaders, builds a grader around operational risk and team efficiency, and revises product-led posts that previously focused only on technical features.
→ The content shifts from generic product promotion to buyer-relevant operational insight.
Common mistakes
Using follower counts as audience evidence
Followers may be inactive, irrelevant, or attracted by unrelated content. Prioritize people who demonstrate meaningful engagement.
Building one giant audience cluster
A broad cluster hides important differences in needs and language. Separate groups until each profile supports distinct editorial choices.
Trusting the grader without calibration
An AI score is a hypothesis rather than an outcome. Compare scores with real performance and adjust the criteria accordingly.
Is it for you?
Best for
It is best for marketers, founders, and creators who publish regularly but lack clarity about whom their content should attract.
Not ideal for
It is not ideal when reliable audience-engagement data is unavailable or the creator has not chosen a specific market.
From the transcript
“But I was really interested in trying to figure out if you could build an ICP for your content.”
“it's going to segment those engaged followers into clusters”
“we're going to try to build a content grader which is basically a model that that allows you to give it content and then it…”
From the episode
Is Grok 4 Worth $30/mo? (Full Test)