agentleFS
MadalaVijayMadalaVijay/marketing-skills-deep/skills/funnel-diagnosis/SKILL.md

funnel-diagnosis

Skill0 starsChanged 2 months agoMITView on GitHub

Find where a funnel actually leaks, separating real drop-off from measurement artefacts and from steps that only look bad. Use when conversion is falling, when asked "where should we focus", or before committing budget to fixing a stage.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before diagnosing, split by at least:

    line 46
  2. Distinctive:

    No segmentation available.

    line 69

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Funnel diagnosis

Most funnel work fixes the stage with the worst-looking number. That is usually the wrong stage.

Order of operations

1. Check the instrument before the reading. A step that reports 3% conversion may be converting at 30% and firing its event once in ten times. Before diagnosing anything, confirm each step's event is recording — compare against a source the front end does not control, such as the database, the CRM or the payment processor. A conversion event that silently stopped looks exactly like a funnel that stopped converting, and it is far more common.

2. Size the leak in absolute terms, not percentages. A step converting at 20% that 500 people reach loses 400 people. A step converting at 60% that 20,000 people reach loses 8,000. Fix the second one. The first has the uglier number and a twentieth of the upside.

3. Compare each step to its own history, not to a benchmark. Published funnel benchmarks are close to useless: they average across price points, categories, traffic sources and intent. A step is underperforming when it is worse than it was, or worse than the same step for a comparable segment in your own data.

Where the leak usually is not

  • The step with the lowest conversion rate. Some steps are meant to be narrow. A checkout that converts 40% of carts is not broken; a homepage that sends 4% to product is doing its job.
  • The last step before the drop. People abandon at checkout for reasons formed three steps earlier: unclear pricing, missing trust, wrong expectation. The visible abandon is often a lagging symptom.
  • The stage everyone argues about. Attention concentrates on the stage with an owner, not the stage with the loss.

Segment before concluding

An overall funnel number is a weighted average of very different behaviours. Before diagnosing, split by at least:

  • Traffic source. Paid, organic and direct convert differently at every step
  • Device. Mobile form completion is routinely half of desktop
  • New versus returning
  • Intent tier, where you can infer it — branded search behaves nothing like a cold interest audience

A funnel that looks mediocre in aggregate is often two funnels: one healthy, one broken. Fixing the average fixes neither.

Quantify before recommending

For each candidate leak, state: people lost per month, the conversion lift you would need to matter, and whether that lift is plausible. "Improve checkout by 30%" is not a plan. "Cut three fields to recover roughly a third of the monthly form abandons, which at current traffic is a few hundred people" is.

When to refuse

  • Fewer than ~100 people through the step in the window. Rates on small numbers swing wildly
  • Event integrity unverified. Say so and stop
  • No segmentation available. Give a provisional read and label it provisional

Output

The three largest leaks by absolute loss, each with the count, the segment it concentrates in, the evidence the tracking is sound, and one specific change with an estimated recovery. Then explicitly: which stage looks worst but is not worth fixing, and why.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.