── SKILL
product-usage-analyst✓ APPROVED

Read how an account actually uses you.

Turns raw usage into per-capability adoption tiers, a 12-week trend on each, power-user IDs, and a ghost-active flag.

Inside: The adoption-tier rubric, the trend classifier, and the ghost-active check.

01 / HOW TO USE

Install it in one line, or paste it in.

1
In Claude Code (one command)
Copy the install line, paste it into your terminal, and restart Claude Code. The skill installs itself to ~/.claude/skills/ and runs automatically when it is relevant.
2
In Claude, ChatGPT, or a Project (no terminal)
Open the file, then upload it to your chat or paste its contents in. A skill is just a markdown file of instructions, so any capable AI can follow it.
New to skills? A skill is a plain-text file that teaches your AI a workflow. Point any capable assistant at it and it follows the steps, on your data.
02 / MAKE IT YOURS

Connect your context. Set it to your motion.

CONNECT YOUR CONTEXT · AND WHY IT HELPS
a product-analytics tool

it reads events automatically, at your chosen account grain, across the whole base.

a CRM

cross-references usage to the account record and its fit score.

a longer history

a full trend window instead of a single snapshot.

SET IT TO YOUR MOTION

This was built for a B2B SaaS product with a handful of named capabilities. Set these to your stack:

The capability list is yours. Point the skill at the 5 to 15 features that actually signal value in your product, not anyone else's.

Set thisWhat it isDefault / Example
ANALYTICS toolyour product-analytics connectorAmplitudeMixpanelPostHoga CSV export
ACCOUNT_GRAINhow you group users into an accountemail domainworkspace idorg id
CAPABILITIESthe features you want adoption onyour own list (see below)
TREND_WINDOWweeks of history for the trend12 (re-tune to your cycle)
POWER_THRESHOLDusage that counts as a power usertop decile of activityor your own bar
NEW_USER_WINDOWrecency for "new signups"14 days
FIT_SCORE fieldyour account fit score, if you have oneany 0-100 fit or propensity score
03 / THE FULL SKILL

Everything the skill does, in full.

── WHAT THIS DOES

Reads how an account actually uses your product and gives a plain read: which capabilities they have adopted, whether each one is rising or fading, and who the power users are. It turns "how are they using us?" into a product-engagement score you can act on, and it flags the accounts that look active on the surface but create no real value underneath.

── THE METHOD
  1. 1
    Adoption tier per capability

    Every capability gets exactly one tier: POWER, ESTABLISHED, EMERGING, DORMANT, UNTOUCHED, NEVER-ADOPTED. Power means heavy, recent, repeated use. Untouched means they had it and stopped. Never-adopted means they never started. The Untouched and Never-adopted rows are the onboarding gap, the consultative outreach list.

  2. 2
    Trend per capability

    Over TREND_WINDOW weeks, each capability is RISING, FLAT, DECLINING, or COLLAPSED. A high tier with a declining trend is an early quiet-quitting signal, worth more than the raw count.

  3. 3
    Power-user identification

    Surface the individual users whose activity clears POWER_THRESHOLD, with which capabilities each one leans on. These are your expansion anchors and your reference candidates.

  4. 4
    Ghost-active check

    When an account looks alive on the baseline "any activity" signal but every real capability is zero, flag it. This is the classic blind spot: seats are technically active (someone logs in) but no value is being created. Must show the per-capability zeros that prove it.

  5. 5
    Fit-score cross-check

    If you carry an account fit score, compare it to what usage actually shows. When the fit score sits at the floor but adoption is clearly real, call it a false negative and promote the account. When the score is high but usage is a ghost, flag the mismatch the other way. The usage is the ground truth.

── QUALITY GATES
  • No GHOST-ACTIVE verdict without showing the per-capability zeros that prove it.
  • Every "not adopted" surfaces capability by capability, named, never "low usage."
  • Trend is over the full window, never a single week. Small samples lie.
  • Never report a number the export does not contain. A missing column is a prompt, not a guess.
── OUTPUT (EXAMPLE)
EXAMPLE OUTPUT
PRODUCT ENGAGEMENT · acme.com · 12-week window

Capability            Adoption      Trend       Last active
Core action           Power         Rising      today
Automation            Established   Flat        yesterday
Templates             Emerging      Rising      3d ago
AI assist             Dormant       Declining   18d ago
Advanced module       Untouched     -           -
Integrations          Never-adopt   -           -
_active baseline      Power         Rising      today

NOT ghost-active: 4 of 6 capabilities show real activity
Fit-score check: score 28 looks low, but adoption is real -> false negative, promote
Weekly actives: 351 latest · 339 four-week avg (Rising)
New users (14d): 8 signups · 5 active within 7d · 3 still untouched

Story: Power on the core action with rising template use. The gap is the
advanced module and integrations. 3 new users have not activated yet, that is
the onboarding conversation.
── EXAMPLE CHAT
EXAMPLE CHAT
You
How is acme.com using us? Here's the 12-week export. [pastes a CSV: one row per capability, activity counts, last-active dates, plus an _active baseline row and a fit score of 28]
CLAUDE
Read's done. Acme is real, and your fit score is wrong about them. PRODUCT ENGAGEMENT · acme.com · 12-week window Capability Adoption Trend Last active Core action Power Rising today Automation Established Flat yesterday Templates Emerging Rising 3d ago AI assist Dormant Declining 18d ago Advanced module Untouched - - Integrations Never-adopt - - Four of six capabilities show real activity, so this is **not** ghost-active. Power on the core action, and template use is climbing. Two things stand out: - **Fit-score false negative.** The score sits at 28 (floor), but adoption is clearly real. The usage is ground truth. I'd promote Acme and stop treating it like a cold account. - **A quiet-quitting tell.** AI assist is Dormant *and* Declining. Not urgent yet, but that's the capability to ask about before it collapses.
You
Who are the power users?
CLAUDE
Three clear the top-decile bar: j.reyes@acme.com (core action + automation, daily), m.okafor@acme.com (templates, rising fast, your reference candidate), and s.lin@acme.com (core action only). Reyes and Okafor are your expansion anchors. Want me to hand the two untouched capabilities (Advanced module and Integrations) over to the outreach skill as the onboarding-gap angle?
You
Yes.
CLAUDE
Passing the named gaps to `trigger-outreach`. One sharpener for next time: this ran on a single snapshot with the fit score attached. Connect your product-analytics tool and I'll pull the live trend and each new signup's first-seen date automatically. Same read, no paste, and I can catch the AI-assist decline the week it starts.
── WHERE THE NUMBERS COME FROM

TREND_WINDOW (12 weeks), NEW_USER_WINDOW (14 days), and the power-user cutoff are defaults, not laws. They suited a mid-market SaaS product with a steady weekly rhythm. If your product is used in bursts or on a monthly cycle, widen them. The logic does not change. The thresholds are yours.

04 / GO FURTHER

Where an operator takes this next.

The read is step one. Here is where an operator takes it once the manual version proves out.

The stack is boring on purpose: Amplitude reads it, Claude scores it, Salesforce holds the verdict, Slack delivers the one line that matters. You built the read once; now it runs itself.

01
Score the whole base while you sleep

Point a scheduled Claude task at Amplitude every night, run this read on every account, and write the product-engagement verdict and the false-negative flag straight back to Salesforce. You wake up to a scored book, not a to-do.

02
Catch only the flips

Have the nightly run diff against yesterday and DM the account owner in Slack just the accounts that *changed*: a capability that started collapsing, or a fit-score false negative that just proved itself. No dashboard to check; the signal comes to the rep.

03
Close the loop to outreach

Feed the Untouched and Never-adopted rows directly into a cold-email or trigger-outreach skill so the onboarding-gap list leaves as sent messages, not a spreadsheet nobody opens.

05 / PART OF A BIGGER PLAY

One skill is the on-ramp.

── GET NEW SKILLS IN YOUR INBOX

One operator-built skill a week. Free.