Install any skill in seconds. Free to start, no credit card required.
Get Started Free →When the user needs to design, improve, or audit a post-signup activation flow to get new users to their first value moment. Activate when activation is lagging, time-to-value feels excessive, or first sessions lack impact.
.claude/skills/mkurman-onboarding-flow/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 61% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 55% | 0% |
-------|------------------------|---------------------| | B2B SaaS (simple) | Complete first core workflow | < 10 minutes | | B2B SaaS (complex) | Import data + run first report | < 24 hours | | PLG / Self-serve | Invite first team member + collaborate | < 48 hours | | Developer tool | First successful API call or deploy | < 30 minutes | | B2C app | Complete first session/transaction | < 3 minutes |
Prioritize onboarding that happens within the product experience itself, not detached from it. Product tours that overlay the UI without context are less effective than:
Diagnose before prescribing. Different stall points indicate different problems:
| Stall Point | Likely Cause | Intervention | |-------------|-------------|--------------| | Signed up, never returned | Unclear value prop or bad timing | Email with specific use case matching signup context | | Started setup, abandoned | Too many steps or hit a blocker | Email linking directly to where they stopped | | Completed setup, never used | No compelling reason to return | Trigger-based email when something relevant happens | | Used once, never returned | First experience was not valuable | Ask what they were trying to accomplish; offer guided call |
All onboarding designs should include:
support-docs — Create help center articles and getting-started guides that support the onboarding flowemail-marketing — Build the full lifecycle email program beyond onboarding (retention, expansion, win-back)churn-analysis — When onboarding completion data reveals early churn patterns needing deeper investigationUser: "We're a B2B project management tool. Users sign up but only 20% create their first project. Help us fix onboarding."
Good output excerpt: > Activation Event: Create first project + add at least one task (correlates with 60-day retention). > > First 30 Seconds Design: > After signup, land on a pre-built sample project (not an empty dashboard). The user sees what the product looks like when populated. A single guided prompt says "Create your first real project" with 3 templates to choose from. After project creation, inline prompt to add a first task with an example. Celebration state: "Your project is live! Invite your team to collaborate." > > Expected impact: Reducing pre-value steps from 4 to 2 should increase project creation from 20% to 45-55%.
User: "We have 3,000 signups from last month and 60% never completed setup. How do we bring them back?"
Good output approach: Segment the 60% by where they stalled. Design different re-engagement emails for each stall point. Include subject lines, body copy, and CTAs linking directly to the abandoned step. Recommend a sunset policy (stop emailing after 30 days to protect deliverability). Build an experiment backlog testing different interventions at each stall point, prioritized by volume of stalled users and estimated recovery rate.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | fail→fail | 13,909 | 16,841 | +21% | 1 | 1 | 0% | 2,194 | 3,700 | +69% | 0 | 0 | — |
case-01 | fail→pass | 30,453 | 23,747 | -22% | 1 | 1 | 0% | 4,275 | 4,794 | +12% | 0 | 0 | — |
case-02 | fail→fail | 28,850 | 28,261 | -2% | 1 | 1 | 0% | 4,480 | 5,679 | +27% | 0 | 0 | — |
case-03 | fail→fail | 25,257 | 25,063 | -1% | 1 | 1 | 0% | 3,794 | 5,019 | +32% | 0 | 0 | — |
case-04 | pass→pass | 17,523 | 21,231 | +21% | 1 | 1 | 0% | 2,902 | 4,500 | +55% | 0 | 0 | — |
case-05 | pass→pass | 26,715 | 34,228 | +28% | 1 | 1 | 0% | 4,102 | 6,424 | +57% | 0 | 0 | — |
case-06 | pass→pass | 27,473 | 25,506 | -7% | 1 | 1 | 0% | 4,343 | 5,064 | +17% | 0 | 0 | — |
case-07 | fail→fail | 18,778 | 18,408 | -2% | 1 | 1 | 0% | 2,517 | 3,753 | +49% | 0 | 0 | — |
case-08 | fail→pass | 13,345 | 13,830 | +4% | 1 | 1 | 0% | 2,033 | 3,183 | +57% | 0 | 0 | — |
case-09 | pass→pass | 13,147 | 16,964 | +29% | 1 | 1 | 0% | 2,007 | 3,696 | +84% | 0 | 0 | — |
case-10 | fail→pass | 14,488 | 14,220 | -2% | 1 | 1 | 0% | 2,220 | 3,314 | +49% | 0 | 0 | — |
case-12 | fail→pass | 13,151 | 15,515 | +18% | 1 | 1 | 0% | 2,147 | 3,460 | +61% | 0 | 0 | — |
case-13 | pass→pass | 8,655 | 9,749 | +13% | 1 | 1 | 0% | 1,219 | 2,604 | +114% | 0 | 0 | — |
case-14 | pass→pass | 9,311 | 8,791 | -6% | 1 | 1 | 0% | 1,555 | 2,527 | +63% | 0 | 0 | — |
case-15 | pass→pass | 18,524 | 20,606 | +11% | 1 | 1 | 0% | 2,735 | 4,185 | +53% | 0 | 0 | — |
case-16 | pass→pass | 12,341 | 13,960 | +13% | 1 | 1 | 0% | 1,915 | 3,207 | +67% | 0 | 0 | — |
case-17 | fail→fail | 16,868 | 22,847 | +35% | 1 | 1 | 0% | 2,380 | 4,367 | +83% | 0 | 0 | — |
case-18 | fail→fail | 13,613 | 13,351 | -2% | 1 | 1 | 0% | 1,833 | 3,091 | +69% | 0 | 0 | — |
case-19 | pass→pass | 12,967 | 12,305 | -5% | 1 | 1 | 0% | 2,024 | 2,825 | +40% | 0 | 0 | — |
case-20 | pass→pass | 9,843 | 9,683 | -2% | 1 | 1 | 0% | 1,443 | 2,498 | +73% | 0 | 0 | — |
case-21 | pass→pass | 13,336 | 13,809 | +4% | 1 | 1 | 0% | 1,967 | 3,105 | +58% | 0 | 0 | — |
case-22 | pass→pass | 11,170 | 11,613 | +4% | 1 | 1 | 0% | 1,746 | 2,825 | +62% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +18 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.