Saas rcd audit

SaaS Audit — Revenue Centric Design (RCD)

Audit any SaaS so its design serves user value AND revenue at the same time. This skill turns a body of behavioral-design knowledge into a repeatable audit that surfaces concrete, prioritized improvement points — each tied to a named cognitive bias or principle, with a recommended fix.

Source & credit. These frameworks synthesize public posts and articles by Richard — "Design for startups" (@richardrx): Revenue Centric Design, the Swiss Knife Index, the 4 startup phases, GBB pricing, the feature filter, and more. The underlying science is credited to its originators (Kahneman & Tversky, Dan Ariely, Eugene Schwartz, Aronson & Mills, Nir Eyal, Fitts, Hick, Nielsen Norman Group). Apply persuasion ethically — see Ethics gate.

Core thesis

Neutral design is not "objective" — it is omission. An interface that does not deliberately guide attention makes the path to value longer and revenue harder. Good product design is never user-value or business-revenue; it is both, by construction. Cognitive biases are the levers; the customer lifecycle is the map; the startup phase sets the priorities.

What this skill does

Given a SaaS (a URL, screenshots, a description, a codebase, or a specific surface like "our pricing page"), produce an audit: a prioritized list of improvement points, each with evidence, the bias/principle it violates, a concrete fix, and an effort×impact estimate. Works for a full audit or a focused one (just pricing, just onboarding…). For a focused request, still do Step 0–1 (phase + ICP) first — they reframe everything — then jump to the relevant stage.

Audit workflow

Copy this checklist into your working notes and fill it in:

Audit progress:
- [ ] Step 0: Startup phase calibrated (ARR/MRR + priorities)
- [ ] Step 1: ICP clarity gated
- [ ] Step 2: Lifecycle walked (Acquisition → Activation → Retention → Monetization → Metrics)
- [ ] Step 3: 9 RCD principles scored
- [ ] Step 4: Master checklist run
- [ ] Step 5: Findings emitted in report template

Step 0 — Calibrate to the startup phase → references/startup-phases.md

The most common failure is applying the wrong phase's advice (scale-stage growth loops on a survival-stage product). Establish ARR/MRR (or a best estimate) and set what matters:

Phase	ARR band	Goal	Where design pays off
Survival	R$0–240k	Prove someone pays	Visual differentiation, trust layer, speed-to-value
Traction	R$240k–1.2M	Luck or system?	Behavioral: onboarding, TTV, value realization, upsell levers, churn control
PMF	R$1.2M–6M	Prove scale (users arrive, stay, refer)	Retention efficiency, margin
Scale	R$6M+	Compete for attention	Continuous optimization + defensibility (network effects, data moat)

Step 1 — Gate on ICP clarity → references/acquisition.md

Everything downstream depends on a sharp, underserved ICP that: (a) feels the pain with real weight (the loss when unresolved is large), (b) is big enough (bottom-up TAM), (c) can pay the ticket your unit economics need, (d) has founder fit. A vague ICP ("anyone who wants the product") makes every later finding ambiguous and makes errors undiagnosable. If the ICP is unclear, that is finding #1.

Step 2 — Walk the lifecycle

For each stage, read the matching reference + the matching checklist section, then record gaps.

Stage	Read	Hunting for
Acquisition	references/acquisition.md	Hero fails the 5-sec test, jargon, generic CTA, no/weak proof, awareness mismatch, LP that doesn't qualify
Activation / Onboarding	references/onboarding-activation.md	No defined aha-moment, slow TTV, empty dashboard, passive tour, admin friction, journey gap
Retention / Churn	references/retention-churn.md	Expectation debt, jargon churn, dead cancellation flow, one-time-job model, reactive comms, low feature adoption, no habit loop, no moat
Monetization / Expansion	references/pricing-expansion.md	No decoy/anchor, >4 plans, free anchored at zero, bad defaults, weak upgrade path, upsell that interrupts instead of meeting a limit
Metrics / Experimentation	references/metrics-experimentation.md	Vanity metrics, underpowered A/B tests, blended LTV, inflated LTV:CAC, leaky-bucket math ignored
Product / Design strategy	references/product-design-strategy.md	Feature creep (Swiss Knife Index), no feature filter, cluttered dashboard, cosmetic-only refactors

Step 3 — Score the 9 RCD principles → references/revenue-centric-design.md

Rate each principle Pass / Partial / Fail with one sentence of evidence. This is the cross-cutting scorecard.

Step 4 — Run the master checklist → checklists/master-audit-checklist.md

The comprehensive, phase-aware gap detector. Every concept from the corpus appears here. Each item is a yes/no check tagged with its bias/principle and a pointer to the reference for the fix.

Step 5 — Emit findings → assets/audit-report-template.md

Fill the template: phase + ICP verdict, findings table (area · finding · evidence · bias/principle · severity · fix · effort×impact), top-5 quick wins, RCD scorecard.

Navigation — which file answers which question

Question	File
"What does each RCD principle mean and how do I check it?"	references/revenue-centric-design.md
"Which bias applies here / what's the study behind it?"	references/cognitive-biases.md
"Is the landing page / hero / copy / CTA / positioning right?"	references/acquisition.md
"Why do trials not activate? What onboarding pattern fits?"	references/onboarding-activation.md
"Why are people churning? How do I retain & build a moat?"	references/retention-churn.md
"How should pricing/tiers/defaults/upgrades be structured?"	references/pricing-expansion.md
"Is this metric or experiment trustworthy?"	references/metrics-experimentation.md
"Too many features? Cluttered dashboard? What to cut?"	references/product-design-strategy.md
"What should I prioritize at my stage?"	references/startup-phases.md
"Give me the full checklist."	checklists/master-audit-checklist.md

How to write recommendations

• Be specific and quantified. "Replace the hero headline 'A plataforma completa de produtividade' with one naming the ICP + outcome, e.g. 'Controle de comandas para oficina mecânica'" — not "improve the headline."
• Tie every finding to a named bias or principle (e.g. Anchoring, Loss aversion, RCD #6 Defaults). The "why" is what makes a recommendation persuasive and trustworthy.
• Respect the phase. Don't recommend scale-stage optimization to a survival-stage product. At survival, prioritize differentiation/trust/speed-to-value; at traction, behavioral onboarding & churn; later, retention efficiency & moat.
• Prefer high-leverage fixes over cosmetics. Pricing structure, defaults, and the onboarding sequence move metrics far more than button color. (1% better price ≈ 8–11% more profit vs ≈2% from 1% better acquisition — ProfitWell.) Flag "repaint" disguised as "refactor."
• Quantify the leak when you can. Use the leaky-bucket and cohort-LTV math in references/metrics-experimentation.md to translate a churn point into R$.

Ethics gate

Persuasion here is for clarity and momentum, not coercion. Enforce these or the audit becomes a dark-pattern manual:

• Defaults must be ethically defensible. A pre-checked box that traps users into spend is churn bait and a ReclameAqui/refund risk — recommend defaults that match what an informed user of the ICP would choose.
• Promise must equal proof (RCD #4). Never recommend a claim the product can't demonstrate — that creates expectation debt and causes churn.
• Scarcity/urgency must be real. Recommend honest scarcity only (true limits, real deadlines), never fabricated countdowns.
• Friction that qualifies ≠ friction that traps. Add friction to qualify intent or deepen commitment; never to obstruct cancellation.

Benchmarks quick-reference

Use as rough goalposts (mostly SaaS; some US data — adjust for the market). Sources in references/.

Metric	Weak	OK	Strong	Note
Activation rate	<20%	30–37%	40%+	<20% = structural onboarding/product problem
D7 retention	<10%	10–15%	>30%
Time-to-Value (TTV)	days	~1d12h (median)	<5 min (top)	Userpilot, 547 cos.
NRR (B2B)	<100%	~100%	>110%	Separates growth from leaky funnel
Trial→paid, card vs no-card	8.9% (no card)	—	31.4% (card)	Chartmogul 2026, US, 200 products
LTV:CAC	<3:1	3:1	>3:1	Recompute by cohort; blended overstates 30–50%
Swiss Knife Index	<0.3	—	≥0.3	features used by >40% of actives/30d ÷ total
Pricing tiers	1 or 5+	—	3 (2–4 max)	Hick's Law; decoy needs a middle plan
Reviews sweet spot	5.0 (looks fake)	—	4.2–4.5★	Northwestern

When citing a number, attribute the source and note it's a benchmark, not a guarantee.