Run competitive analysis

Run Competitive Analysis

Turn competitors' shipped products into evidence — usability evidence and positioning evidence — that changes a specific design or product decision.

When to use / when NOT to use

Use when:

• Starting a project and you need the landscape before designing (often the correct first research activity — NN/g)
• About to design a flow (booking, checkout, onboarding, AI chat) that competitors have already shipped — their live products are free prototypes
• A stakeholder asks "why don't we just do what X does?" and the answer needs evidence
• A redesign is planned and you want to test alternative directions without building them

Do NOT use when:

• The question is about the domain/market in general, not specific competitors → conduct-business-research (desk research + gap list)
• You need a deep expert review of ONE product against usability principles → run-heuristic-evaluation
• You already have teardown evidence and need the differentiation story → position-product
• You need to decide what to DO about a found opportunity (sizing, feasibility, risk-reward) → identify-business-problems

Evidence rule: every claim about competitors, users, or the market in this analysis is handled per craft-critique's evidence-discipline protocol. Load craft-critique before writing conclusions — its protocol governs how each external claim is sourced or flagged.

The method

Step 1 — Anchor to a decision (or stop)

Write one sentence: "This analysis exists so we can decide ______."

• Good: "…decide whether our booking flow should require account creation before showing available times."
• Good: "…decide which of 3 concept directions to build for the conversational shopping screen."
• Bad: "…understand the competitive landscape." That is not a decision. If no decision exists, this is desk research — hand it to conduct-business-research.

Everything downstream is filtered by this sentence. A finding that cannot change the decision does not go in the report.

Step 2 — Choose the mode (decision tree)

What does the decision need?
│
├─ Overall standing: "how good are we vs. them, broadly?"
│   → HOLISTIC evaluation
│     Same core tasks on every product, rank on overall UX
│     quality per dimension. Best at kickoff or annual checkpoint.
│
├─ One flow/feature: "how do others handle X?"
│   → FOCUSED evaluation
│     One flow, decomposed into steps, compared step-by-step
│     across products. Best when a specific design is imminent.
│
└─ A redesign with real users available?
    → COMPETITIVE USABILITY TESTING
      Put real users on competitors' LIVE products — they are
      free, fully-built prototypes of alternative design ideas.
      Use `plan-usability-test` for sessions/logistics; this
      skill defines what to compare and how to report it.

If unsure: focused. Holistic analyses sprawl; focused ones ship implications.

Step 3 — Pick 3–5 competitors across three rings

Ring	What it is	Why it must be included
Direct	Same product, same audience	The obvious comparison — but studying only these breeds convergence
Indirect / analogous	Different product, same job (e.g. a restaurant-reservation app when studying salon booking)	Where the actually-new patterns come from
Status quo / non-consumption	The workaround: phone call, WhatsApp message, spreadsheet, "do nothing"	The real competitor for most products — if you can't beat the phone call, the UI doesn't matter

Hard rules: 3–5 total (fewer = anecdote, more = never finishes). At least one non-direct ring represented. Name why each was chosen in one line.

Step 4 — Define dimensions and tasks BEFORE touching any product

Deciding what to compare after browsing competitors invites cherry-picking (confirmation bias — you'll "find" support for the direction you already like; name and control it per name-and-control-bias).

• Holistic: 4–7 dimensions tied to the decision (e.g. first-run experience, core-task efficiency, error recovery, trust signals, accessibility, mobile parity). Plus 2–3 core tasks executed identically on every product.
• Focused: one flow, decomposed into its steps (entry → … → confirmation). Each step is a row.
• Per dimension, write the evidence you'll capture: step count, taps to complete, forced pre-work (account? payment?), error/empty states encountered, recovery paths, load feel, copy tone. For AI products add: how uncertainty is shown, fallback behavior when the model fails, latency masking.

Scoring: use behavior-anchored marks, not star ratings from feel: ✓ works without friction · ~ works but with named friction · ✗ fails or blocks · — absent. Every mark must have an observation attached. Never average marks into a single number — the tradeoffs ARE the finding.

Step 5 — Run the UX pass

Execute the same tasks on every product, in the same order, capturing as you go:

• Screenshot every step of the flow (dated — products change)
• Count steps/taps from entry to goal
• Note every point of friction verbatim ("forced signup before showing prices")
• Deliberately trigger one error (bad input, back button mid-flow) and record recovery
• Check the flow on mobile if the decision touches mobile
• Note what each product does NOT have — absence is data, but never conclude absence = opportunity without asking why it's absent (Step 7)

Step 6 — Run the positioning teardown

For each competitor, answer five questions from public evidence (homepage, pricing page, app-store listing, public reviews):

1. Who is it for? — the audience the homepage headline actually addresses
2. What is the promise? — the differentiation claim in their own words (quote it)
3. How does it monetize? — pricing model: subscription / commission / freemium / ads; who pays, per what
4. What is the moat claim? — network, data, integrations, brand, price
5. Where do they push growth? — which surfaces carry upgrade/expansion pressure (onboarding, paywalls, share loops)

This layer explains WHY their UX is shaped the way it is. A marketplace that monetizes per-booking will bury direct-contact info; that is a business constraint, not a UX oversight — and copying the pattern without the business model copies the cost without the benefit.

Step 7 — Interrogate every pattern before adopting it

For each notable competitor pattern, ask in order:

1. Does it actually work? Observed evidence or public signal (reviews complaining about it?) — or unknown → flag under-evidenced.
2. Why does it exist? Genuine user value / their business model / legacy debt / regulation. Only the first is a reason to adopt.
3. Does it transfer? Their audience, scale, and model vs. ours.
4. If everyone does it the same way — is it a convention users now expect (follow it), or a shared blind spot (the opportunity)? State which, with reasoning.

Step 8 — Build the decision-ready matrix

Rows = dimensions/steps. Columns = competitors + own product (always include your own product as a column — the analysis is comparative, not touristic). Every cell = mark + observed evidence. Every row ends in a design implication: "so for us, ______."

A row with no implication gets cut. A matrix where any implication row is empty is a feature checklist wearing a costume.

Step 9 — Extract the bets and end with a verdict

• Maximum 3 opportunities, each traceable to specific matrix rows.
• Each opportunity states what evidence supports it and what's still under-evidenced.
• End with ONE recommendation against the Step-1 decision — not a list of options. Sizing and feasibility of the opportunities belong to identify-business-problems.
• Date the analysis and stamp an expiry (default: 90 days). Competitor products change; an undated matrix becomes misinformation.

Worked example (focused mode, condensed)

Decision: "Decide whether our salon booking flow should require account creation before showing available slots." Mode: Focused — booking flow, entry → confirmed appointment. Competitors: A = marketplace booking app (direct) · B = vertical salon SaaS (direct) · C = restaurant-reservation app (analogous) · Status quo = phone call to the front desk.

Step	A (marketplace)	B (salon SaaS)	C (reservations)	Phone call	→ Implication for us
See availability	~ visible, but only after location + service picker (4 taps)	✗ demands signup first — availability hidden	✓ calendar on screen 1, zero pre-work	✓ "does 3pm work?" — instant	Show slots first; every gate before availability loses to the phone call
Identify yourself	~ full account (email+password)	✗ account + salon code	✓ name + phone only, at confirm step	✓ "it's Priya"	Identity comes AFTER slot choice; phone-number-only clears the bar
Confirm + recover	✓ confirm screen, easy cancel link	~ confirm buried in email	✓ SMS confirm + one-tap cancel	~ no artifact — misremembered bookings happen	SMS confirmation is the convention users now expect (all digital players ship it)

Positioning teardown (one line each): A monetizes per-booking commission → hides the salon's phone number (model-driven, don't copy). B charges the salon a subscription → gates everything behind accounts because its buyer is the owner, not the client. C monetizes seat volume → optimizes for zero-friction entry; its audience overlaps ours in expectation.

Pattern interrogation: "Account before availability" exists in A and B — driven by their monetization/CRM needs, not user value; public reviews of B complain about it (cite the review source in the real report). Convention vs. blind spot: blind spot → our opportunity.

Opportunities (max 3):

1. Availability-first flow — beats the phone call at its own game. Evidence: rows 1–2. Under-evidenced: whether OUR clients abandon at signup — pull product analytics or test before betting.
2. Phone-number-only identity — evidence: row 2, C's pattern transfers (same expectation register).
3. SMS confirm — convention, table stakes, follow it. Evidence: row 3.

Recommendation: Do not require account creation before showing slots; collect phone number at confirmation. Analysis dated 2026-07-11 · expires 2026-10-09.

Anti-patterns / red flags

Red flag	Why it fails	Instead
Feature checklist ("they have X, we lack X")	Parity racing copies costs, not outcomes; features ≠ usability	Compare task performance and evidence, end each row in an implication
Only direct competitors	Guarantees convergent, me-too design	Three rings: direct, analogous, status quo (Step 3)
Skipping the non-consumption competitor	Most users' alternative is the workaround, not another app	Always evaluate the phone call / spreadsheet / WhatsApp
"Competitor does it, so it works"	Their pattern may exist for their business model or legacy reasons	Interrogate: works? why exists? transfers? (Step 7)
Star ratings from feel / averaged scores	Hides tradeoffs, unfalsifiable	Behavior-anchored marks + evidence per cell; never average
Choosing dimensions after browsing competitors	Cherry-picking to confirm the pet direction	Lock dimensions and tasks before touching any product (Step 4)
No own-product column	Tourism, not comparison	Your product runs the same tasks, same marks
Undated analysis reused for months	Competitors ship weekly; stale matrix = confident misinformation	Date + expiry stamp, default 90 days
Confident market claims with no source	Violates the evidence protocol	Handle per craft-critique's evidence protocol; load it before concluding
Analysis with no recommendation	Data without a verdict is unfinished work	One recommendation against the Step-1 decision

Output format

# Competitive Analysis — [product/flow]
**Decision this serves:** [one sentence]
**Mode:** holistic / focused / competitive usability testing
**Competitors:** [3–5, with ring + one-line why each]
**Date:** YYYY-MM-DD · **Expires:** YYYY-MM-DD

## Matrix
| Dimension/Step | Comp A | Comp B | Comp C | Us | → Implication |
(✓ / ~ / ✗ / — + observed evidence in every cell)

## Positioning teardown
| | Audience | Promise (quoted) | Monetization | Moat claim | Growth surfaces |

## Pattern interrogation
[Notable patterns → works? / why exists? / transfers? / convention or blind spot]

## Opportunities (max 3)
[Each: matrix rows it traces to + what's still under-evidenced]

## Recommendation
[ONE — answers the decision sentence]

Sources

• Competitive usability evaluations (holistic vs. focused): https://www.nngroup.com/articles/competitive-usability-evaluations/
• Competitors as free prototypes in redesigns: https://www.nngroup.com/articles/redesign-competitive-testing/
• Conducting a competitive usability evaluation: https://www.nngroup.com/videos/conduct-competitive-usability-evaluation/
• Secondary/desk research before primary research: https://www.nngroup.com/articles/secondary-research-in-ux/
• Confirmation-bias countermeasures (lock criteria before looking): https://www.nngroup.com/articles/confirmation-bias-ux/
• UX strategy (research-based, market-viable positioning): https://www.nngroup.com/articles/ux-strategy/

Boundaries

• conduct-business-research owns desk/secondary research and the gap list for a domain; this skill owns hands-on evaluation of specific competitor products. Run desk research first when the domain itself is unfamiliar.
• position-product consumes this skill's teardown as evidence for the differentiation story; this skill produces evidence, never the story.
• identify-business-problems owns opportunity sizing, risk-reward quadrant, and feasibility for what to do about the findings.
• run-heuristic-evaluation owns deep single-product expert review; invoke it per-competitor only if one product needs principle-level depth.
• plan-usability-test owns sessions and logistics when real users test competitor products (competitive usability testing mode).
• craft-critique owns the evidence-discipline protocol and verdict language; every external claim here follows it.