5.3 Competitive Analysis

In Brief

Description

Competitive analysis is a structured comparison of the players already trying to solve the same job for the same kind of customer. The point is not to count competitors; it is to understand how each one is positioned for a target segment you care about, and to find a gap where your offering produces evidence of stronger fit than theirs.

Two principles run through the method.

The first is relational comparison. A feature is not a strength in isolation. SOC2 compliance is a strength against an enterprise buyer and irrelevant against a prosumer; a free tier is a strength against a price-sensitive solo user and a liability against a procurement-led buyer. Every cell in your matrix is scored against a named segment, never in the abstract.

The second is citation-backed claims. AI assembles competitive material faster than any human team, but it invents funding rounds, customer counts, and feature lists at a meaningful rate. Every numeric and named claim on a finished profile must trace to a verifiable URL. The verification discipline this demands is spelled out in Biases & Tips and built into the AI prompts below.

For market sizing (TAM / SAM / SOM, segment growth, willingness-to-pay benchmarks), use Secondary Market Research. This page is about who is in the field and where the gaps are; that one is about how big the field is.

How to

Prep AI Prompt

1. Define the target segment first, not the competitor list. Write a one-paragraph description of the customer you are analyzing competitors against: who they are, what job they are trying to do, what they currently use, and what they care about most. Every later judgment — a feature being a strength, a price being a problem, a channel being saturated — is judged against this paragraph. Skip this step and you produce a generic map that scores every competitor against no one in particular.

2. Pick which competitors are worth deep-research time. Don’t try to map “everyone in the space.” Aim for 8–15 names: 3–5 direct competitors, 3–5 indirect competitors solving the same job differently, and 2–3 substitutes from adjacent markets. Cross-reference at least two of these candidate-pool sources before locking the shortlist:

   • Customer discovery interview transcripts — whom did your interviewees name as alternatives, and what did they switch from?
   • Search trend analysis — which brands rank for the pain queries your target segment types into a search engine?
   • Keyword competition — who is bidding on your keywords in a keyword-intelligence tool?
   • Open-ended survey responses — answers to “what tool or method do you currently use for X?” and “who else did you consider?”
   • App-store category browsing — which apps appear in the categories your customer browses?
   • Social listening — which names keep coming up in community forums, chat communities, and review threads your customer reads?

   If a name shows up in only one source, treat it as a weak candidate. If it shows up in two or more, it earns a slot.

3. Decide your dimensions. Pick the 6–10 dimensions that matter most to your target segment, not a generic SaaS rubric. Examples: feature coverage, pricing model, time-to-value, integrations with the buyer’s existing stack, support quality, deployment options, compliance posture, channel mix, partner ecosystem. Drop dimensions that don’t map to a real concern of your segment — “supports SOC2” is dead weight if you sell to indie creators, and “free tier” is dead weight if you sell to procurement-led enterprises.

4. Sketch the artifacts you will produce. Pick the visualizations that will actually drive the decision in front of you, and stub them out before you collect data. Common artifacts:

   • Feature × segment matrix — competitors on rows, dimensions on columns, scored against the target segment.
   • 2×2 positioning grid — pick the two dimensions your segment cares about most; plot competitors; look for empty quadrants.
   • Petal diagram — for new categories, place yourself at the center and draw petals for each adjacent market your customers will switch from.
   • Pricing-tier comparison — list price, packaging, what’s included at each tier, common discounting patterns.
   • Channel-mix breakdown — paid search, content, partnerships, marketplaces, outbound, community — what share each competitor leans on.

   You do not need every artifact. Pick the two or three that answer the actual decision you are about to make.

5. Decide where verified primary sourcing matters. Some claims drive decisions (pricing, funding round size, feature presence on a paid plan, named integrations); others are background. Tag each dimension as decision-grade (must be verified to a primary source) or directional (a published roundup or AI summary is fine). Decision-grade claims will go through fact-checking later; directional claims will not. Without this triage, fact-checking either takes forever or doesn’t happen.

Execution AI Prompt

The verified shortlist from Prep is now the input to deep profiling. The goal here is not to write essays per competitor; it is to fill a fixed schema so that the synthesis stage has clean inputs.

1. Use a fixed profile schema. Each profile records the same fields in the same order. A reasonable default:

   • Company name, URL, year founded, HQ location.
   • Funding stage and total raised (from a funding database), most recent round.
   • Customer segments served (named, not “everyone”).
   • Top 6–10 features mapped to your dimension list.
   • Pricing model and published tier prices.
   • Channel mix (paid search, content, marketplaces, partnerships, outbound, community) — qualitative description plus traffic share if traffic-analytics data is available.
   • Integrations and named partners.
   • Sentiment summary from review sites and forums (positives, complaints, common deal-breakers).
   • Any recent product or strategy moves worth noting.
   • Inline [1], [2] citation markers for every numeric and named claim.
   • A Sources list at the end of the profile resolving every marker to a URL.

2. Collect each field from its canonical source.

   • Funding, stage, founding year → funding databases, SEC filings, press releases → spreadsheet column.
   • Features and positioning copy → competitor’s homepage hero, product pages, pricing page → feature matrix row.
   • Channel mix and GTM motion → traffic-analytics services (traffic sources), keyword-intelligence tools (keyword bids), ad-transparency libraries (creative), app-store presence, the competitor’s own integrations / partners page → channel-mix table.
   • Customer segment and sentiment → software review sites, community forums, and app-store reviews → segment notes + sentiment column.
   • Pricing and revenue model → published pricing page; review threads mentioning negotiated discounts → revenue model column.

3. AI does the first pass; humans verify decision-grade claims. A capable LLM with web access can fill most of the profile, but the remainder — and any claim you tagged as decision-grade in Prep — must be confirmed by visiting the cited source yourself or by routing through a fact-check subagent. Pricing changes often; funding round sizes are routinely overstated in press releases; “X customers” claims rarely include the basis on which a customer was counted.

4. Flag anything gated. Anything that requires a login, a paid database (full paid-database reports, paid analyst reports, gated review-site detail, sales-intelligence tools) gets flagged as MANUAL in the profile. AI cannot fetch behind authentication, and a confidently-stated number that the agent could not actually retrieve is the most common failure mode of AI-generated profiles.