Make a competitive-intelligence platform prove four things before you sign

Write requirements as jobs

A requirement is a sentence about your work that the platform can visibly fail. "Tracks competitor pricing pages" is a feature. "Tells one product marketer within a week when any of five competitors changes a price, with the old and new page side by side" is a requirement, because you can watch it fail. The numbers are an example of the form; use your own.

Four facts settle most of the list: how many competitors matter and how fast they change; how quickly a change has to reach sales to be useful; who reads the output and in which tool they already work; and what a wrong output costs. A seller quoting a stale price to a buyer is a different risk from an old number in a quarterly slide.

Write these down before a shortlist exists, so every vendor call tests the product against your sentences.

Four things to demand evidence for

Every vendor will say it sources claims, keeps them fresh, supports review and takes security seriously. The difference is what each can show. Ask all four with a live account open, on a competitor you named.

A platform that shows all four on your competitor has answered a different question from a feature comparison: not only what the product says it does, but what you can verify. A roadmap answer on any row tells you where to discount.

Why a feature comparison alone will mislead you

Current buyer guides offer useful feature frames. Klue's guide asks buyers to assess automated data collection, deal-level delivery, win-loss integration, AI research and CRM or revenue-tool integrations. Quid's guide organizes its feature check around real-time monitoring, AI analysis, centralized data aggregation and integrations. Use those frames to learn the category, then apply the four evidence tests above to each vendor you shortlist.

Any price in a comparison list needs a source and checked date before you rely on it. A list also covers only the vendors its author chose. Read these pages for the category's vocabulary, then decide with your own requirement sentences.

Check vendor claims yourself with Claude

An ordinary Claude chat or a Claude Project is enough, because this is synthesis over pages you open yourself. If you keep a Project for product marketing, the setup lesson covers instructions and files; one chat is fine for a single evaluation.

1. Draft the requirements in your own words

   Paste the four facts and ask Claude for requirement sentences you can test. Check that each names a failure you could observe. A sentence that cannot fail is a preference and belongs lower on the list.

2. Open each vendor's own pages

   Paste the URL of each shortlisted vendor's pricing, security and product pages, or turn on web search so Claude fetches them during the conversation. Research does this across several pages at once and needs a paid plan (Pro, Max, Team or Enterprise), per Anthropic's help article checked 14 September 2026. Never ask Claude what a vendor offers from memory.

3. Extract only claims with a source on the fetched page

   Ask for a table of the vendor's claims against your requirements, one source line per row, plus a separate list of claims the page states without support. Open every source line yourself and confirm the page still says it. A claim Claude cannot cite is dropped, never softened into "likely".

4. Save the evidence file for the next decision

   Keep one file per vendor: the URLs, the dates you opened them and the claims that held. It is the same habit the evidence-first research plan and the claim ledger lesson teach for competitors, and it is ready at renewal.

The extraction instruction in step three is the one prompt worth keeping verbatim, because its wording is what stops Claude filling gaps from memory.

Using only the pages I have pasted or you fetched in this conversation, list
each claim this vendor makes about sourcing, freshness, approval workflow and
data handling. For each claim give the exact sentence and the URL it came
from. Put any claim you cannot tie to a fetched page in a separate list
labelled "unsupported". Do not add anything from your own knowledge of this
vendor.

When DIY is still the right call

Keep the job in-house while you can keep the relevant sources current and review the evidence without missing material changes. The solo competitive-intelligence lesson shows what one person can run, and the build-versus-buy guide compares hand checks, your own automation and a platform as ongoing work. Judge by missed changes, maintenance effort and review effort; competitor count alone decides nothing.

Two sub-decisions have their own guides because their evidence questions differ from the platform's: the battlecard software guide for generation and approval, and the win-loss software guide for buyer-feedback programs. Once your requirements exist, the evaluation checklist turns them into a short test on your own competitors.

Before any signature, the evidence file should answer these. A blank row is your next question for the vendor.

• Every requirement is a sentence that names an observable failure.
• Each vendor showed a live claim with a source link on a competitor I named.
• Each vendor gave a last-checked date and a schedule for at least one source.
• The approval path was shown on a real edit, with the role that accepts it.
• A security document exists that I can attach to a vendor-risk review.
• Pricing is in writing, dated, for the scope I described.
• The pass conditions for testing on my own competitors are written before it starts.

What Steve's published scope shows

Steve's product page says it checks a competitor's website daily, and its battle cards show a suggested update beside the current text; nothing changes until someone chooses. Data handling and public pricing need separate checks.

Steve's SLA page states that Trissino maintains a SOC 2 Type 2 attestation. Ask for the report as you would from any vendor.

Pricing sits outside the four rows, and Steve does not publish it; the pricing page is a quote request. If your requirement is a published, comparable price, Steve does not publish one; put the same requirement to every vendor.

Steve competitive intelligence describes the monitoring and battlecard scope; if the requirements hold, contact us with the competitors you track.