All Claude Rundowns

article·9 min read

Rundown 04 — Run Competitive Research That Stays Current

Builds a sourced, dated profile of your real competitors, including the spreadsheet your customers use instead of you, then runs a monthly check that reports only what actually changed, with a link for every claim.

  • Workflows
  • Product discovery

What it does: builds a competitive landscape where every fact has a source and a date, then runs a monthly check that tells you what moved: pricing, positioning, launches, hiring, and what their customers are complaining about.

When to use it: you're about to make a positioning, pricing or roadmap decision and "I think they're cheaper" is the current state of the evidence. Or someone made a competitor deck last year and it's still being quoted.

What you need: Claude with web search switched on, and a list of the decisions this research is meant to inform. Forty-five minutes once, then ten minutes a month.

Time: the monthly run is shorter than the meeting where someone asks "has anyone looked at what they're doing lately?"

---

Why most competitive research is worse than useless

Three failure modes, and AI makes the third one much easier to fall into:

1. The wrong competitor set. The research covers the three companies that look most like you and misses what customers actually choose instead: a spreadsheet, an agency, the incumbent they already pay for, or doing nothing. You lose more deals to the status quo than to any logo on your slide.

2. It's a deck, not a system. It's accurate the week it's made, nobody owns it, and six months later a pricing page that has changed twice is still being quoted in sales calls.

3. Unsourced claims. "They're moving upmarket." "Their pricing starts at about €50." Neither comes with a link or a date. A language model will fill a gap with something plausible, and a plausible competitor price is worse than no price, because people act on it.

The setup below fixes the set before it profiles anyone, makes every fact carry its source, and treats "not found" as a perfectly good answer.

---

Part 1 — the landscape prompt (run once)

Paste this into Claude with web search switched on.

Help me build a competitive landscape I can keep current with a monthly check.
Accuracy matters more than completeness. A gap I can see is fine; a guess
presented as fact is not.

STEP 0 - CHECK WHAT YOU CAN REACH
Tell me whether you can search the web and open pages in this session, and
whether you can create and edit a Google Sheet or will give me tables to paste.
Check, don't guess.

STEP 1 - ASK ME
In one message, ask for:
- What my product does, in one sentence, and who it's for
- The job customers hire it to do, in their words if I have them
- Competitors I already know about, and what customers use when they don't
  buy anything
- The two or three decisions this research needs to inform (for example:
  pricing, which segment to target first, what to build next)
Then wait.

STEP 2 - AGREE THE SET
Propose a competitor set in four groups:
- Direct: same job, same customer, similar product
- Indirect: same job, done a different way (a services firm, a feature
  inside a bigger platform)
- Status quo: what customers do today without buying anything (spreadsheets,
  email, an intern, doing nothing). Always include this group.
- Adjacent: companies serving the same customer that could plausibly move into
  this job
One line on each explaining why it's there. Recommend no more than eight to
profile in depth and say why those eight. Wait for my approval.

STEP 3 - PROFILE EACH ONE
For each approved competitor, find:
- Positioning: their homepage headline, quoted verbatim
- Who it's for: the customer they say they serve, and who appears in their
  case studies (these often differ, and the case studies are more honest)
- Pricing: model and published prices, quoted from the pricing page. If
  pricing isn't published, write "not published". Never estimate a price.
- How they sell: self-serve, sales-led, partners, marketplace
- What customers praise and complain about, verbatim from review sites,
  communities or app stores, with the source. At least three of each where
  available.
- Recent moves in the last 12 months: launches, pricing changes, funding,
  senior hires, acquisitions, each with a date.

RULES
- Every factual claim gets a source URL and the date you accessed it.
- Label everything as one of: FACT (sourced from somewhere independent or
  observable), CLAIM (what the company says about itself), or INFERENCE (your
  reading). Never blend them in one sentence.
- If you can't find something, write "not found". Don't fill the gap.
- Treat job postings as signals, not facts. "Hiring three enterprise AEs" is a
  fact; "moving upmarket" is an inference and gets labelled as one.

STEP 4 - OUTPUT
1. ONE-PARAGRAPH SUMMARY - how the market is divided up, in plain language.
2. COMPARISON TABLE - competitor, group, positioning, target customer, pricing
   model, entry price, sales motion.
3. WHERE THEY'RE WEAK - complaint themes that appear for two or more
   competitors, with counts and quotes. These are the openings.
4. WHAT THIS MEANS FOR MY DECISIONS - for each decision I gave you in step 1,
   what the evidence suggests, labelled as INFERENCE, and what would change it.
5. WHAT YOU COULDN'T FIND - every "not found", in one list, so I know where
   to look myself.

STEP 5 - WRITE IT DOWN
Create a Sheet (or give me tables to paste) with two tabs:
- Profiles: one row per competitor with every field above, plus source URLs
  and a "last checked" date.
- Changes: competitor, date spotted, what changed, previous state, new state,
  source. Empty for now; the monthly check appends to it.
Then stop.

---

Part 2 — the monthly prompt

Competitors don't change weekly, and a weekly check mostly reports noise. Monthly is enough; run it on the first working day.

Run my monthly competitor check against <link to my Competitor Sheet>.

FOR EACH COMPETITOR IN THE PROFILES TAB, CHECK
- Homepage headline
- Pricing page
- Blog, changelog or release notes since the last-checked date
- Job postings, by function and seniority
- New reviews since the last-checked date
- News: funding, acquisitions, senior hires, partnerships

RULES
- Report a change only if you can show both the previous state (from the
  Profiles tab) and the new state, with a source for the new state. If you
  can't show the before, it isn't a change; it's a new observation, and gets
  labelled as one.
- "No change" is a good result. Don't pad.
- Label anything that is your reading of a change as INFERENCE.
- Rank changes by how much they bear on the decisions listed at the top of the
  Profiles tab, not by how interesting they are.

OUTPUT - keep it to this, nothing more:

Competitor check - <month>
  Changed (<count>):
  - <competitor> - <what changed> - was: <before> / now: <after> - <source>
    Why it matters: <one line, labelled INFERENCE, tied to a decision>
  (ranked as above)

  New this month (<count>):
  - <observations with no previous state to compare against>

  Complaint themes rising: any complaint theme that appears in new reviews
  for two or more competitors, with quotes. If none, write "none".

  No change: <competitors with nothing to report, in one line>

  Couldn't check: <anything blocked, missing or paywalled>

THEN
Append each confirmed change to the Changes tab. Update the matching fields
and "last checked" date in Profiles. Don't touch anything you couldn't verify.

---

Getting more out of it

  • Read their reviews before their website. The website is what they want to be; the complaints are what they are. A complaint theme that shows up across three competitors is a positioning line waiting to be written.
  • Sign up yourself, once a quarter. Claude can read a pricing page; it can't feel a bad onboarding. Go through the trial of your top two competitors and write down where you got stuck.
  • Ask switchers. In every customer interview, ask what else they looked at and why they didn't pick it. Put the answers through Rundown 01, and they become the best competitive evidence you have, because they come from buyers rather than websites.
  • Watch their moves, don't copy their roadmap. A competitor launching something tells you where they think the market is going. It isn't a ticket for your backlog.
  • Keep it to public information. Don't pose as a prospect to get pricing or a demo. Apart from the ethics, it's a small market, and people talk.

---

If you do this more than twice

Which you will, because the whole point is the monthly check. Save it as a skill in Claude Code, Cowork or Codex, schedule the monthly run, and give it one owner. The skill version of this Rundown is [link], with the Sheet template included.

A competitor deck shows you one month. The Changes tab builds up month after month into a record of how your market is moving, and that's the part nobody else in the room will have.