--- name: company-diligence description: >- Read one company's trajectory from public evidence: branded demand year over year, the acquisition engine underneath it, whether that demand was earned or bought, and what payroll says ships next — then name the questions only a data room can close. Use when: Before an investment, a partnership, an offer, or a dependency you cannot cheaply reverse — any decision where the subject is someone else's company. Use positioning-teardown instead when the question is how to sell against them, and demand-validation when it is whether the market exists at all. --- # Is this company actually growing? **The job.** I am about to bet on this company — invest in it, partner with it, sign a contract with it, or take the offer. Tell me from the outside whether it is actually growing, which engine is carrying it, and what I would only learn after I signed. **The method.** Bill Gurley's characteristics of a business worth a multiple, narrowed to the four an outsider can actually measure — growth, organic demand versus bought demand, concentration, and lock-in — and read as a series rather than a snapshot. The cheapest disqualifier runs first, so a run can end at call one with a real answer. **What it costs.** A disciplined run is about 13 billed calls, $2.60 at 20¢ per successful call. Stop at 22 ($4.40) — past that you are confirming, not learning. Failed calls are not charged, and reading this playbook is free. **Tools it uses.** `keyword_history`, `domain_overview`, `domain_traffic_history`, `domain_history`, `ads_creatives`, `ads_advertisers`, `domain_top_pages`, `web_search`, `company_jobs`, `company_profile`, `company_funding`, `company_firmographics` Each one is `GET https://wutheringai.com/v1/?…&token=`. The live list, with every parameter, is at https://wutheringai.com/v1. **Inputs.** company (required) — The company you are diligencing: its name and its domain. rivals — Two or three rival domains, comma-separated. They ride in the same history call so a collection gap is separable from a real decline. bet — What you are about to commit, and what would make you walk away. market — Country to measure, if not the United States. None were supplied on this request, so the placeholders below are yours to fill. **Deliver.** A trajectory verdict with the series under it: branded demand year over year against the rival set, organic value and footprint over 24 months, which engine carries the growth and whether it was earned or bought, what payroll says ships next, and the three questions only a data room can answer. --- ## The subject **FINDING URLS.** Several tools take a URL and cannot search for one — `social_linkedin_posts`, `social_reddit_posts`, `social_reddit_comments`, `company_jobs`, `company_person`, `web_fetch`. When you need a page, find its URL with YOUR OWN web search first, using a site: filter. That costs nothing and is exact. `web_search` does the same job for a billed call and is the fallback when you have no search of your own. Never construct a URL from a company name: a wrong URL comes back as a dead page, which reads as an absence of evidence rather than as your mistake. **BATCHING.** The keyword tools bill once per call however many terms you send — `keyword_volume` takes 1,000, `keyword_intent` 1,000, `keyword_overview` 700, `keyword_ideas` 200. Pool your terms and send them in one call rather than looping. A batch whose rows exceed one response comes back trimmed and says how many rows it held versus returned: size the next batch from that figure and send the terms you did not get. A term missing from a trimmed response was withheld, NOT measured at zero. Scoring it as zero is the single most common way one of these runs reaches a confident wrong answer. Every other playbook here is about your own product. This one is about someone else's, which changes the rules: you cannot ask them anything, you cannot see their numbers, and everything they publish is written to be read by you. So work from instruments rather than claims, and remember that **a single month is a number and only a series is a fact.** Two companies with the same traffic this month are not in the same condition if one of them is halfway up and the other halfway down. Write the bet down before the first call: what you are committing, and the one finding that would make you walk. A diligence run that discovers its own threshold after the data arrives has not tested anything. **THE EVIDENCE LADDER.** Rank every signal by what it would cost someone to fake. Prefer signals higher up, and never let a lower rung outvote a higher one: 1. Money someone is paying a competitor, or sustaining ad spend 2. Salary a company is paying humans to do this job 3. Search commercial-intent queries with real, repeated volume 4. Time people build workarounds — spreadsheets, templates, macros 5. Attention views, engagement, followers, upvotes 6. Words someone said they would like it. Never sufficient on its own. Rungs 1 and 2 cost money to fake. Rung 6 costs nothing, which is why so much of it exists. This run does not work straight down that ladder, and the reason is cost: branded search is rung 3, but it is one call, it cannot be bought cheaply at scale, and it disqualifies more subjects than anything below it. So the cheap disqualifier goes first, and the money and salary rungs follow in steps 4 and 6, where they explain the trajectory rather than merely confirm it. ## 1. BRANDED DEMAND — the cheapest disqualifier, and it runs first One `keyword_history` call, up to 700 terms for a single charge. Send the subject's brand, two or three rivals' brands, and the intent-bearing compounds: ` pricing`, ` alternatives`, ` login`, ` reviews`. More people typing a company's name is the hardest-to-fake public signal that it is growing. Nobody types a brand by accident, nobody buys the habit cheaply, and it moves months before revenue does. Read the **yearly averages**, not the last month. Then: - **Rising year over year, and rising faster than the rivals in the same call** — the company is taking the category. Continue. - **Flat while a rival rises** — they are losing share to a named party. That is the finding; the rest of the run explains it. - **Flat, and the whole rival set is flat** — the category is not growing. Stop here and say so. This is a result, not a failed run, and it cost 20¢. **KILL CRITERION, fixed now:** if branded demand for the subject is not higher than it was twelve months ago, nothing later in this run rescues it. Traffic, hiring and funding are all lagging or purchasable. This one is not. **The trap.** A brand that is also an ordinary word cannot be measured this way — the volume belongs to the word, not the company. When the brand is a common noun, throw the bare term away and keep only the compounds above. Reporting a common-word brand's search volume as company demand is the fastest way to be confidently wrong in this run. ## 2. THE TRAJECTORY — and put the rivals in the same call `domain_overview` on the subject first: one row, current position bands and estimated traffic value, and the cheapest way to learn whether this domain has enough footprint to be worth measuring over time at all. Then `domain_traffic_history` with `targets=,,` and `date_from` at least 24 months back. **Batch the rivals in deliberately, even if you do not care about them.** Public coverage has gaps, and a gap looks exactly like a collapse. A month where every domain in the batch falls together and recovers the next month is the instrument, not the market — and you can only see that because you asked about more than one company. A solo history call cannot tell a bad month for the company from a bad month for the index. Two series come back per domain and they move independently: etv estimated value of the organic traffic — the money count how many keywords the domain ranks for — the footprint Rising value on a shrinking footprint, and falling value on a growing one, are both common and mean opposite things. Do not report one without the other. ## 3. RESOLVE THE CONTRADICTION WITH POSITION BANDS `domain_history` on the subject. It returns the last few months only — it is not the long series, step 2 is — but it splits the footprint by rank, which is what settles step 2's ambiguity: pos_1, pos_2_3, pos_4_10 the positions that earn essentially all of it count everything, dominated by ranks 11-100 The tail is most of `count` and almost none of the money. So: - **Top-10 bands holding or rising while `count` falls** — a content programme was cut, or the index pruned a tail nobody was reading. The business did not lose anything it was being paid for. Reporting this as a collapse is the most common error in an outside-in read. - **Top-10 bands falling** — that is a real decline, whatever `count` does. - Read the `paid` block in the same response. It is the measured paid footprint, and step 4 has to agree with it. ## 4. EARNED OR BOUGHT — Gurley's ninth characteristic Demand a company created is worth a multiple. Demand it rents each month is worth its margin. From outside you can tell them apart. - **`ads_creatives` with `target=` is the first call, not the second.** It goes straight at the domain and returns the creatives with the dates each was first and last seen. - `ads_advertisers` is for disambiguation only, and it matches loosely on the name: run it on a short brand and it will return verified advertisers whose names merely contain the string. **Check the `title` or `domain` on every row against your subject before you attribute one ad to them.** Borrowed spend is worse than no data, because it reads as evidence. Then read the dates, because days live is the only performance signal public ad data carries — an advertiser stops paying for a creative that does not work: - **A small set of creatives, most of them running a year or more** — brand defence. Cheap, permanent, and not an acquisition engine. It should agree with a near-empty `paid` block in step 3. - **A large set with a short median life, churning constantly** — bought growth. Real, and it stops the month the budget does. Discount the growth curve accordingly, and ask what the payback period is. - **Sustained spend against a rival's brand terms** — someone is fighting for the same customer at renewal. If step 3 says almost no paid keywords and step 4 shows long-lived creatives, the two agree: the growth in step 2 was earned. Say that plainly — it is the single most valuable sentence an outside-in read produces. ## 5. WHERE THE VALUE SITS — concentration `domain_top_pages` on the subject. Two things to take from it: - **The homepage's share of total organic value is the branded share.** A large share means the search channel is really a brand channel: the number will follow whatever creates the brand and will not survive it. A small share on a large total means an acquisition engine that works without the brand carrying it. - **Which pages carry the rest.** A handful of pages holding most of the value is a concentration risk one ranking change can undo. The careers page appearing at all is hiring volume showing up in search. ## 6. WHAT PAYROLL SAYS SHIPS NEXT Hiring is the roadmap a company has already paid for, visible months before anything launches. Find postings with your own search, or `web_search` on `site:linkedin.com/jobs ""`. Then `company_jobs` on one or two URLs, which takes the posting's own URL. Read the **shape of the list** before any single posting: - A cluster of roles around one function is a bet being placed. - Roles named for a second geography, or for an expansion motion, say the growth is meant to come from existing customers and new countries rather than new logos. - Internal platform and sales-ops roles say the company is scaling process, which is what a company does after it has found something that works. - `job_posted_time` on a live posting separates a company hiring today from a careers page nobody has pruned. **Honest gap.** `base_salary` is empty far more often than not — bands are published where the law requires them and rarely elsewhere. When it is absent, the salary rung of the ladder degrades from "what they pay" to "which functions they are buying", which is still a fact and still costs them money. ## 7. THE COMPANY RECORD — fire early, read last `company_profile`, `company_funding` and `company_firmographics` resolve the company upstream before they answer. They are the slowest calls here by an order of magnitude and the most likely to time out. Start them at the beginning of the run and read them at the end; retry a failure once, because failed calls are refunded and a retry frequently succeeds. Let none of them gate the run: the verdict has to stand on steps 1 to 6 whether or not these three answer. What they are good for: - `company_profile` — headcount and follower count. Headcount tracked against the last time anyone looked is the momentum signal; a single reading is only a size. - `company_funding` — stage, number of rounds, investors, and whether they have been acquiring. A company making acquisitions is spending from strength. - `company_firmographics` — revenue band, employees by function, and the technologies in use. The function split is the useful part: it says where the headcount actually went. It is also the least reliable of the three, so treat a run where it never answers as normal rather than as a blocked one. **Honest gap.** The funding record reliably carries the *stage*, not the *clock*. Round dates and amounts are frequently missing for private companies, so months-of-runway is usually not computable from here. Do not infer it, and do not let a stale round type stand in for one. Where a total-visits figure comes back on the funding record, treat it as a second, independent instrument — it counts all channels, not organic search — and say so when you cite it. Two instruments agreeing is worth more than either alone; two disagreeing is a finding you owe the reader. ## 8. THE VERDICT Score only what you measured, on the four characteristics an outsider can see: GROWTH branded demand YoY, and the 24-month value series EARNED/BOUGHT creative longevity and the paid footprint CONCENTRATION share of value in the homepage and the top pages LOCK-IN evidence of expansion motion in hiring and product surface Then name what you could not see. Growth accounting — new, resurrected and expanded revenue against churned and contracted — is the thing you actually want, and it needs numbers from inside the company. So finish by writing the three questions the data room has to answer, each one aimed at a gap this run identified rather than at a checklist. An outside-in run that ends with the right questions has done its job even when it cannot close the answer. **RULES THAT MAKE THE OUTPUT TRUSTWORTHY** - Cite the tool and the argument behind every number you state. - Distinguish measured (a tool returned it) from inferred (you reasoned to it). - State the strongest evidence AGAINST your conclusion before your conclusion. - Absence of a signal is not evidence of absence. Name which instrument was blind and why. - Coverage is public data and varies by query, country, and date. A thin result is a thin result, not an empty market. - If the data does not settle it, say INCONCLUSIVE and name the one call that would settle it. An unspent budget and an uncertain answer at the same time is a failed run; so is a confident answer the data did not support. ## Deliver The verdict on the four characteristics, each with the series behind it and the call that produced it. The strongest evidence against your own conclusion. The month you decided was a collection artefact and why. And the three questions only a data room can close.