Ooak Data growth playbook

Turn the homepage into a two-way entrance. Add a second hero button, "Have company data? See what it's worth", next to "Get in touch", plus a short strip under the hero with the partner promise (up to $1M, 2-4 weeks, about an hour of your time, raw data never leaves Ooak) linking to /partner. Add "Partner with us" to the header and footer. Split the contact page into two clear choices, "I'm building AI" and "My company has data", with the second going to /partner/request. Add a partner call to action at the end of the trust article. Tag every link so you can see which entrance produces calls.

The partner page already does the persuading (pay range, process, FAQ, sample datasets). The gap is routing. The site is already fast (mobile 98, real visitors pass Core Web Vitals), so the problem is that partners can't find the right page, not speed. One extra entrance on the most-visited page and on the contact page puts the offer in front of the founders who arrive through word of mouth, press coverage of data sales, or the AE's outreach and then look Ooak up. It costs a few hours, and every extra signed partner is worth $20K to several hundred thousand dollars of data that labs are already contracted to buy.

Hero buttons:
[Get in touch]  [Have company data? See what it's worth →]

Strip under the hero:
Your company's history is training data. We license it, anonymize it on our own infrastructure and pay you, up to $1M in 2-4 weeks, for about an hour of your time. The raw data never leaves Ooak. → What is my data worth?

Contact page:
What brings you here?
( ) I'm building AI and need real-world data → current form
( ) My company has data to license → /partner/request

• Labs may feel the site now looks like a marketplace

  Keep the hero for AI teams and make the partner entrance clearly secondary, as a second button and strip.

• More low-fit requests (too small, no rights)

  Show the fit criteria (team size, years of history, number of tools) next to the button.

• Your websiteHomepage (checked November 9, 2026): title "Ooak Data — Applied AI Research Lab", H1 "The data infrastructure for frontier AI"; three blocks (Real Companies, Digital Twin, RL Environments) for frontier AI labs, enterprise AI teams and AI startups; "Backed by Y Combinator". The only calls to action are "Explore our work", "Read our research" and "Get in touch" (/contact), and the footer links Careers, Contact, Privacy, Terms and Legal. Nothing on the homepage links to /partner, the page for companies that want to license their data.Source
• Your websiteContact page: "Get in touch" invites people building AI to describe their project, with [email protected] as the fallback. It does not mention data partners or link to /partner.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• HiringChief of Staff, Paris: contracts signed with three frontier AI labs; raw data to an RL-ready environment in about three weeks, three times faster than at launch; contracted delivery volume growing by an order of magnitude; "The three of us are the bottleneck"; customers are American and data partners are European; the role covers closing data partnerships and new vertical launches.Source
• PageSpeedHomepage, November 10, 2026: mobile performance 98, desktop 100, accessibility 95; real visitors pass Core Web Vitals (field LCP 1.06 s, CLS 0); the one accessibility issue is touch targets without enough size or spacing.Source

Add a proof band to /partner: "30+ companies licensed", "0 re-identification incidents", "EU processing available", "Exclude any source before anything moves", "Raw data deleted within 30 days, confirmed in writing". Add two or three short anonymized partner quotes (role, company size and sector only) and a link to the full chain-of-custody article.

The first question founders ask is who will see their data. Answering it with numbers on the decision page removes the main objection before the call, and it sets Ooak apart from micro1, whose page promises payouts but is English-first and US-focused.

• Partners don't want to be quoted

  Use role, sector and size only, and launch with the numbers if quotes are slow.

• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• Competitor researchmicro1's data partnerships page: "Monetize your company's operational data", paying "$100k-3M+"; wants 30+ full-time employees and data "primarily in English", with demand strongest in the US, UK and Canada; a "Get estimate" estimator; payment within 30 days of acceptance; a referral programme paying $50,000 per enterprise data deal up to $700,000 and $75,000 for defunct companies. micro1 announced a $1 billion, 12-month enterprise data commitment on October 9, 2026.See the research Source

Publish the real schedule as a simple timeline: call (day 0), scope and contract (week 1), extraction and anonymization (weeks 2-3), approval of the sample, then payment in a stated number of days, with any split such as a portion on signature. Add the same timeline to the contract summary the AE sends after the first call.

A concrete date turns a vague promise into a planning input, and it competes directly with the one specific number micro1 offers.

• Promising a date the pipeline can't meet

  Publish the worst case, not the median.

• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• Competitor researchmicro1's data partnerships page: "Monetize your company's operational data", paying "$100k-3M+"; wants 30+ full-time employees and data "primarily in English", with demand strongest in the US, UK and Canada; a "Get estimate" estimator; payment within 30 days of acceptance; a referral programme paying $50,000 per enterprise data deal up to $700,000 and $75,000 for defunct companies. micro1 announced a $1 billion, 12-month enterprise data commitment on October 9, 2026.See the research Source
• HiringChief of Staff, Paris: contracts signed with three frontier AI labs; raw data to an RL-ready environment in about three weeks, three times faster than at launch; contracted delivery volume growing by an order of magnitude; "The three of us are the bottleneck"; customers are American and data partners are European; the role covers closing data partnerships and new vertical launches.Source

Add a booking calendar after the request form, showing the EU AE's and US counterpart's slots by time zone, and a plain promise: "A real person replies within one working day." Send an instant confirmation email with the three questions the call will cover and a link to the trust article.

Fast, specific next steps raise show-up and close rates, and the AE is measured on building pipeline from zero, so every request that books itself saves outreach time.

• Low-fit companies book calls

  Ask team size and years of history before showing slots.

• Your websiteRequest a call page: "Tell us about your company. We review every request and get back to you as soon as possible." No reply time or booking slot is given.Source
• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• HiringData Partnerships - Account Executive (EU), Paris: covers "UK, France, Spain, Germany", owns the "full deal cycle" with founders and C-level executives and builds pipeline from scratch; French strongly preferred, Spanish or German a plus; works alongside a US counterpart. The post says Ooak has "8 figures contracts signed with frontier AI labs".Source

Add a two-button block to the end of every post: "Building agents? Request a sample environment" (to /contact with a tag) and "Have company data? See what it's worth" (to /partner#estimator). Add a light email sign-up for new research, since the site already publishes an RSS feed.

Readers who finish a 10-12 minute technical post are the most engaged visitors Ooak gets, and today they leave without a next step.

• Calls to action make research look like marketing

  Keep the block small and below the references.

• #1 Give data partners a front door: the homepage and contact page speak only to AI buyers, and nothing on the homepage links to /partner

• Your websiteResearch hub: four posts (January 22, February 11, March 18 and October 27, 2026), no calls to action and no newsletter sign-up. "Why synthetic benchmarks are holding AI back" cites only third-party studies (for example diagnostic accuracy falling from 82% on vignettes to 62.7% in multi-turn dialogue) and reports no results from Ooak's own environments.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• Your websiteHomepage (checked November 9, 2026): title "Ooak Data — Applied AI Research Lab", H1 "The data infrastructure for frontier AI"; three blocks (Real Companies, Digital Twin, RL Environments) for frontier AI labs, enterprise AI teams and AI startups; "Backed by Y Combinator". The only calls to action are "Explore our work", "Read our research" and "Get in touch" (/contact), and the footer links Careers, Contact, Privacy, Terms and Legal. Nothing on the homepage links to /partner, the page for companies that want to license their data.Source

Add a cookieless, EU-hosted analytics tool (to stay consistent with Ooak's privacy stance) and track five events: partner page view, estimator started, estimator finished, request submitted and call booked, split by entry source. Review the funnel weekly with the sellers.

Every other idea in this playbook changes the partner funnel. Without measurement, none of the changes can be judged.

• Analytics conflicts with the privacy message

  Use a cookieless tool and say so in the privacy policy.

• Site scanChanges since December 2025: the site moved from Webflow to SvelteKit and dropped Google Analytics and Google Tag Manager, with no analytics tool detected now; the headline changed from "Your AI agent's reality check." to "The data infrastructure for frontier AI"; no LinkedIn, X or other social profile is linked from the site; no news mentions in the last 30 days.Source
• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• Your websiteRequest a call page: "Tell us about your company. We review every request and get back to you as soon as possible." No reply time or booking slot is given.Source

Publish /fr/partenaire, /de/partner and /es/socio, local versions of the partner page, request form and custody article, written by native speakers rather than machine-translated. Lead each with the EU processing option and the jurisdiction check the FAQ already describes. Add hreflang between versions and use them in the AE's outbound.

A CEO deciding whether to license years of internal messages is more likely to trust a page in their own language, and Ooak can claim the non-English European market before English-first rivals adapt.

• Legal terms translated loosely

  Counsel signs off each version, and the English contract remains the binding text unless agreed otherwise.

• #1 Give data partners a front door: the homepage and contact page speak only to AI buyers, and nothing on the homepage links to /partner

• HiringData Partnerships - Account Executive (EU), Paris: covers "UK, France, Spain, Germany", owns the "full deal cycle" with founders and C-level executives and builds pipeline from scratch; French strongly preferred, Spanish or German a plus; works alongside a US counterpart. The post says Ooak has "8 figures contracts signed with frontier AI labs".Source
• Competitor researchmicro1's data partnerships page: "Monetize your company's operational data", paying "$100k-3M+"; wants 30+ full-time employees and data "primarily in English", with demand strongest in the US, UK and Canada; a "Get estimate" estimator; payment within 30 days of acceptance; a referral programme paying $50,000 per enterprise data deal up to $700,000 and $75,000 for defunct companies. micro1 announced a $1 billion, 12-month enterprise data commitment on October 9, 2026.See the research Source
• Your websiteSitemap (12 URLs, checked November 9, 2026): home, research and four posts, partner, careers, contact and three legal pages. No page for AI buyers beyond /contact, no industry pages, no case studies and no French or other-language pages (/fr returns 404).Source
• HiringChief of Staff, Paris: contracts signed with three frontier AI labs; raw data to an RL-ready environment in about three weeks, three times faster than at launch; contracted delivery volume growing by an order of magnitude; "The three of us are the bottleneck"; customers are American and data partners are European; the role covers closing data partnerships and new vertical launches.Source

Build a short adviser kit: a one-page explainer of how a data licence can add value to an estate, what rights checks Ooak runs, the deletion promise, and a single-contact intake. Have the AE meet insolvency firms, restructuring lawyers and wind-down advisers in Paris, London, Madrid and Berlin. Offer a fast valuation for any estate they bring.

One adviser sees many closing companies a year, so each relationship is a repeat source of partners. A valuation that adds money for creditors is easy for them to say yes to.

• Estate data may carry rights problems

  Counsel's rights check before any valuation is promised.

• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• PressMIT Technology Review (October 13, 2026) on labs buying data from shuttered startups: SimpleClosure's Asset Hub has handled over 150 deals, 47% now including code or data; a rival broker, Sunset, is growing a data resale business; Surge AI buys data to build RL environments; one airline's data sold for $10 million; labs want archives with disagreements and hard judgment calls; a lawyer calls the data "legally radioactive" and a law professor says "Employers should have express consent before selling workers' data". Ooak is not mentioned.Source
• Competitor researchSimpleClosure, a startup shutdown service: its AssetHub lets founders "Sell or license source code, workspace data, and other startup assets to vetted buyers."See the research Source
• HiringData Partnerships - Account Executive (EU), Paris: covers "UK, France, Spain, Germany", owns the "full deal cycle" with founders and C-level executives and builds pipeline from scratch; French strongly preferred, Spanish or German a plus; works alongside a US counterpart. The post says Ooak has "8 figures contracts signed with frontier AI labs".Source

Publish /partner/refer with a simple rule, such as a percentage of the licence value paid after delivery, and a unique link per referrer. Invite every past partner, Ooak's investors and the advisers from the wind-down channel. Keep the terms clear: paid only on signed and delivered deals.

Introductions from a peer who has been paid carry the trust that cold outreach lacks, and a fee aligned to deal value keeps the cost proportional.

• Fee-driven referrals of companies without rights to their data

  Pay only after the rights check and delivery.

• Competitor researchmicro1's data partnerships page: "Monetize your company's operational data", paying "$100k-3M+"; wants 30+ full-time employees and data "primarily in English", with demand strongest in the US, UK and Canada; a "Get estimate" estimator; payment within 30 days of acceptance; a referral programme paying $50,000 per enterprise data deal up to $700,000 and $75,000 for defunct companies. micro1 announced a $1 billion, 12-month enterprise data commitment on October 9, 2026.See the research Source
• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source

Offer investors a standing service: a free data valuation for any portfolio company that is winding down or pivoting, a one-page note for the board pack, and the referral fee from idea 9 paid to the company rather than to the investor where that is cleaner. Start with the YC network and five European funds.

Investors want to recover value for every company, and one platform team can introduce many partners.

• Investors worry about reputation if a sale goes wrong

  Lead with the custody process and the deletion confirmation.

• #9 Launch a referral fee for founders, advisers and investors who introduce a data partner

• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• HiringYC Work at a Startup (checked November 9, 2026): Summer 2026 batch, team of 5, 11 open roles: Lead ML Engineer, Lead SWE, Lead Counsel - Data Licensing & Privacy, Data Partnerships - Account Executive (EU), Data Quality Specialist, Operations Manager - Data & AI, Head of Engineering, Chief of Staff, Senior Data Engineer and Senior Software Engineer in Paris, and GTM Manager (US) in San Francisco.Source
• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source

Publish one page per proven sector: what kinds of records sell (dispatch logs, project tickets, support threads), a range of past deal sizes, the tools typically connected, a redacted before-and-after example, and sector-specific exclusions. Add new sectors as the chief of staff's "new vertical launches" happen.

Specific examples answer "would my company qualify?", and sector pages give the AE something targeted to send in outbound.

• A redacted example could hint at the source company

  Use synthetic stand-ins for the example and get the source partner's approval.

• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• Your websiteSitemap (12 URLs, checked November 9, 2026): home, research and four posts, partner, careers, contact and three legal pages. No page for AI buyers beyond /contact, no industry pages, no case studies and no French or other-language pages (/fr returns 404).Source
• HiringChief of Staff, Paris: contracts signed with three frontier AI labs; raw data to an RL-ready environment in about three weeks, three times faster than at launch; contracted delivery volume growing by an order of magnitude; "The three of us are the bottleneck"; customers are American and data partners are European; the role covers closing data partnerships and new vertical launches.Source

Rebuild the estimator around Ooak's four value drivers (headcount, years of history, number of tools, how connected they are). Show a range instantly with no email, then offer "Get an exact figure", which opens the request form already filled in with the answers and the range.

A free number is the hook for curious founders, and passing the answers to the form gives the AE a qualified, sized lead before the call.

• High ranges create disappointment

  Show a conservative range and the drivers that raise it.

• #6 Measure the partner funnel again with privacy-friendly analytics

• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• Competitor researchNyne's company data page: for operating companies with 20+ US W-2 employees and three years in business; a "typical partnership" of $100K+; a free three-question estimator (team size, years in business, headquarters) that answers in about 30 seconds with no email; an offer to beat a qualifying written offer.See the research Source
• Your websiteRequest a call page: "Tell us about your company. We review every request and get back to you as soon as possible." No reply time or booking slot is given.Source

Publish /for-ai-teams: what is in a digital twin (sources, years, people, tool mix), the environment format and grading, how tasks are calibrated against frontier models, delivery options (datasets, environments, evals), the delivery chain from evaluation to delivery, and a "Request a sample environment" form that asks the GTM manager's diagnostic questions.

Technical buyers qualify vendors before they take a call. A spec sheet shortens the first meeting and filters for real projects, and a sample is the strongest proof a data vendor can give.

• Revealing too much to competitors

  Publish the shape and share the sample under NDA.

• HiringGTM Manager (US), San Francisco: "Own the full sales cycle with frontier labs and agent companies", selling to ML researchers, post-training leads and AI founders; diagnoses needs for "agent training data, RL environments, and evals" (workflows, tools, task difficulty, volume); six- to eight-figure deals; relays which "environments, domains, and data types" buyers want next; experience at a data vendor such as Scale, Surge, Mercor, Turing or Snorkel is a plus.Source
• Your websiteSitemap (12 URLs, checked November 9, 2026): home, research and four posts, partner, careers, contact and three legal pages. No page for AI buyers beyond /contact, no industry pages, no case studies and no French or other-language pages (/fr returns 404).Source
• HiringOperations Manager - Data & AI, Paris: owns each partner delivery "from signature to shipment" across evaluation, ingestion, anonymization, enrichment and delivery, with delivery scaling "aggressively over the next months", and automates the manual steps.Source

Write one data-led post: take a set of digital-twin tasks (multi-step, multi-tool, across the sample sectors), run several publicly available models and agents on them, and report pass rates against their public benchmark scores, with failure examples. Keep the companies anonymous and the tasks private. Publish the method and the headline numbers.

Original evidence is what researchers share and what labs remember, and it turns Ooak's thesis into proof. It also feeds the GTM manager's first conversations.

• Lab customers dislike public scores for their models

  Report results across anonymized models, or only for models whose terms allow it.

• Your websiteResearch hub: four posts (January 22, February 11, March 18 and October 27, 2026), no calls to action and no newsletter sign-up. "Why synthetic benchmarks are holding AI back" cites only third-party studies (for example diagnostic accuracy falling from 82% on vignettes to 62.7% in multi-turn dialogue) and reports no results from Ooak's own environments.Source
• Your websiteHomepage (checked November 9, 2026): title "Ooak Data — Applied AI Research Lab", H1 "The data infrastructure for frontier AI"; three blocks (Real Companies, Digital Twin, RL Environments) for frontier AI labs, enterprise AI teams and AI startups; "Backed by Y Combinator". The only calls to action are "Explore our work", "Read our research" and "Get in touch" (/contact), and the footer links Careers, Contact, Privacy, Terms and Legal. Nothing on the homepage links to /partner, the page for companies that want to license their data.Source
• HiringGTM Manager (US), San Francisco: "Own the full sales cycle with frontier labs and agent companies", selling to ML researchers, post-training leads and AI founders; diagnoses needs for "agent training data, RL environments, and evals" (workflows, tools, task difficulty, volume); six- to eight-figure deals; relays which "environments, domains, and data types" buyers want next; experience at a data vendor such as Scale, Surge, Mercor, Turing or Snorkel is a plus.Source

Publish /trust: the licence model (scope, duration, revocation, ownership retained), the anonymization standard in plain words, the chain of custody, a sample deletion certificate, sub-processor policy (no subcontractors on raw data), EU processing, and a status line for SOC 2 and ISO 27001 as they progress. Link it from /partner and /for-ai-teams.

Both sides' lawyers ask the same questions. One page answers them before the call, shortens legal review and makes Ooak look like the safer choice.

• Public commitments create liability

  Publish only what the contracts already promise.

• #2 Move the trust proof from the October research post onto the partner page

• HiringLead Counsel - Data Licensing & Privacy, Paris: owns EU and US licensing templates and negotiations with data partners and frontier labs, defines "anonymized" in law for the pipeline and defends it to partner counsel, lab counsel or regulators, and leads SOC 2 and ISO 27001 end to end.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• PressMIT Technology Review (October 13, 2026) on labs buying data from shuttered startups: SimpleClosure's Asset Hub has handled over 150 deals, 47% now including code or data; a rival broker, Sunset, is growing a data resale business; Surge AI buys data to build RL environments; one airline's data sold for $10 million; labs want archives with disagreements and hard judgment calls; a lawyer calls the data "legally radioactive" and a law professor says "Employers should have express consent before selling workers' data". Ooak is not mentioned.Source

Turn the audit into a one-page quality card attached to every delivery: leak-scan results, the human-review sample size, entity consistency rate, structural checks passed and known limits. Publish aggregate numbers (datasets shipped, leaks caught before release, incidents since launch) on /trust and /partner.

Labs get evidence they would otherwise build themselves, and partners see that privacy is measured rather than promised. The same work serves sales on both sides.

• Publishing caught leaks sounds alarming

  Frame them as caught before release, which is the point of the process.

• HiringData Quality Specialist, Paris: writes acceptance criteria and error thresholds that "barely exist" today; audits each delivery for PII leakage, replacement consistency, structural integrity and whether business value survived anonymization; tracks quality metrics per dataset and per vendor; "Nothing ships without passing your bar."Source
• HiringOperations Manager - Data & AI, Paris: owns each partner delivery "from signature to shipment" across evaluation, ingestion, anonymization, enrichment and delivery, with delivery scaling "aggressively over the next months", and automates the manual steps.Source
• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source

Write a short kit, reviewed by counsel for each market: a plain-language notice to staff and alumni explaining what is licensed, what is excluded, how identities are replaced and how to ask questions. Include a FAQ, an optional opt-out for private channels and a contact at Ooak. Offer it during scoping, and in local languages once idea 7 ships.

It removes the most personal objection a founder has, turns a press risk into a differentiator, and supports the legal posture counsel is hired to define.

• Notices trigger staff objections that stop the deal

  Make it optional and let the partner choose timing and scope.

• PressMIT Technology Review (October 13, 2026) on labs buying data from shuttered startups: SimpleClosure's Asset Hub has handled over 150 deals, 47% now including code or data; a rival broker, Sunset, is growing a data resale business; Surge AI buys data to build RL environments; one airline's data sold for $10 million; labs want archives with disagreements and hard judgment calls; a lawyer calls the data "legally radioactive" and a law professor says "Employers should have express consent before selling workers' data". Ooak is not mentioned.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• HiringLead Counsel - Data Licensing & Privacy, Paris: owns EU and US licensing templates and negotiations with data partners and frontier labs, defines "anonymized" in law for the pipeline and defends it to partner counsel, lab counsel or regulators, and leads SOC 2 and ISO 27001 end to end.Source

Make the close a routine: send the deletion certificate with a short thank-you from the founders, ask two questions (would you recommend us, may we quote you by role and sector), and ask for two founders they know who might qualify, with the referral fee from idea 9. Log answers for the proof band in idea 2.

Founders know other founders, and a just-paid partner whose data was handled as promised is Ooak's most credible salesperson.

• Asking too soon after a sensitive project

  Send after the certificate, when the work is fully complete.

• #9 Launch a referral fee for founders, advisers and investors who introduce a data partner

• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source

Offer each lab a refresh plan: every quarter, recalibrate difficulty against the newest models, retire saturated tasks, add new tasks from fresh partner data, and report how the lab's agents moved. Price it as an annual subscription on top of the initial licence.

It ties revenue to the very problem Ooak describes (saturation) and turns eight-figure one-off contracts into recurring expansion.

• Delivery capacity is already stretched

  Automate recalibration before selling at scale.

• Your websiteHomepage (checked November 9, 2026): title "Ooak Data — Applied AI Research Lab", H1 "The data infrastructure for frontier AI"; three blocks (Real Companies, Digital Twin, RL Environments) for frontier AI labs, enterprise AI teams and AI startups; "Backed by Y Combinator". The only calls to action are "Explore our work", "Read our research" and "Get in touch" (/contact), and the footer links Careers, Contact, Privacy, Terms and Legal. Nothing on the homepage links to /partner, the page for companies that want to license their data.Source
• HiringGTM Manager (US), San Francisco: "Own the full sales cycle with frontier labs and agent companies", selling to ML researchers, post-training leads and AI founders; diagnoses needs for "agent training data, RL environments, and evals" (workflows, tools, task difficulty, volume); six- to eight-figure deals; relays which "environments, domains, and data types" buyers want next; experience at a data vendor such as Scale, Surge, Mercor, Turing or Snorkel is a plus.Source
• HiringChief of Staff, Paris: contracts signed with three frontier AI labs; raw data to an RL-ready environment in about three weeks, three times faster than at launch; contracted delivery volume growing by an order of magnitude; "The three of us are the bottleneck"; customers are American and data partners are European; the role covers closing data partnerships and new vertical launches.Source

Offer operating partners a recurring licence: after the first delivery, Ooak pulls a new read-only slice each quarter (same scope, same exclusions), anonymizes it with the same entity dictionary so the twin stays consistent, and pays a quarterly fee. Sell labs "living twins" that extend over time.

Recurring supply lowers acquisition cost per dataset, the consistent entity dictionary makes the data more valuable over time, and partners get predictable income.

• Retaining a mapping table for consistency conflicts with the deletion promise

  Design the refresh so consistency survives without keeping raw data, and have counsel approve it before piloting.

• #15 A public trust centre: the licence model, the legal definition of "anonymized", a sample deletion certificate and SOC 2 and ISO 27001 progress

• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• Competitor researchmicro1's data partnerships page: "Monetize your company's operational data", paying "$100k-3M+"; wants 30+ full-time employees and data "primarily in English", with demand strongest in the US, UK and Canada; a "Get estimate" estimator; payment within 30 days of acceptance; a referral programme paying $50,000 per enterprise data deal up to $700,000 and $75,000 for defunct companies. micro1 announced a $1 billion, 12-month enterprise data commitment on October 9, 2026.See the research Source
• Your websitePartner page: "Turn your data into revenue" and "You get paid for something you already have"; up to $1M, 2-4 weeks, no engineering work; deals so far from $20K to "several hundred thousand dollars"; a "What is my data worth?" estimator that gives a range; "We pay fast" with no timeframe; three anonymized sample datasets (edtech: 400+ people, 6 years, 15 tools; construction tech: 200+ people, 4 years, 7 tools; fleet logistics: 15+ people, 4 years, 7 tools); an FAQ saying a director, liquidator or administrator can sign for a company in wind-down; "zero re-identification incidents to date". English only, with no testimonials and no mention of GDPR, the EU, SOC 2, a referral programme or how many companies Ooak has worked with.Source

Launch a small public leaderboard: a held-out set of multi-step, multi-tool tasks drawn from twins whose licences allow it, scored on Ooak's infrastructure (tasks never released), refreshed each quarter to avoid saturation. Publish the method, a few example tasks, and scores for publicly available agents. Labs can request private runs.

A respected benchmark makes Ooak the reference point for "real work" evaluation, brings in buyers who want private runs, and gives partners a visible reason their data matters.

• Distracts the team from contracted delivery

  Start with 100 tasks and a quarterly cadence.

• Lab customers dislike public rankings

  Let models opt in, and offer private runs.

• #14 Publish Ooak's own numbers: how frontier models score on real-company tasks compared with public benchmarks

• Your websiteResearch hub: four posts (January 22, February 11, March 18 and October 27, 2026), no calls to action and no newsletter sign-up. "Why synthetic benchmarks are holding AI back" cites only third-party studies (for example diagnostic accuracy falling from 82% on vignettes to 62.7% in multi-turn dialogue) and reports no results from Ooak's own environments.Source
• Your websiteHomepage (checked November 9, 2026): title "Ooak Data — Applied AI Research Lab", H1 "The data infrastructure for frontier AI"; three blocks (Real Companies, Digital Twin, RL Environments) for frontier AI labs, enterprise AI teams and AI startups; "Backed by Y Combinator". The only calls to action are "Explore our work", "Read our research" and "Get in touch" (/contact), and the footer links Careers, Contact, Privacy, Terms and Legal. Nothing on the homepage links to /partner, the page for companies that want to license their data.Source
• HiringGTM Manager (US), San Francisco: "Own the full sales cycle with frontier labs and agent companies", selling to ML researchers, post-training leads and AI founders; diagnoses needs for "agent training data, RL environments, and evals" (workflows, tools, task difficulty, volume); six- to eight-figure deals; relays which "environments, domains, and data types" buyers want next; experience at a data vendor such as Scale, Surge, Mercor, Turing or Snorkel is a plus.Source

For operating partners, offer an add-on: the right to run their own (or a vendor's) agents against the twin of their company, with Ooak-built tasks from their real workflows, before going live. Price it as a credit against the licence fee, so the partner both earns and gets a test bed.

It turns a one-sided sale into a two-sided relationship, gives Ooak an enterprise product built on work it already does, and gives a hesitant partner a practical reason to say yes.

• Exclusivity conflicts with lab licences

  Write partner self-use into the licence from the start.

• #20 Recurring licences for operating companies: a quarterly refresh of the digital twin instead of a one-off sale

• Your websiteHomepage (checked November 9, 2026): title "Ooak Data — Applied AI Research Lab", H1 "The data infrastructure for frontier AI"; three blocks (Real Companies, Digital Twin, RL Environments) for frontier AI labs, enterprise AI teams and AI startups; "Backed by Y Combinator". The only calls to action are "Explore our work", "Read our research" and "Get in touch" (/contact), and the footer links Careers, Contact, Privacy, Terms and Legal. Nothing on the homepage links to /partner, the page for companies that want to license their data.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source

Package a multilingual offer: twins and environments built from non-English European partners, processed only in the EU, sold to European agent companies and enterprise AI teams as well as to labs that want multilingual coverage. Test it with a handful of European agent startups before building more.

It turns the language of Ooak's supply from a limitation into a differentiated product, and gives the EU partner pipeline a second set of buyers.

• European buyers have smaller budgets than frontier labs

  Price per environment and keep labs as the priority.

• #7 French, German and Spanish partner pages for the four markets the EU AE covers

• HiringChief of Staff, Paris: contracts signed with three frontier AI labs; raw data to an RL-ready environment in about three weeks, three times faster than at launch; contracted delivery volume growing by an order of magnitude; "The three of us are the bottleneck"; customers are American and data partners are European; the role covers closing data partnerships and new vertical launches.Source
• Your websiteResearch post "How we protect a company's history when training AI" (October 27, 2026): Ooak has worked with more than 30 companies of 50 to 500 people; "the raw data never leaves Ooak"; read-only extraction, consistent stand-ins across Slack, email and Jira, automated leak scans plus human review, an EU option for GDPR-aligned processing, sources that can be excluded up front, about three weeks end to end. It is written for founders but ends only with "Continue reading" links, with no link to /partner.Source
• Competitor researchmicro1's data partnerships page: "Monetize your company's operational data", paying "$100k-3M+"; wants 30+ full-time employees and data "primarily in English", with demand strongest in the US, UK and Canada; a "Get estimate" estimator; payment within 30 days of acceptance; a referral programme paying $50,000 per enterprise data deal up to $700,000 and $75,000 for defunct companies. micro1 announced a $1 billion, 12-month enterprise data commitment on October 9, 2026.See the research Source