Founding Data Journalist

Turn one of the largest first-party records of what actually works at online checkouts into the studies journalists cite and AI engines quote.

Product.ai is the verified truth layer for shopping: when a person or an AI agent needs to know what is actually true about a purchase, we answer with proof. SimplyCodes is the first proof at scale, the code verification service whose robots run real checkouts and test discount codes so shoppers only see codes that work. It earns about $22 million a year at roughly 60% margins. Profitable. Bootstrapped. Founder-owned since 2009. No outside investors. No board. Fewer than twenty operators, outbuilding companies 10x our size.

Why This Role Exists

We sit on one of the largest first-party records of online savings anywhere: years of robot-run checkouts, code tests, and shopper behavior across hundreds of thousands of stores. When we turn a slice of that record into a study, it gets cited. We tested codes at 500 stores: ours worked 66.8 percent of the time, and codes from the open web worked 23.4 percent. That finding now travels the industry as "only about 1 in 4 coupon codes actually work," and we never had to push it.

The engine those studies feed is compounding. Our studies earned hundreds of pickups and brand mentions across our two brands this year, earned coverage is where our citations in AI answers come from, and the distribution behind it is widening: proactive story pitches, plus a standing reactive lane where reporters come to us for savings data. A widening engine spends studies faster than it can mint them today. This seat exists to be that supply. Every citation makes Product.ai a source those engines trust about a purchase, and that, not traffic, is how we intend to win. It is a founding seat: the pipeline, the standards, and the beat are yours to define.

The System You'll Need to Model

• Data-driven public relations as a supply chain. Proprietary first-party data becomes a citable study; the study rides two distribution lanes, proactive pitches and reactive reporter queries, into press coverage and AI answers. Every serious research-communications shop runs this loop. Ours runs on data nobody else has, and the study is the unit of leverage.
• A real warehouse with real traps, and a moat that was measured rather than scraped. The estate is BigQuery-resident: Search Console, GA4 behavioral events, affiliate commissions, code-test outcomes. Large estates lie to the careless through settlement lags, event-taxonomy breaks, and dishonest denominators, so the craft is finding the trap before the number goes to print. What you print stands on checkouts we measured ourselves, ground truth rivals cannot buy.
• Two citation markets with different physics. Journalists are pitch-driven and relationship-gated; AI answer engines are retrieval-driven and quotability-gated. Being named and being cited are separate outcomes in both, and a study that wins one can lose the other.
• Agents as research staff, and verification as the craft that makes them usable. Warehouse pulls, research sweeps, and first drafts run through agents here; the job is proving what they hand back. What a golden check looks like for a research claim, which tests catch a silent join fanout or a survivorship artifact, when you re-pull a number by hand. Generation is the cheap part. The verdict is yours, and an unverified agent number in print is the one unrecoverable mistake in this seat.
• Cortex, the newsroom you publish inside: the shared AI brain that runs the company and the product family we sell, it answers its own questions from more than 8,600 documents. A newsroom data desk like this ran three people in 2023; the agent tooling is why it is one seat here. The company moves weekly, nobody hands you a brief, and reading where the system is going is part of the job.

If reading that energizes you, keep going. If it feels overwhelming or underspecified, this isn't the right fit.

What You Will Own

• The citable-research pipeline, end to end. From warehouse query to published study to the story brief a pitch rides on: four to six first-party data studies a month across SimplyCodes and Product.ai, so neither distribution lane ever idles. Working rates, code-test outcomes, shopper behavior, survey panels. The craft you must already own: SQL against a production warehouse, statistical honesty about denominators, time windows, and what a number can claim, and ledes an editor would keep. What you grow into here: agents as your research staff, adversarial verification as a production discipline, and AI-engine quotability as a designed property of what you publish.
• Primary reporting. Expert interviews, shopper panels, merchant conversations, with one live reporting thread open at all times, because "we spoke with 10 AI-shopping experts" is a story class no rival can scrape. The measured checkout record stays the ground-truth spine; interviews and panels supply the human context, and every study states which kind of evidence each claim rests on.
• The canonical stat surfaces. The numbers the press and the answer engines quote, kept fresh, sourced, and honest. A stale stat is a broken promise with our name on it. When a study calls for social copy, you supply the data moment; our brand and content team owns that channel.
• The seat, chartered. Within your first quarter you co-sign a seat charter anchored to one machine-checkable number that proves the seat works: published studies that clear a stated method bar, at the cadence this page promises. Citations are the shared scoreboard you move together with the distribution lanes; the studies are yours alone. The charter also writes down what you publish on your own authority and what the founder reads before it goes out - a wrong number in print is the expensive thing here, never the compute.

Who You Are

You reason about numbers the way a good editor reasons about sources: what would make this wrong, what is the honest denominator, what time window the claim covers. You state those without being asked. You can feel the difference between a finding and a query artifact, and when your model of the data is wrong you update fast. You write clearly, because on a team this small the written study is the meeting.

You go to the data first. Agents are your research staff: warehouse queries, research sweeps, and verification passes run through them; you direct the machinery. But you can do every step yourself - pull, sanity checks, draft, pitch note - and that mastery is what lets you trust, or reject, what an agent hands back.

You have personally done all four: pulled the data, found the story, written the piece, and called the source. Your record is newsroom-data-desk shaped: computational journalism, an investigations desk, a brand research team that published real studies, or an independent beat you ran with a scraper and a spreadsheet. "We scored ten thousand machine-generated answers against a rubric" reads like a normal Tuesday to you. We care about the artifact and the reasoning, not where you did it; there is no degree to check.

Who this isn't for. This seat is wrong if you need the data handed to you in a brief; here the story starts in the warehouse, with you holding the query. It is wrong if you analyze but never publish, or publish takes instead of datasets; the portfolio that wins this seat is cited work standing on numbers you pulled yourself. It is wrong if the panel seat, the personal brand, or a masthead logo matters more to you than the dataset under the byline, and it is wrong if this beat is a clip portfolio for your next job; the work here compounds, and a study still cited two years out is the point. And it is wrong if AI-native means a chat window and a subscription; here agents do the grunt work, you own the verdicts, and verification is most of the job. You will be happiest here if your idea of a good month is five studies shipped, two of them cited somewhere that matters, and one interview thread open.

How We Evaluate

We don't run traditional editorial interviews. We evaluate demonstrated performance on work-relevant tasks, in four steps.

Async video screen. Brief and on your own time, about fifteen minutes. We want to see how you think, not how you present.

Calls with company stakeholders. Short calls with the operators you would publish alongside.

Conversation with the founder. How you find a story in a dataset, where the honest denominator lives, and where you push back.

Paid work trial. Three days, paid, on real work in our real environment: a small first-party dataset, find the story, write the study. It stress-tests stack fluency, statistical honesty, and judgment in one artifact. We watch how you get grounded, how you verify what agents hand back, whether the lede survives contact with the data, and whether your self-assessment is honest.

If the work above reads like yours but your resume is unconventional, apply anyway. We hire on the work and the reasoning, not the pedigree.

Compensation & Ownership

Total first-year comp: $250,000 - $350,000 (base, plus performance-based ownership and profit-share programs). Base: $160,000 - $210,000, top of market for a data journalist.

Beyond base: eligibility for the company's ownership and profit-share programs, with grants performance-based and terms discussed at the offer stage, plus 100% company-paid family health premiums and an AI tooling budget steered by return, never capped.

This is a partnership, not a pay grade. The model is built to mint partners: when the company wins, you win, in real and liquid dollars, every year.

Based in Santa Monica, Los Angeles. In person, five days a week. The rooms are real rooms.