Polymarket: the raw research

# New Recording 9 Date: 2026-07-09 Duration: 5 minutes Speakers: Speaker A, Speaker B, Speaker C --- **Speaker A** [0:04] How you been? I got. I was actually talking to Adam, so I just passed you in. How you doing? **Speaker B** [0:08] Oh, hey, guys. Good. **Speaker A** [0:13] So I wanted to just catch up **Speaker B** [0:15] to see where you guys are at with the product. And then I'm at Polymer right now helping them launch all the CPD and Programmatic internally. **Speaker A** [0:25] Oh, incredible. **Speaker B** [0:27] Yes. **Speaker A** [0:28] At Polymarket. No kidding. So we're spending quite a bit of **Speaker B** [0:32] money, you know, like tens of millions of dollars. **Speaker A** [0:35] Yeah. **Speaker B** [0:36] I just got out of the meeting and you guys popped into my head and I said, there's proud, there's multiple DSP partners, multiple data sources. It's, you know, whole conundrum of different things. I kind of remember where you guys were going directionally with the product. And I was like, oh, this might be a fit to help them optimize out, get log data, you know, all the things you guys previously discussed. But I don't know if you're still in that direction or if it is a fit. So I just wanted to check. Well, so. **Speaker A** [1:04] So, I mean, I'll start and Adam's film. But so it is, I think, expanded in the scope of what we're trying to achieve on an enterprise level. And I say that because, geez, it went from, you know, we can have this like, kind of like retention widget, as you saw in the beginning. That can kind of help you, you know. Oh, look, I can kind of figure out, like, what's the most optimal way to run the campaign if it's recalibrated. You saw maybe the release. It's called the decision economics. But the same tech. We're now kind of like we're talking to Bill Wise, for example, who's an investor, put a quarter million bucks into the last round. You may have saw to be like, hey, can you guys be effectively the, you know, can we. Can we slowly become the Nielsen for agentic media? Meaning, like constantly, just one case study by another getting into the work stream. Start to value every decision these agents are making while. And I know you have a back end blockchain making sure that you have received so that it's never like in question and it's actually governed and has like, you know, mathematical proof of what happened. So we're kind of like, those are big, like kind of founder to founder conversations. We have like two or three of those going on. Peter. And so I guess my first question to you is if you tell us what you're trying to solve for, I think that will kind of open up **Speaker B** [2:20] for what we can do. **Speaker A** [2:23] These are conversations becoming bigger than Like **Speaker B** [2:24] a product or a widget. Okay. So effectively, you know, one of the guys that I worked with that disrupted one of their fundament to be a head of growth and I spoke to him about scaling the connected TV and the programmatic. And we came in probably 30 days ago, they had a whole slew of different partners they wanted to test. We started flighting the Programmatic and the CTV for the World cup and for the UFC fight at the White House. And now they're going into the NFL and the football season. They're working with every trade desk, stack, adapt, you name it, it's on a media plan. And the budgets are collectively hundreds of millions of dollars is being spent through these channels. So you know the attribution problem they have, okay, you know, they're using Apps Flyer, it's an app you download, you do a first deposit, you're running across a bunch of different networks. What's working, what's not working? Where do you do the optimization? Where should the money be spent across which partners, which channels, where's the granularity of the log? So like, you know, it's less of just fighting media at this point and now it's understanding how the media is working and defining which partners are going to receive allocation. It's not following through tool assist from like a PMX perspective or APEX or what have you. It is more of a consultative project. So we have like five or six people on our team that are helping them stand up. Basically like an in house variant of this, you know, so they're working with agencies today to maybe flight and run media campaigns. But I instantly thought of you guys because I'm like, okay, as they're going through this discovery phase, is there a plug in here? Because I was trying to find an opportunity where it makes sense because we do have, we can get all the raw access logs, we can do all the integrations, you can do anything that you need to do. But yeah, it's effectively how do they stand up the right next version of what they need internally? **Speaker A** [4:23] Wow. Seems like in our wheelhouse, Adam. **Speaker C** [4:27] Yeah, I need to think about it. But yeah, no, it's definitely something, I mean that kind of like fits our wheelhouse for sure. I need to know a little bit more like details of the inputs and outputs. But it's definitely, you know, it also sounds. Feels like it'll be like a, like kind of a foot in the door to our capability. Right? **Speaker A** [4:44] Yeah. **Speaker B** [4:45] And I'm going through all the log stuff now manually, so I'm like downloading the apps, fire logs. **Speaker C** [4:49] I'm just, I mean, that's stuff I can also, like, help. Help you do behind the scenes and, and mechanize, you know, with a custom, custom, like, harness built on top of our platform, you know, so, **Speaker B** [5:04] yeah, whatever. Whatever you guys think. I just know I can get the, you know, I'm going to be in their office the next two days, and then next week I'm going to be in New York as well. So I'm happy to, you know, I know you guys probably need more, have more questions.

########## LANE corporate ########## ## Polymarket corporate + growth (2025–Aug 2026) **Money** - Oct 7 2025: ICE invests **up to $2B at ~$8B pre-money**; ICE becomes global distributor of Polymarket event data + tokenization partner ([ICE IR](https://ir.theice.com/press/news-details/2025/ICE-Announces-Strategic-Investment-in-Polymarket/)). - Mar 27 2026: ICE adds **$600M**, capping ~$1.6B deployed ([Reuters](https://www.reuters.com/business/nyse-parent-intercontinental-exchange-invests-600-million-polymarket-2026-03-27/)). - Apr 2026: closed round at **$15B** (included ICE's $600M). - **Aug 4 2026 (today): in talks to raise ~$1B at >$20B**, Bloomberg first, CNBC confirmed ([CNBC](https://www.cnbc.com/2026/08/04/polymarket-seeks-fundraising-round-at-more-than-20-billion-valuation.html)). - Revenue: told CNBC late June 2026 **annualized revenue "well above $1B."** **US re-entry** - Jul 21 2025: bought **QCEX** (QCX DCM + clearinghouse) for **$112M** after DOJ/CFTC probes dropped ([PR](https://www.prnewswire.com/news-releases/polymarket-acquires-cftc-licensed-exchange-and-clearinghouse-qcex-for-112-million-302509626.html)). CFTC amended order Sept 2025; [no-action letter 25-48](https://www.cftc.gov/csl/25-48/download) Dec 11 2025. - US platform debuted **waitlisted Dec 2025**; app left beta **March 2026**; **waitlist dropped / open iOS May 12 2026**. Sportico: the product team "fumbled" the 2025 NFL season — that miss is why NFL 2026 is loaded. **Volume / share (the number that matters for the pitch)** - July 2026 = record **$50.6B** combined: Kalshi **$37.7B (74.5%)**, Polymarket intl **$7.9B (−26%)**, **Polymarket US $5B (+54%)** ([Block Stream](https://blockstreammedia.com/2026/08/03/prediction-markets-hit-record-50-6b-july-volume/)). Open interest fell to $1.2B post-World Cup. - Among CFTC-regulated venues: Kalshi ~83%, **Polymarket US ~10%** (8.5% in May), Robinhood's Rothera 4.6%, Crypto.com's Nadex 2.4% ([Sportico](https://www.sportico.com/business/sports-betting/2026/prediction-market-data-volume-polymarket-kalshi-world-cup-1234939529/)). - Reach vs. dollars: Apr 2026 ~678K unique Polymarket users vs. ~85K implied Kalshi — more people, less money per ticket ([pm.wiki](https://pm.wiki/news/kalshi-overtakes-polymarket-volume) — affiliate source, treat as directional). - Product: **custom parlays/"combos"** rolling out — Sportico notes parlays inflate notional, so Polymarket has been at an accounting disadvantage vs. Kalshi. **Marketing posture (direct pitch surface)** - World Cup: industry spent **hundreds of millions** on ads + signup bonuses. Kalshi ~**$32M** national linear TV YTD (iSpot); **Polymarket aired 45 ads during live World Cup matches on Fox/Telemundo vs Kalshi's 36** — they are outspending on in-game inventory and still losing share. That is the allocation question Peter is living in. - First brand campaign ever — "Questions Are Everything," agency **Better Half** with **Rick Rubin** ([LBB](https://lbbonline.com/news/polymarket-rick-rubin-world-cup)). No public Publicis media AOR announcement — the Publicis presence reads as an embedded in-house build, not an agency award (inferred). - Influencer spend is a live liability: CMO **Matthew Modabber** sent **$350K+ from a personal PayPal** to creators posting without disclosure (Politico, via [Verge, Jun 5 2026](https://www.theverge.com/policy/943986/polymarkets-cmo-sent-thousands-to-influencers-like-nick-shirley-and-riley-gaines)). Attribution hygiene has a reputational edge here, not just a budget one. **People in the room** - **Shayne Coplan** — founder/CEO. **Matthew Modabber** — CMO. **Jacob Shamberger** — also on the hiring side of growth. - **jenner@ = Jenner Kearns**, leads **performance marketing**, joined ~Apr 2026, ex-VP Strategy at Dysrupt (perf agency) — he is the buyer of a spend-allocation story. - **anthony.khaiat@ = Anthony Khaiat**, Senior Data Scientist, Research — joined **Feb 2026 from Kalshi** (data scientist there Sep 2024–Feb 2026), MIT MBAn. He will grade the method, and he knows the rival's numbers. **Distribution / rivalry** - Official prediction market of **X** since Jun 6 2025, with xAI/Grok integration (Grok 4 demo pulled Polymarket data); also Stocktwits. - **Kalshi**: $22B round May 2026; FT reported Q3 2026 talks at **$40B**. Polymarket is raising into a competitor valued 2x. ########## LANE media ########## ## Polymarket media spend & marketing — findings **Spend, measured.** Kalshi ~**$32M** national linear TV YTD (iSpot via [Sportico, Jul 2026](https://www.sportico.com/business/sports-betting/2026/prediction-market-data-volume-polymarket-kalshi-world-cup-1234939529/)) — airtime only, excludes talent (Chalamet). Prior marks: $17.7M since Sept ([Marketing Brew, Apr](https://www.marketingbrew.com/stories/2026/04/20/polymarket-and-kalshi-are-turning-tv-programming-into-one-big-casino)), $10.6M in Q1 ([MediaPost](https://www.mediapost.com/publications/article/414156/prediction-markets-do-networks-need-more-transpar.html)). **No public estimate exists for Polymarket's spend** — yet Polymarket ran **45 ads** in live World Cup matches vs Kalshi's 36 (rest of US betting industry: 22). Sportico: the industry spent "hundreds of millions" on World Cup ads + sign-up bonuses. Baseline: Polymarket spent ~$1M on US ads in all of 2025 ([Sportico 2025](https://www.sportico.com/business/sports-betting/2025/polymarket-advertising-spend-united-states-election-1234863286/)). The step-change is ~100x in twelve months, and it is unmeasured publicly — Peter's "collectively hundreds of millions" is consistent with the trade press. **Campaigns.** *Questions Are Everything* — first-ever brand campaign, client-led by creative director Daf Orlovsky with Better Half + Four Four; Coplan takes a story credit. Hero film narrated by Rick Rubin; "Flags" directed by Gabriel Moses to Kanye's "Runaway," 30M views in 24h; Future and Peso Pluma, **no athletes**. 15s cut ran all tournament; the **30s premiered during the UFC White House card**; country editions (ARG/POR/SEN/GBR/FRA) each backed by **local billboards** ([Marketing Chief](https://marketingchief.com/news/polymarkets-world-cup-bet-puts-culture-before-media-spend)). Polymarket branding sat **on the Octagon itself** at UFC Freedom 250, the $60M White House event ([Front Office Sports](https://frontofficesports.com/ufc-freedom-250-white-house-donald-trump-60-million/)). NFL: **$21M-guaranteed survivor contest** with Splash Sports ([Jun 18](https://globalgaminginsider.com/news/5210/polymarket-sponsoring-record-21m-nfl-survivor-contest-through-splash-sports-partnership)), plus Genius Sports/ATP rights and $50-bonus referral codes syndicated through CBS Sports, nj.com et al. **Channel mix beyond CTV.** Golden Globes on-screen live odds (26/28 correct); "million-dollar deals" with **WSJ, Substack, StockTwits, RealClearPolitics**; Dow Jones data distribution; Dear Media's *What Are the Odds?*; own weekly podcast; participatory OOH. Kalshi counters with CNBC (multiyear exclusive), CNN, Fox, Times Square, pitch-side signage, J Balvin, Mourinho. **The exposure.** CMO **Matthew Modabber** personally wired influencers (Nick Shirley $3,100, Riley Gaines) with thin or no disclosure ([Business Insider](https://www.businessinsider.com/how-paid-influencers-hype-polymarkets-odds-prediction-markets-2026-6), [The Verge](https://www.theverge.com/policy/943986/polymarkets-cmo-sent-thousands-to-influencers-like-nick-shirley-and-riley-gaines)); BBB referred Kalshi to state AGs ([Slate](https://slate.com/technology/2026/06/polymarket-kalshi-ads-sponsored-content-t-pain.html)). **Wedge for us:** the KOL lane is their largest unattributed, highest-legal-risk spend. A defensible per-partner allocation record is a compliance asset, not just a media one. **Compliance constraint on creative.** Polymarket US geo-blocks ~8+ states (AZ, IL, MA, MD, MI, MT, NV, OH; WA/LA/SD restricted) — NV under preliminary injunction. Every impression bought nationally in CTV or programmatic is partly waste by law, and waste they can prove only with log-level geo. That is a Precise-native argument. **Room intelligence.** `jenner@` = **Jenner Kearns**, hired ~Apr 2026 to *lead performance marketing*, ex-VP Strategy at Dysrupt ([LinkedIn](https://www.linkedin.com/posts/jennerkearns_last-month-i-officially-joined-polymarket-activity-7460142893501054976-K3z1)) — he is the buyer. `anthony.khaiat@` = **Senior Data Scientist, Research, joined Feb 2026 from Kalshi** (Data Scientist, Sep 2024–Feb 2026), MIT Operations Research MBAn ([LinkedIn](https://www.linkedin.com/in/anthonykhaiat)) — he will check the math, he knows the competitor's measurement stack, and Shapley/attribution rigor will land with him rather than bounce. **Hiring signals** (live on Ashby): Growth Engineer — the *first* dedicated one, owning funnel instrumentation and A/B calls; Partnership Manager, Media — "track and report weekly on referral traffic and placement performance"; Head of Partnership Marketing; Creators & Culture manager; Staff Data Analyst. Translation: measurement is being stood up right now, by hand, by people who don't exist yet. (inferred) No public AOR/Publicis announcement — the embed is unannounced, so treat it as confidential. Context: $21B traded 2025; Kalshi ~83% of CFTC-regulated notional volume in July vs Polymarket US ~10%; Polymarket US left beta only in March. They are outspending into a share gap — which is exactly when allocation truth is worth most. ########## LANE people ########## ## Polymarket side **Anthony Khaiat** — Senior Data Scientist, Polymarket since Feb 2026. Immediately prior: **Data Scientist at Kalshi** (Sep 2024–Feb 2026). MIT MBAn (Operations Research Center); UChicago Econ + Media Arts; Publicis Sapient PM intern 2022. Python/R/Julia/SQL, econometrics, optimization. ([LinkedIn](https://www.linkedin.com/in/anthonykhaiat)) *His seat:* the technical veto. He'll price the method, not the story. Bring the estimand, the counterfactual, holdout design, and exactly how log granularity becomes an allocation. Marginal-contribution/Shapley math is native — don't simplify it. He came from the competitor's stack, so "we've seen how the other side measures" is a real conversation. **Jenner Kearns** (jenner@) — leads **performance marketing**, joined Apr 2026, announced May 13 thanking CMO Matthew Modabber and Jacob Shamberger. **He is the ex-Dysrupt hire Peter referenced: VP of Strategy at Dysrupt Apr 2024–Apr 2026 — Peter's own person.** Before that co-founder of Half Past Nine (acq. 2024), and Common Thread Collective, where he deployed $40MM+ media / $150MM+ revenue in 2020. Chapman BBA. ([profile](https://www.linkedin.com/in/jennerkearns) · [announcement](https://www.linkedin.com/posts/jennerkearns_last-month-i-officially-joined-polymarket-activity-7460142893501054976-K3z1)) *His seat:* he owns the allocation and is judged on install→first-deposit CAC. He is DTC/Meta-native now buying CTV + programmatic at nine figures — that's the gap Precise fills. His own vocabulary is "precise forecasting, creative testing, multi-channel growth" (CTC language). "Price the spend like you price the world" lands hardest here. He needs a defensible answer to *which partner earns the next dollar* — and cover when he gives it. *Escalation path (not in room):* **Matthew Modabber**, CMO since Feb 2024, ex-BeReal first growth hire, claims 25x'ing daily volume via a digital media playbook ([LinkedIn](https://www.linkedin.com/in/matthew-modabber-40a473122)). ## Publicis side **Peter Muzzonigro** — **CEO of Dysrupt**, the performance agency Publicis Media US acquired Jan 2025; he reports to Chris Boothe, CEO Publicis Media US, alongside CRO Jarod Haness and COO Nate Lorenzen. Dysrupt's pitch is the "Impact Advertising System" — media quality, creative performance, cookieless measurement. ([Adweek](https://www.adweek.com/agencies/publicis-media-us-acquires-digital-marketing-agency-dysrupt/) · [Marketing Dive](https://www.marketingdive.com/news/publics-media-us-acquires-performance-digital-marketing-agency-dysrupt/738228/)) *His seat:* a founder-CEO inside a holdco whose alum now runs performance at the client. He needs the embedded team to read as capability, not arms-and-legs — Precise is what makes the IAS measurement claim true at CTV scale. **Stephen Metzger / Daine Tamayo** — **no public footprint found** (Exa + web, LinkedIn and trade press; the Metzgers/Tamayos surfaced are all wrong people). (inferred) Dysrupt/Publicis activation-side — programmatic/CTV activation lead plus planner-analyst on the embedded team. Ask Spencer or Peter to confirm titles pre-meeting. *Their seat:* they carry the work. Tell them Precise reads logs they already pull, needs no re-tagging, and hands them the defensible answer when Jenner asks why StackAdapt got that budget. ########## LANE assets ########## ## 1. What each surface pitches **`/docs/opportunities/polymarket-brief-site/`** (5 live pages, not 3; `onepager/read/workfield` are redirect shims). Deployed copy: `~/projects/polymarket-brief` → polymarket-brief.vercel.app. - `index.html` — external forwardable brief. Spine: "Run every media dollar on economics." Connect (AppsFlyer + DSPs + cost feeds) → Price the next dollar (marginal CPA, saturation, pacing, modeled headroom) → Move (cut→fund pair) → Operate (propose/approve/API write, audit trail). Closes on a **sample request** (5 items: AppsFlyer install/in-app rows, one DSP delivery+cost at movable grain, field dictionary + join keys, first-deposit definition/window/maturation, current decision report) → Precise returns a coverage-and-joinability read. - `demo.html` + `demo.js` — "Growth opportunity book." All numbers hardcoded in `demo.js`: World Cup flight, day 12/28, $4.82M spend, 18,940 first deposits, $254 blended CPA, 91% of spend joined, freshness 4–18m; 3 ranked moves ($180k / $120k / $90k) with source→dest marginal CPA pairs and "equivalent cost reduction" ($38k–$52k top row). Row 03 sits in "Capacity review" as the credibility beat. Labeled "Demo data · shadow mode." - `deck.html` (10 slides, presenter mode) + `adamspence.html` (operator index mapping every Peter ask from the Jul 9 calls to a slide) + `meeting.html` (room sheet: Peter's position, sponsor = head of analytics, language guardrails). **`/docs/opportunities/publicis-polymarket-opportunity-2026-07-09.html`** — earlier internal room brief, same economics spine but with the **three-party framing**: Polymarket (allocation), Publicis (repeatable agency-owned decision layer), Precise (economics engine). Adds "later, the model compounds" (cross-campaign response memory, priors that don't override new flights) and 7 discovery questions. Same illustrative $180k / $38k–$52k move. **`scratchpad/pm-liked-deck.html`** (the one PM liked) — different spine entirely: **"Price the spend the way you price the world."** Analogy to their own mechanism (prices earn accuracy by correction → fees should too). Decompose → Score → Classify (Contributing / Partial / Redundant / Waste) via TraceScore; independent referee, not a player; Trace Receipt sealed to AliceNet. Stat block: $20B–$33B recovery (ANA 2023–25), 36¢ working media (ANA 2024), 51% targeting-data accuracy (Truthset), **15–40% overhead→working media "Precise pilot data."** Credential: "grew Madhive's decisioning stack to $100M+." Ask: one scoped slice of paid growth spend, single integration point. ## 2. What the observatory actually is `/Users/adamhelfgott/projects/precise/packages/polymarket-observatory/` — a **real, large, running public-data research lab**, ~54 GB of sealed artifacts, US + Global lanes, harness-backed (`harness/examples/polymarket-observatory.harness.ts`, 12 ablatable seams + 4 Global nodes), lazy-mounted from root harness via `npm run precise -- init poly`. It **cannot trade**: no wallet, signer, secret loader, or order path; every ledger row is `authority: shadow/paper`, `order_authority: none`; live size zero. Real today (per `PRODUCT.md` + `/docs/labs/polymarket-observatory.md`): finalized Polygon collector; complete Gamma registry (1,713,823 markets / 3,427,513 tokens); pinned TimeSeventeen V1 Parquet archive (1.2B rows); 7.5M-row Goldsky sample; US all-market panel (7,819 → 7,471 markets, 7,901 $25 recipe watches, 6,245 complete exits); Markov/Merkle information court; Claude-planned paper agent. Results are **negative or fragile, by design**: favorite bucket −8.9%, longshot −13.4% on complete-exit capital; first 10-position paper portfolio **lost $31.72** after fees; six-window wallet study = 118 cases, CIs cross zero, labeled fragile, no winner; Markov court finds public-flow information but **no held-out wallet increment**. `no recipe has passed historical confirmation`. Freshest artifact `data/us-panel-markouts.json` is dated **2026-07-13** — the loop is not currently running. Planned/blocked: BigQuery actor backfill (ADC `invalid_rapt`, ~6.03 TiB / $37.70 dry run), full Polygon actor history (17.1 TiB estimate), PMXT v2 GCS materialization, GCP lake (disabled by default). ## 3. Claims needing care - **"We already run a Polymarket lab" is true and defensible — but it is a research lab that has found no edge and is three weeks stale.** Safe framing: rigor, refusal machinery, sealed evidence. Unsafe: "we trade / we found signal / it's live." - The lab is **prediction-market research**, not media measurement. There is zero shared pipeline with the media-spend pitch. Don't let "we run a Polymarket lab" imply we already touch Polymarket's ad data. - **Every demo number is invented** (`demo.js`). `$38k–$52k`, `$342 → $244–269`, `91% joined`, `4–18m freshness`, `$4.82M`. Publicis page says "illustrative"; brief `index.html` says only "Demo" in one badge. State it aloud once, per `meeting.html`. - **`15–40% of overhead converted to working media — "Precise pilot data"`** in the liked deck is the highest-risk claim in the corpus; it needs a named pilot and denominator or should carry a range-of-observed caveat. ANA/Truthset stats are third-party and fine. - AliceNet "tamper-evident record sealed to blockchain" — `meeting.html` explicitly holsters blockchain/mathematical-proof language for this room; the liked deck leads with it. Pick one register per audience. - "$100M+ Madhive decisioning stack" and Bill Wise's $250k figure — the room sheet says confirm permission with Spencer before voicing. - Peter's raw-log commitment ("you can do anything you need to do") is **secondhand and unverified**; the walkthrough is what converts it. - Shapley: per `/docs/labs/POLYMARKET-SHAPLEY-AUDIT-2026-07-13.md`, the exact reducer is verified correct (agrees with canonical Python to 8.67e-19) but games are 5–6 feature buckets over 4,502 out-of-fold evaluations, **not individual-wallet attribution**, and the target is not executable P&L. Don't say "we run Shapley on all Polymarket history." ## 4. Strongest reusable pieces for the dossier 1. **The sample request** (`index.html` bottom, copyable) — the single best conversion object in the corpus; audience-agnostic, low-ask, produces a coverage read that prices the deal. 2. **The cut→fund row as the unit of work** — one row = complete decision (source, destination, amount, expected result, confidence, reason). Best single visual asset; already three renderings (hero card, table, deck slide). 3. **The "Capacity review" state** (`move-03`) — the refusal beat made visible; maps directly to the repo's "refusals are the credibility feature" principle. 4. **Their-mechanism analogy from the liked deck** — "every market forces a price to prove itself against a real outcome; we do that for spend." This is the one line PM responded to; it should anchor the dossier. 5. **Four-verb loop** (Connect / Price / Move / Operate) — stable across both decks, survives audience swap. 6. **Observatory as proof-of-discipline** — the negative results (−$31.72 portfolio, CIs crossing zero, winner: null, closed promotion gates) are the strongest available evidence that Precise refuses to overclaim. Use as character, not capability. 7. **Publicis three-party frame** — Polymarket / Publicis / Precise; the only place the agency-repeatability upside is articulated. 8. **`adamspence.html` ask→slide map** — verbatim Peter quotes with timestamps; the raw material for a "built from your own words" dossier section. Paths: `/Users/adamhelfgott/projects/precise/docs/opportunities/polymarket-brief-site/{index,demo,deck,meeting,adamspence}.html`, `.../demo.js`, `/Users/adamhelfgott/projects/precise/docs/opportunities/publicis-polymarket-opportunity-2026-07-09.html`, `/Users/adamhelfgott/projects/precise/docs/labs/polymarket-observatory.md`, `/Users/adamhelfgott/projects/precise/packages/polymarket-observatory/{README,PRODUCT}.md`, `.../data/us-panel-markouts.json`. ########## SYNTHESIS ########## # POLYMARKET MEDIA-SPEND PITCH BRIEF ## 1. The wedge Polymarket went from about $1M of US ad spend in 2025 to nine figures across World Cup, the White House UFC card, and now NFL, with every major DSP on the plan and AppsFlyer sitting on an install to first-deposit funnel. Peter's stated pain on Jul 9 is not reporting. It is allocation: which partners and channels earn the next dollar, at a grain fine enough to act on, with raw logs available. That is exactly the Precise loop. Connect (AppsFlyer plus DSP delivery and cost feeds), price the next dollar (marginal CPA, saturation, headroom), move (a cut and fund pair), operate (propose, approve, write back, keep the decision record). We are the independent referee that makes each dollar prove itself against a real outcome, which is the same discipline their own order book runs on. ## 2. Five facts that sharpen it 1. **They are outspending and losing share.** Polymarket ran 45 ads inside live World Cup matches vs Kalshi's 36, yet July volume was Kalshi $37.7B (74.5%) vs Polymarket US $5B ([Sportico](https://www.sportico.com/business/sports-betting/2026/prediction-market-data-volume-polymarket-kalshi-world-cup-1234939529/); [Block Stream](https://blockstreammedia.com/2026/08/03/prediction-markets-hit-record-50-6b-july-volume/)). Allocation truth is worth most when you are buying into a share gap. 2. **The spend step-change is roughly 100x in twelve months**, from ~$1M US in 2025 ([Sportico 2025](https://www.sportico.com/business/sports-betting/2025/polymarket-advertising-spend-united-states-election-1234863286/)) to Peter's "collectively hundreds of millions." No public estimate of their spend exists. Nobody outside the room can grade it, including them. 3. **Geo-blocked states make national buying partly waste by law.** US Polymarket is blocked or restricted in roughly eight to eleven states. Log-level geo turns that from a legal footnote into recoverable budget. 4. **The KOL lane is the largest unattributed spend and carries legal exposure.** Their CMO wired $350K-plus to creators personally, with thin disclosure ([The Verge](https://www.theverge.com/policy/943986/polymarkets-cmo-sent-thousands-to-influencers-like-nick-shirley-and-riley-gaines)). A per-partner allocation record is a compliance asset here, not only a media one. 5. **Measurement is being stood up by hand, right now.** Live roles: first-ever Growth Engineer owning funnel instrumentation, Partnership Manager Media required to "track and report weekly on referral traffic and placement performance." We arrive as the thing they are hiring for, not a replacement for it. ## 3. What each seat needs to hear - **Jenner Kearns (performance marketing, and Peter's own Dysrupt alum):** he owns install to first-deposit CAC and is DTC and Meta native, now buying CTV at nine figures. Give him the defensible answer to which partner earns the next dollar, plus cover when he gives it. Lead line: price the spend like you price the world. - **Anthony Khaiat (Senior Data Scientist, ex-Kalshi, MIT MBAn):** the technical veto. Do not simplify. Bring the estimand, the counterfactual, holdout design, and the exact path from log grain to allocation. Marginal contribution and Shapley are native vocabulary for him. He knows how the competitor measures, which makes that a real conversation, not a jab. - **Peter Muzzonigro (CEO, Dysrupt, inside Publicis Media US):** he needs the embed to read as capability, not arms and legs. Precise is what makes the Impact Advertising System measurement claim true at CTV scale, and it is repeatable across his other clients. - **Stephen Metzger and Daine Tamayo (activation side, titles unconfirmed):** they carry the work. Tell them we read logs they already pull, require no re-tagging, and hand them the answer when Jenner asks why StackAdapt got that budget. ## 4. The first proof A retrospective read on one closed flight, ideally World Cup, before NFL money commits. Scope: AppsFlyer install and in-app event rows, one DSP's delivery and cost at movable grain, field dictionary and join keys, first-deposit definition with window and maturation, and their current decision report. We return a coverage and joinability read first, then a ranked set of cut and fund rows with expected result, confidence, and reason, with a randomly held-out slice of partners scored blind and revealed after. Shadow mode throughout, no spend authority, no re-tagging, one integration point. Easy to say yes to because the data already exists, nothing in market changes, and the holdout means the result is checkable by Anthony rather than asserted by us. ## 5. Observatory and the liked deck The **liked deck** (`scratchpad/pm-liked-deck.html`) anchors the dossier, because "price the spend the way you price the world" is the line Peter responded to. Keep its mechanism analogy, keep the ANA and Truthset third-party stats, and either name a pilot behind the 15 to 40 percent overhead-to-working-media figure or carry it as range-of-observed. The **observatory** (`packages/polymarket-observatory/`) is character, not capability. It is a real harness-backed research lab that has published its own negative results: a first paper portfolio down $31.72 after fees, confidence intervals crossing zero, winner null, no recipe passing historical confirmation. Say plainly that it is prediction-market research, not media measurement, that it does not touch their ad data, and that we publish the nulls. That is the proof of refusal discipline. Do not imply live trading, found edge, or a running loop; the freshest artifact is dated 2026-07-13. ## 6. Three pins and the knock-downs - **"We already have MMM and AppsFlyer."** Good, keep both. Those tell you what happened. We price the next dollar at the partner and placement grain the DSP logs already carry, and we do it against a holdout so the answer is testable. - **"You are another vendor between us and our DSPs."** We do not buy, do not touch tags, and cannot move money without approval. We are the referee, not a player, and the record we produce belongs to them. - **"Your demo numbers are made up."** Correct, and we say it aloud once. Every figure in the demo is illustrative shape, not measurement. The sample request exists precisely so the first real numbers are theirs. ## 7. Vocabulary Their side: price discovery, correction, resolution, settlement, edge, notional, open interest, the book. Performance side: marginal CPA, incrementality, holdout, saturation curve, headroom, pacing, blended vs marginal, install to first deposit, log-level, movable grain, cut and fund. Bridge line: every market forces a price to prove itself against a real outcome, and we do that for spend. **Paths:** `/Users/adamhelfgott/projects/precise/docs/opportunities/polymarket-brief-site/{index,demo,deck,meeting,adamspence}.html`, `/Users/adamhelfgott/projects/precise/docs/opportunities/publicis-polymarket-opportunity-2026-07-09.html`, `/Users/adamhelfgott/projects/precise/packages/polymarket-observatory/`