Zeta Global Holdings Corp. (NASDAQ: ZETA) — An AI-Data “Moat” That Is Really a Dilution-Funded Roll-Up in a Platform Costume

Independent Equity Research — Analytical Note Report date: 2026-07-10 · Price: $21.92 (2026-07-09 close) · Market cap: ~$5.45B · Enterprise value: ~$5.33B Sector: Information Technology · Data-Driven Marketing / AdTech Software · FY-end: December · CIK: 0001851003 Coverage: Independent initiation

⚡ Claude’s Take

This is the author’s own independent opinion and general information only — not investment advice, and not a recommendation to buy or sell any security. The analytical body that follows takes no position and sets no price target; it discusses valuation only as embedded expectations and scenarios.

Verdict: AVOID at ~$22 — not a short. A well-run, fast-growing martech roll-up whose “profit,” “free cash flow,” and “moat” are all accounting-and-narrative constructions, priced as a moaty AI-data compounder it has not proven itself to be. Constructive only in the low-to-mid teens (~2.0–2.5x sales), where the roll-up would be honestly priced.

Zeta is the rare stock where both GAAP bears and adjusted-EBITDA bulls are wrong, in opposite directions. The GAAP net loss (−$31.5M FY25) overstates the weakness — it is dragged below the line by a non-cash mark on acquisition earn-out liabilities. But the company’s headline “$279M of Adjusted EBITDA (21% margin)” and “$165–179M of free cash flow” vastly overstate the strength: 64% of that Adjusted EBITDA is added-back stock compensation, and that same $178M of SBC is larger than all of the free cash flow — so free cash flow net of the true cost of paying employees in stock is roughly zero to slightly negative. The real economics are ~$100M of EBITDA-ex-SBC on $1.3B of revenue (7.7% margin), a ~0.8% return on invested capital that sits below any reasonable cost of capital, a gross margin that has compressed as the company scaled (63.5% → 60.6%), and share count up ~52% in two years. This is not operating leverage; it is a good sales organization and a serial-acquisition engine transferring value from shareholders to employees via ~19%/yr dilution. The “proprietary opted-in first-party data moat” is, on the company’s own 10-K language, a mix of “proprietary, partner, and publicly available” data harvested from web widgets (Disqus, LiveIntent) — commodity-adjacent and regulation-exposed, not a defensible intangible. Culper Research’s 2024 short thesis over-reached on fraud, but its softer, durable point — that the data-moat story is overstated and part of the revenue is media/data pass-through dressed as SaaS — is corroborated by the filings.

Why not a short: the growth is real (~21% organic), the balance sheet is roughly net-cash, sell-side is supportive (BofA PT $28), a June-2026 Palantir partnership re-ignited the AI narrative, and — critically — this is a ~2.0-beta name that already squeezed the last short-seller (the stock round-tripped to new highs after Culper). Betting against a founder-controlled momentum story with a live AI catalyst and a −70% max-drawdown volatility profile is a good way to be right on the analysis and wrong on the P&L. Framing: a high-beta, V-recovered story stock at the rich end of the martech cohort, whose reverse-DCF demands a 0%→15%+ true-margin expansion it has not begun to demonstrate. I would only underwrite it in the low-to-mid teens, where it prices like the LiveRamp/Braze-quality roll-up it actually is.

• Conviction: Medium. Framing tag: “A great sales org and a serial acquirer, marketed as an AI-data monopoly.”
• Bullish trigger (flips me constructive): a clean, disclosed organic-growth number sustained ≥18–20% with SBC/revenue falling below ~10% and true (SBC-inclusive) FCF margin visibly climbing into the teens — i.e., the margin transformation actually arriving, plus the buyback exceeding SBC so net share count falls.
• Bearish trigger (flips me to short-candidate): organic growth decelerating into the low teens as the roll-up laps out, SBC re-accelerating (unrecognized SBC already jumped $254M→$428M in one quarter), or independent corroboration of Culper’s data/revenue-quality allegations.

📈 Stock Price Action — Five-Year Event Map

Factual price history from the AZI five-year adjusted-price series. Price moves are FACT; attributed causes are INTERPRETATION. No recommendation, no target.

Zeta is a violent, high-beta round-trip. From its June-2021 IPO at $8.89, it slid to a $4.27 low (Jul-2022), ground back through 2023, then ripped roughly +362% in 2024 to an all-time-high close of $36.74 (2024-11-11) — before a short-seller report vaporized more than half its value in three sessions. It has since twice round-tripped through the mid-$20s and the mid-teens. Today at $21.92 it sits ~40% below its ATH, inside a 52-week range of roughly $13.74–$25.95, and above its rising 21/50/200-day EMAs (~$20.29 / $19.60 / $18.57) — a stock that has recovered ~51% off its March-2026 low on an AI-data catalyst but remains well shy of its 2024 peak.

Cycle narrative. (1) IPO & drift — Zeta debuted at $8.89 in June 2021 and leaked lower as the profitless-software trade cooled. (2) 2022 bear — the rate shock roughly halved it to $4.27; a cash-burning, SBC-heavy martech name was precisely what 2022 punished. (3) 2023 stabilization — narrowing losses and platform traction lifted it back toward $9. (4) The 2024 melt-up — a near-quadruple to $36.74 as consecutive beats-and-raises and an “AI-powered marketing cloud” story re-rated the multiple from ~2x to ~7x sales. (5) The Culper crash — on 2024-11-13 short-seller Culper Research alleged Zeta operated “consent farms” and round-tripped data revenue; the stock lost ~53% in three sessions on 4–5× volume [move=FACT; the validity of the claims is an OPEN QUESTION — ]. (6) The Feb-2025 rebound — a solid Q4’24 print, buyback activity, and partial dismissal of Culper carried it back to $24.69. (7) The double round-trip — two high-beta risk-off episodes undid the recovery, bottoming at $14.55. (8) The Palantir V-recovery — a June-2026 Palantir partnership re-ignited the AI-data-collaboration narrative and the stock climbed ~51% off the March low to $21.92.

1. Executive Summary

Zeta Global is a New-York-based, founder-controlled, data-driven marketing-technology company that sells the Zeta Marketing Platform (ZMP) — a bundle of customer-data platform (CDP), consumer identity graph, omnichannel campaign activation, and, since 2025, an agentic-AI interface (“Athena by Zeta”). It went public in June 2021 and has grown revenue from $458M (FY21) to $1,304.7M (FY25, +29.7%), with Q1-26 revenue of $396.3M (+49.9% YoY). On the surface this is a scaled, accelerating, and — as of FY2025 — finally-profitable AI-marketing platform.

Beneath the surface, three things are true and they define the thesis. First, the growth is materially inorganic. Stripping the LiveIntent (Oct-2024, $277M) and Marigold-Enterprise (Nov-2025, $302.8M) acquisitions, pro-forma FY25 revenue growth is ~21.4%, and the +49.9% Q1-26 headline is roughly half a full-quarter of Marigold that was absent a year earlier. Zeta is a serial roll-up (Disqus, Apptness, ArcaMax, Kinetic, LiveIntent, Marigold) buying largely mature revenue with its own stock. Second, the reported profitability is an add-back construction. FY25 Adjusted EBITDA of $278.7M (21.4% margin) is 64% stock-based compensation; excluding SBC it is $100.8M (7.7% margin). Reported free cash flow of ~$165M is smaller than the $177.8M of SBC — so free cash flow net of the true economic cost of equity compensation is −$12.8M. As-reported GAAP operating income reached a mere +$5.4M in FY25 (its first positive year) and already reverted to a −$18.8M operating loss in Q1-26. Third, there is no moat visible in the financials. Gross margin has compressed as the company scaled (63.5% FY22 → 60.6% FY25) — an anti-moat signature — and return on invested capital is ~0.8%, below any plausible cost of capital. The “proprietary opted-in first-party data” narrative that anchors the bull case is, by the 10-K’s own description, a “mix of proprietary, partner, and publicly available data” harvested from web widgets — commodity-adjacent and squarely in the crosshairs of tightening privacy law.

Capital allocation and governance reinforce the skeptical read. Zeta is founder-controlled through a dual-class structure (Class B carries 10 votes; Steinberg-affiliated entities hold ~40% of total voting power). Management earns 200%-of-target cash bonuses on an Adjusted-EBITDA metric that is majority-composed of the very stock compensation they are paid, and PSUs vest on share-price hurdles rather than returns on capital. The $120M FY25 buyback offset well under half the year’s dilution (a “repurchase and RSA-withholding” program), so net share count still rose. And across a corpus of 188 Form 4s and 101 Form 144s, there is not a single open-market purchase by any insider — while the CEO and CFO have gifted ~6.9M shares combined into family/estate trusts that filed ~$46.9M of Rule 144 sales in 2026 alone.

At ~$5.3B enterprise value the stock trades at ~4x trailing / ~3.4x forward sales and ~20x Adjusted EBITDA (or ~53x EBITDA on a GAAP basis) — the rich end of the martech cohort, above LiveRamp, Braze and The Trade Desk on EV/sales despite a lower gross margin and no proven moat. A reverse-DCF shows the price already embeds a transformation of true (SBC-inclusive) free-cash-flow margin from ~0% today to ~12–20% — a leap of faith the company has not begun to demonstrate. The base case is roughly fair value; the risk is asymmetric to a re-rating toward the LiveRamp/Braze zone if organic growth decelerates or SBC re-accelerates. This report takes no position and sets no target; the sections below argue the evidence.

2. Business Overview

What Zeta does. Zeta Global operates the Zeta Marketing Platform (ZMP), an enterprise “marketing cloud” that combines (a) a customer-data platform that ingests and unifies first- and third-party data into a “single view” of a consumer; (b) the “Zeta SuperGraph,” an identity graph the company markets as covering 245M+ US and 535M+ global individuals with 2,500+ attributes each; © omnichannel activation across email, display, social, connected TV, and mobile; (d) an insights/measurement layer (“Zeta Answers”); and (e) since 2025 an agentic-AI interface, “Athena by Zeta.” The commercial pitch is vendor consolidation — “One Zeta” — replacing a stack of point tools (an email service provider, a CDP, a demand-side platform, an identity vendor) with a single platform. [Source: FY2025 10-K, filed 2026-02-25.]

How it makes money — not clean recurring SaaS. Revenue arises from “subscription fees, volume-based utilization fees and fees for professional services.” A material portion is therefore consumption-linked rather than fixed subscription, which matters for both quality and forward visibility. Zeta discloses revenue by delivery mechanism: Direct platform 74% / Integrated platform 26% in FY2025 (70/30 in FY2024; 72/28 in FY2023; 75/25 in Q1-26). “Integrated platform” revenue is “generated by leveraging its platform’s integration with third parties” — media buys (notably social) and API partners such as LiveIntent, where third-party media/data cost is a pass-through that sits in cost of revenue. On the Q1-26 call, the CFO tied a 190-bps YoY rise in cost of revenue (to 41%) to “new agency wins driving a higher initial mix of social as a channel.” Translation: a non-trivial slice of “platform revenue” is media arbitrage flowing through the platform at pass-through economics — the structural reason Zeta’s gross margin (~60%) sits below pure-SaaS peers. [Source: FY2025 10-K MD&A; Q1-26 earnings call, 2026-04-30.] (Interpretation.)

Customers and KPIs. FY2025: 2,651 total customers, 602 “scaled,” and 184 “super-scaled” (≥$1.0M trailing-12-month revenue, +24% YoY); 87% of revenue comes from super-scaled customers. Scaled ARPU was $2.11M (+13%); super-scaled ARPU $6.16M. Q1-26: 189 super-scaled (+19% YoY), super-scaled ARPU +21% YoY, net revenue retention above the 110–115% target. These are genuinely strong land-expand-extend metrics. Two cautions: (i) management has signaled it will stop reporting scaled-customer metrics and report only the flattering super-scaled cohort going forward — a narrowing of disclosure a skeptic should note; and (ii) FY2025 saw one customer exceed 10% of total revenue (none did in FY24) and one customer exceed 10% of receivables — an emerging concentration that did not exist a year ago. [Source: FY2025 10-K; Q1-26 10-Q, Note 10.] (Fact; Interpretation on the disclosure change.)

A serial acquirer. Growth has been powered by a steady acquisition cadence: Disqus (2017, the comment-widget property), Apptness and ArcaMax (2022), Kinetic, LiveIntent (email/inbox identity, closed 10/21/2024, $277.0M), and Marigold’s Enterprise Business (the carved-out Selligent/Cheetah Digital/Campaign Monitor/Sailthru enterprise-email assets, closed 11/24/2025, $302.8M). Both large deals were funded substantially in equity plus seller notes and are heavy with goodwill. The consequence for the top line is that reported growth overstates the organic rate. [Source: FY2025 10-K, Note 7.]

Cyclicality. Zeta’s political/advocacy customers make revenue modestly cyclical — higher in presidential election years (2024) and, to a lesser degree, midterms (2026). Management now reports growth “ex-political candidate revenue” to normalize comparisons. [Source: FY2025 10-K, “Seasonality.”]

** read:** A genuine, scaled, accelerating martech platform with a strong cross-sell motion — but with mixed revenue quality (26% integrated/media pass-through, heavy inorganic growth) and an “opted-in first-party data” framing that, on the company’s own disclosure, is overstated. The structure that follows tests whether any of this is durable and whether it reaches the owner.

3. Industry Dynamics

A large, elastic, and definitionally-fragmented pool. “Martech” is one of the most crowded categories in software: chiefmartec’s landscape counts roughly 14,000+ tools, up ~100x since 2011 — a proxy for how unconsolidated and low-barrier the space is. Total-addressable-market figures are almost useless because they range from ~$175B (narrow) to ~$580–860B (broad) depending on where one draws the line. The relevant sub-segment — CDP / data-driven marketing / identity — is perhaps $25–40B growing in the mid-20s-to-30s percent range, genuinely fast. So the demand-side tailwind (enterprises spending more on data-driven marketing and consolidating onto fewer platforms) is real. [Source: industry estimates, chiefmartec 2025 landscape; interpretation.]

The walled-garden problem — the profit pool sits upstream. The highest-signal first-party data and the terminal ad inventory sit inside the walled gardens — Google, Meta, Amazon, TikTok — which every independent (Zeta, The Trade Desk, LiveRamp, Salesforce, Adobe, Braze, Klaviyo) must buy from and integrate with, not disintermediate. The independents compete for the orchestration / identity / activation layer around the gardens. That is a real market, but it is one where the most valuable data and the durable pricing power belong to someone else. Amazon’s growing commerce-media and data assets tighten this further. [Interpretation.]

Signal loss and privacy — a double-edged sword, not a clean tailwind. The bull framing is that cookie deprecation and signal loss raise the value of any deterministic identity graph (Zeta’s SuperGraph). Partly true. But the same forces (a) concentrate power further in the logged-in walled gardens, and (b) escalate the regulatory risk aimed squarely at Zeta’s collection model — CCPA/CPRA, the California Delete Act (which lets consumers bulk-delete from data brokers via a single request), GDPR, state privacy laws proliferating, and rising FTC scrutiny of data brokers and “dark-pattern” consent. A business whose data is harvested via comment widgets and lead-generation properties is more exposed to this than one built on merchants’ own opt-in CRM lists. Zeta’s own 10-K concedes stricter EU opt-in enforcement “could adversely affect our margins,” and that adverse publicity about its data collection “has in the past adversely impacted the price of our Class A Common Stock.” [Source: FY2025 10-K risk factors.]

Competitive intensity. Fragmented and feature-competitive, with low switching costs at the point-solution level and every incumbent (Salesforce, Adobe, Oracle) bundling overlapping capability into broader suites. Generative AI has lowered, not raised, the barrier to shipping a “marketing copilot.” [Interpretation.]

** Verdict — structurally mixed to below-average.** There is real secular growth in digital-marketing spend and a genuine “consolidate onto fewer platforms” tailwind. But the industry is extremely fragmented, the most valuable data and pricing power sit upstream in the walled gardens, switching costs at the point level are weak, AI is commoditizing the copilot layer, and a privacy/regulatory overhang points directly at Zeta’s sourcing model. This is not a structurally attractive oligopoly like payments, exchanges, or ratings; it is a crowded, contested layer where a well-run operator can grow but where the industry does not confer durable excess returns.

4. Competitive Position

Name the mechanism — and its absence. Management asserts two Greenwald-taxonomy advantages: an intangible (proprietary data + 125+ AI patents) and a scale/data-flywheel effect (“data that improves with every interaction, intelligence that compounds with every decision”). Pressure-tested against the frameworks and, more importantly, the financials, neither qualifies as a durable moat.

The data-flywheel claim fails its own key tell. A genuine economies-of-scale-in-data advantage should surface as rising gross margin as a largely-fixed data asset is amortized over more revenue. Zeta’s gross margin has gone the other way: 63.5% (FY22) → 62.3% (FY23) → 60.3% (FY24) → 60.6% (FY25). Roughly 120% cumulative revenue growth over three years produced no gross-margin expansion — because incremental revenue increasingly carries integrated/media pass-through and freshly-acquired legacy-email cost. Margin compressing as the business scales is the opposite of a scale moat. A moat is supposed to show up in returns on capital; Zeta’s GAAP operating margin only just crossed into positive territory (+0.4% as reported in FY25), ROA is still negative (−2.4%), and ROIC is ~0.8% — below cost of capital. [Source: ROIC.ai profitability ratios; FY2025 10-K.]

The data “intangible” is commodity-adjacent and regulation-exposed. By the 10-K’s own account, the “principal way that we collect individual data is directly from consumers when they register or interact with our platform (such as the DISQUS commenting system and LiveIntent inbox advertising), or with partners’ services,” and the dataset is “a comprehensive mix of proprietary, partner, and publicly available data.” In plain terms, the “first-party opt-in” asset is largely harvested from Zeta-owned web widgets and lead-gen funnels and supplemented by licensed third-party and public data — not obtained through direct brand-CRM relationships the way a Klaviyo (a merchant’s own customers) or a bank’s in-house CDP would be. Anyone with capital can license third-party data and buy consent inventory; the barrier is not a patent or a network effect but a data-ops operation exposed to the same privacy laws as every data broker. If the SuperGraph were truly unique and irreplaceable, it would command pricing power visible in the margin — it does not. [Source: FY2025 10-K.]

“AI” is table-stakes. Every peer ships gen-AI copilots (Salesforce Einstein/Agentforce, Adobe Sensei/AEP, Braze Sage, Klaviyo AI). Zeta’s Athena is, by management’s own account, an effective sales tool (“Athena demos were a crucial differentiator in each of our marquee wins,” Q1-26 call) — but a sales accelerant is not a structural barrier to entry.

Competitor comparison (with numbers).

The tell is the margin column: Zeta’s ~60% gross margin sits below every software peer because a chunk of its “platform” revenue is media/data pass-through, not software. The Trade Desk — the genuine independent-adtech moat — earns ~80% gross margin and ~24%+ EBITDA margins on a broadly comparable independent model. Zeta is not that. Net revenue retention above 110% and 21% super-scaled ARPU growth are respectable, but 110-115% NRR is good, not best-in-class (elite SaaS runs 120%+), and it is boosted by cross-selling acquired products. [Sources: company filings; TTD FY25 8-K; Klaviyo/LiveRamp/Braze IR.]

Where the advantage is real. The go-to-market motion — “land, expand, extend / One Zeta,” a hunter/farmer sales model, industry-specialized sellers, and the Athena demo — is genuinely effective and drives the ARPU expansion. That is worth something. But it is a good sales organization plus an aggressive M&A engine, which is replicable and confers no pricing power — not a Greenwald moat.

Marathon capital-cycle lens. Zeta is a stock-and-debt-funded roll-up compounding external capital into acquired revenue in a fragmented, capital-attracting space with commoditizing AI — precisely the setup Marathon warns is prone to mean-reversion and value destruction when returns on deployed capital do not clear the cost of capital. Zeta’s do not.

** Verdict — crowded market, weak differentiation, no durable moat proven in the financials.** The data “moat” is, on the evidence, a commodity-adjacent, regulation-exposed data pool + an excellent sales org + serial M&A, not a defensible intangible or network effect. This is a good business being marketed as a great one.

5. Growth History and Forward Opportunities

History. Revenue: $368M (FY20) → $458M (FY21, +24.5%) → $591M (FY22, +28.9%) → $729M (FY23, +23.3%) → $1,005.8M (FY24, +38.0%) → $1,304.7M (FY25, +29.7%), with Q1-26 at $396.3M (+49.9% YoY). At face value, a rare acceleration at scale. [Source: ROIC.ai; FY2025 10-K.]

But the headline overstates the organic rate. Zeta’s own pro-forma disclosure — revenue as if both LiveIntent and Marigold had closed on 1/1/2024 — is FY25 $1,515.8M vs. FY24 $1,248.6M, i.e. ~21.4% apples-to-apples growth, not 29.7% (10-K Note 7). The Q1-26 +49.9% is even more inorganic: Marigold adds roughly $211M annualized (~16% of the base) and contributed a full quarter in Q1-26 versus zero in Q1-25, on top of a partial LiveIntent lap. Management’s own “ex-Marigold, ex-political” framing put organic growth nearer +29% in Q1-26 — still healthy, but the point stands that a large and rising share of reported growth is acquired. Call true organic growth low-20s%, decelerating as the deals lap. [Source: FY2025 10-K Note 7; Q1-26 call.] (Fact + interpretation.)

Quality of the organic growth. The organic engine is real: super-scaled customer count +19–24% YoY, super-scaled ARPU +21% YoY, NRR above 110–115%, driven by cross-selling more ZMP modules (data, channels, AI) into existing enterprise accounts. This is legitimate land-and-expand. The caveat is that a portion of “expansion” is Zeta selling acquired products (LiveIntent inbox media, Marigold email) into its base — cross-sell of bought capability, which is good execution but blends organic and inorganic.

Forward opportunities. (1) Cross-sell / consolidation — the strongest lever; enterprises collapsing point tools onto “One Zeta.” (2) AI / Athena monetization — agentic-marketing features that could lift ARPU if customers pay for them (unproven at scale). (3) The Palantir partnership (announced 2026-06-23) — a data-and-AI-infrastructure collaboration management frames as a potential ~$100M annual-revenue opportunity; note this is commercial PR, not a material 8-K transaction in the filing corpus, so treat the number as aspirational until it appears in results. (4) The Snowflake / “Open Standards” (OSI) marketing-data initiative — positioning Zeta as a data-interoperability layer for enterprise AI. (5) Continued M&A — the roll-up will likely continue, which keeps headline growth up but raises the goodwill and dilution questions in .

** Verdict — high reported growth of mixed quality.** The underlying ~20%+ organic growth is genuine and above-average, but the headline is inflated by acquisitions, and the forward levers that would justify a premium multiple (AI monetization, Palantir/Snowflake) are narrative-stage, not yet in the numbers. Good growth; not yet high-quality growth in the sense of durable, organic, margin-accretive expansion.

6. Financial Quality

The margin structure does not improve with scale. Gross margin (revenue less cost of revenue ex-D&A) is flat-to-down — 62.3% (2023) → 60.3% (2024) → 60.6% (2025) — so ~80% cumulative revenue growth produced no gross-margin expansion, inconsistent with a scaling software platform and consistent with a mix that includes low-margin integrated/media revenue and freshly-acquired legacy assets. The GAAP operating “inflection” (2023 −$167.7M → 2024 −$67.9M → 2025 +$5.4M as reported) came almost entirely from (a) SBC rolling off ($242.9M → $195.0M → $177.8M) and (b) operating leverage on sales & marketing (39.6% → 26.1% of revenue) and G&A (28.2% → 17.9%) — not from gross margin. And it did not stick: Q1-26 reverted to a −$18.8M operating loss and a −$13.2M net loss despite +49.9% revenue, as SBC re-accelerated to $53.0M in the quarter, Marigold amortization pushed D&A to $23.5M, and restructuring hit $6.8M. [Source: FY2025 10-K and Q1-26 10-Q statements of operations.]

Note on the “$28.8M operating income” figure. Some aggregators (and the initial data pull) show FY25 operating income of $28.8M. That is operating income before $20.3M of acquisition-related and $3.2M of restructuring expense. The as-reported GAAP income from operations in the 10-K is $5.4M — the number this memo uses. [Filing wins; ROIC.ai reclassifies deal/restructuring costs below the line.]

Quality of earnings — five forensic checks.

(a) The non-operating item, precisely identified. The ~$62M gap between operating income and net loss is not a tax-receivable-agreement charge and not a warrant/derivative mark — the 10-K states there were “no outstanding warrants and derivatives” at year-end and there is no TRA anywhere in the filings. FY25 “other expenses/(income)” of $38.1M is 96% a non-cash “change in the fair value of acquisition-related liabilities” ($36.7M) — the remeasurement of LiveIntent/Marigold contingent-consideration and earn-out liabilities, marked up in 2025 because Zeta revised the LiveIntent forecast higher. It is the mirror-image of a “GAAP loss that’s just a paper mark”: it drags reported net income below the operating line but is added back in both Adjusted EBITDA and the operating-cash-flow bridge, so it costs no cash, and it is volatile — in Q1-26 it reversed to a $3.0M other income. For valuation, strip it out both ways. This is why the GAAP net loss understates the business’s underlying operating profitability. [Source: FY2025 10-K, other-expense note; Q1-26 10-Q.]

(b) SBC vs. FCF — the illusion, with numbers. This is the crux of the entire thesis. FY25 operating cash flow was $198.9M; capitalized software ($20.1M) + capex ($13.8M) = $33.9M, so FCF ≈ $165.0M (the ~$179M figure some sources cite counts only software-dev capex). Stock-based compensation was $177.8M — larger than FCF. Therefore FCF less SBC = −$12.8M. The cash is real cash in the bank, but it is “funded” by ~$178M/yr of shareholder dilution; net of the real economic cost of paying employees in equity, Zeta generates roughly zero-to-negative free cash. Q1-26 makes it starker: SBC ($53.0M) alone exceeded operating cash flow ($49.7M). [Source: FY2025 10-K cash-flow statement and SBC note.]

© The GAAP-loss-vs-Adjusted-EBITDA gap. Net loss −$31.5M to Adjusted EBITDA +$278.7M is a $310M swing, of which SBC is $177.8M — 64% of the entire Adjusted-EBITDA number — plus the $38.1M contingent-consideration mark, $72.0M of D&A, and $23.4M of deal/restructuring. Adjusted EBITDA excluding SBC is $100.8M, a 7.7% margin — not 21.4%. Management’s headline profitability metric is majority-composed of the cost that is diluting the people it is presented to. [Source: FY2025 10-K Adjusted-EBITDA reconciliation.]

(d) Receivables vs. revenue. Accounts receivable grew to $322.4M from $235.2M (+37.1%) against revenue +29.7%; DSO rose from ~85 to ~90 days, and $16.2M of the balance is unbilled contract assets. A modest but real working-capital drag and a revenue-quality yellow flag — the more so given Culper’s (rebutted) round-trip/related-party allegations; treat as an open question for a channel check. [Source: FY2025 10-K; Q1-26 10-Q.]

(e) Capitalization and deferred revenue. Capex is genuinely light (~$34M all-in, ~2.6% of revenue) and software capitalization (~$20M) is not aggressive — a point in Zeta’s favor. But deferred revenue is not a tailwind: the $10.3M → $35.4M jump is Marigold’s acquired balance; organic deferred revenue actually fell ~$9.3M in FY25. This is a consumption/usage model — billings ≈ revenue, minimal forward visibility. Remaining performance obligations of $376.4M ($227.5M next-twelve-months) are a thin backlog on a ~$1.6B run-rate. [Source: FY2025 10-K revenue note.]

Returns and balance sheet. GAAP ROE is negative; ROIC is ~0.8% on $5.4M of GAAP operating income over ~$682M of invested capital, and even a generous “cash EBIT” construction (Adjusted EBITDA ex-SBC less capex, ~$67M) lands below a reasonable adtech WACC of ~10–11%. Full valuation allowances remain on US and UK deferred-tax assets given cumulative losses; the accumulated deficit is −$1,059.8M. Reported net cash is $122.7M (cash $319.8M less a $197.1M term loan) — but adding the $188.5M of acquisition-related liabilities ($149.0M current, including Marigold seller notes due ~Feb-2026) swings the adjusted position to a ~$66M net obligation. Book equity of $804.6M is 93% intangible: goodwill $527.9M + intangibles $217.9M = $745.8M, leaving tangible common equity of just $58.8M (it was negative as recently as 2023). [Source: FY2025 10-K balance sheet.]

** Verdict — economics do not improve with scale in any way that reaches the owner.** Gross margin is flat, the GAAP profit was $5M and already reversed, “Adjusted EBITDA” is 64% stock comp, and free cash flow net of that stock comp is negative. Returns on capital are below cost of capital. This is not a business demonstrating operating leverage; it is a business whose reported profitability is an accounting-and-add-back construction resting on ~19%/yr dilution.

7. Capital Allocation

A serial equity issuer. Zeta raised $229M net at $23.50/share in a 2024 follow-on, refinanced into a $550M facility ($200M term + $350M revolver) in August 2024, and continues to issue stock for both compensation and acquisitions. Weighted-average basic shares went 156.7M (2023) → 186.0M (2024) → 220.7M (2025) → 238.6M (Q1-26) — roughly +52% in two years. Because Zeta reports net losses, diluted equals basic in the EPS math, which hides a ~47M-share overhang (16.3M RSUs + 13.2M market-PSUs + 13.9M options + 11.6M non-market PSUs) that would surface the moment sustained GAAP profitability arrives. [Source: FY2025 10-K equity notes; Q1-26 10-Q.]

M&A — buying revenue with stock, banking goodwill. Both large deals were substantially equity-funded and goodwill-heavy: LiveIntent ($277.0M consideration — $55.8M net cash + 5.84M shares + $16.9M earn-out + $14.35M holdback; $176.5M goodwill) and Marigold Enterprise ($302.8M — $89.1M net cash + 5.33M shares at $17.27 + seller notes; $207.9M goodwill, 69% of the purchase price). Marigold is legacy Selligent/Cheetah Digital/Campaign Monitor enterprise email — a mature-to-declining carve-out, evidenced by the $273.1M of UK net operating losses acquired (a chronic-loss asset). At ~1.4x its ~$211M annualized revenue the price is optically cheap, but Zeta is paying with its own equity and ~$188M of contingent/seller obligations to acquire slowing assets that keep the growth narrative alive. Tellingly, the named executives received a discretionary “Marigold Deal Bonus” for closing it — management is paid extra to do dilutive deals. [Source: FY2025 10-K Notes 7–8; DEF 14A.]

The buyback vs. SBC — capital return or optics? Zeta repurchased $120.1M / 7.9M shares in FY25 (Q1-26: $25.7M / 1.5M). But the program is explicitly a “Stock Repurchase and RSA Withholding Program” — it doubles as the tax-withholding vehicle on executive RSA vesting. Against $177.8M of SBC and ~11.2M shares issued for the two acquisitions, the buyback offset well under half of the year’s dilution, and net share count still rose. This is not capital return to owners; it is a partial, cosmetic offset to the dilution treadmill — buying back stock near ~$15–20 while insiders distribute it (below). There is no dividend, and the credit agreement restricts distributions. [Source: FY2025 10-K; DEF 14A.]

Incentive alignment — circular and share-price-driven. The proxy is unflattering on alignment. NEO annual cash bonuses are tied to revenue and Adjusted EBITDA vs. budget and were paid at 200% of target in FY25 (revenue $1,304.7M and Adjusted EBITDA $278.7M beat guidance). So management earns double bonuses on an Adjusted-EBITDA metric that is 64% the very stock compensation they are paid — paid in stock, add the stock cost back to define “profit,” collect 200% on that “profit.” PSUs vest on a 20-day-VWAP share-price hurdle (190.1% earned when the stock hit a $21.54 VWAP), rewarding price rather than returns on capital. The tax note flags $31.9M of non-deductible officer compensation under Section 162(m) of the tax code, confirming outsized pay. To management’s credit, hedging and pledging are prohibited and a clawback policy exists — but the metric design is the problem. [Source: 2026 DEF 14A, CD&A.]

Governance — founder control. Zeta is founder-controlled through a dual-class structure: Class B carries 10 votes/share versus one for Class A, and Steinberg-affiliated entities (ACI Investment Partners) hold ~40% of total voting power. Minority holders have limited governance leverage. [Source: 2026 DEF 14A beneficial-ownership table.]

Insider behavior — systematic distribution, zero conviction buys. Across the 2025–26 corpus (188 Form 4s + 101 Form 144s), there are zero open-market purchases (code P) by anyone. Direct open-market Form 4 sales are minimal, but the real monetization runs through gifts: CEO Steinberg gifted ~4.7M shares and CFO Greiner ~2.2M shares (code G, which reports no sale price) into family/estate trusts (Family Trust IX/C5/S4/III, Southbeach Trust). Those trusts then file Rule 144 notices — ~$46.9M of sales in 2026 YTD across 16 filings (e.g., Family Trust III: 715,000 shares / $15.7M on 6/10/26). Layering distribution as “gifts” and then selling from the trusts under 10b5-1 is how the two people who know the business best are steadily reducing exposure — while the company spends $120M buying stock that partly provides their liquidity. [Source: EDGAR Form 4/144 corpus, 2025–2026.]

** Verdict — no, capital has not been allocated intelligently for outside owners.** The pattern is: acquire mature revenue with stock, add the stock back to manufacture “profit,” pay management 200% bonuses on that profit and PSUs on the share price, buy back a fraction of the resulting dilution for optics, and let insiders distribute stock through trusts. Value is being transferred from shareholders to employees and management via dilution, not compounded for them.

8. Changes and Headwinds — Last Two Years

Strategic / M&A. The two defining moves are LiveIntent (Oct-2024) and Marigold Enterprise (Nov-2025), which together added ~$500M of annualized acquired revenue, ~$384M of goodwill, and ~11.2M shares. The strategy is explicit consolidation of the enterprise-email/identity landscape onto ZMP. Concurrent product moves: the Athena by Zeta agentic-AI interface (2025) and the Palantir (2026-06-23) and Snowflake/OSI data-collaboration partnerships — narrative-stage but sentiment-relevant.

The Culper short report (2024-11-13) and its aftermath. Culper Research alleged Zeta operated “consent farms” (bait sites harvesting data) and round-tripped data revenue to flatter growth, estimating consent-farm economics at ~56% of LTM Adjusted EBITDA. The stock fell ~53% in three sessions. Zeta rebutted forcefully (“categorically false”; Apptness + ArcaMax combined <3% of revenue; Deloitte reaffirmed unqualified opinions). No restatement followed and the stock recovered — but multiple securities class actions (Rosen, Levi & Korsinsky, Hagens Berman; Bragar Eagel investigating as of Aug-2025) remain live, and a pre-IPO SEC subpoena concerning a data vendor Zeta says it stopped using in 2020 (disclosed on the Q2-2023 earnings call) shows regulatory scrutiny of the data model predates Culper. Assessed independently: not proven fraud, but a substantively-corroborated critique of the data-moat marketing and revenue quality that over-reached on the fraud conclusion. [Sources: Culper Research report; Zeta IR response, Nov-2024; class-action filings; Zeta Q2-2023 earnings call transcript.]

Financing / capital structure. New $550M credit facility (Aug-2024); $229M net equity raise at $23.50 (Nov-2024); buyback authorization raised to $200M (Jul-2025); Marigold seller notes (~$188M acquisition-related liabilities) added in Nov-2025.

Leadership / board. CEO David Steinberg (co-founder, chairman) and CFO Chris Greiner remain in place; the founder-control structure is unchanged.

Headwinds. (i) Organic-growth deceleration risk as the roll-up laps out; (ii) SBC re-acceleration — unrecognized SBC jumped $254.4M → $428.1M in a single quarter (Q1-26), signalling a fresh mega-grant and a forward SBC run-rate of ~$210–230M/yr, ending the multi-year decline; (iii) privacy-regulatory tightening aimed at the data model; (iv) emerging single-customer concentration (>10% of revenue in FY25); (v) the live litigation overhang; (vi) macro/rate sensitivity given a ~2.0 beta.

** Verdict — the last two years strengthened the reported growth story and weakened the quality story.** Acquisitions bought top-line and optical profitability; the Culper episode, the SBC pipeline, and the insider distribution deepened the skepticism a rigorous owner should hold.

9. Risk Analysis

Net risk read: the dominant risks are quality and valuation risks (growth deceleration, SBC/dilution, cosmetic profitability, regulation) amplified by a ~2.0 beta — not solvency risk. The tail risk is a validated data/revenue-quality finding; the base risk is a de-rate toward peers if the organic-growth or margin story disappoints.

10. Valuation Discussion (Embedded Expectations)

Framing. Zeta is a fast-growing, GAAP-thin, SBC-funded platform, so no single multiple governs; the valuation stands or falls on how one treats stock-based compensation and how durable the (partly acquired) growth is. P/E is not meaningful (FY25 GAAP EPS −$0.13; TTM −$0.10). The honest lenses are EV/Sales, EV/Adjusted-EBITDA, EV/FCF, and — the reframing lens — EV/FCF after charging SBC as the cash cost it economically is.

The multiples, reconciled. At $21.92 on ~248.7M shares, market cap is ~$5.45B; with $319.8M cash against $197.1M debt (net cash ~$123M), EV ≈ $5.33B. Against FY25 revenue of $1,304.7M that is ~4.1x trailing EV/sales; on the Q1-26 TTM base (~$1,437M) ~3.7x; on an estimated FY26 ~$1.55–1.60B, ~3.4x forward sales (assumption — confirm against guide). On EBITDA the picture splits by definition: management’s Adjusted EBITDA (~$279M) implies ~19–20x; GAAP/ROIC EBITDA ($100.8M) implies ~53x; the ~$178M gap is almost entirely SBC. FY25 FCF of ~$165M implies ~32x EV/FCF — but that FCF ≈ 100% of SBC, so on an SBC-inclusive basis EV/true-FCF is effectively infinite. [Sources: ROIC.ai enterprise value & cash flow; AZI price.]

Own-history read (AZI valuation_index). On the stock’s own five-year range, the P/S of ~3.49x sits in the 78th percentile — toward the rich end versus its own history, even though the stock is ~40% below its 2024 ATH; the reconciliation is that revenue has more than doubled since 2022, so a lower absolute price still embeds a higher sales multiple than the 2021–2023 norm (~1.9x EV/sales). P/B (5.54x) is only the 16th percentile and the composite is a middling 47th (no P/E). On sales, then, Zeta is not cheap against itself — it is near the top of its own band, just short of the 2024 blow-off (~6.9x intra-2024). [Source: AZI valuation_index.]

Comparable companies.

*Reported net cash; adjusted for ~$188.5M acquisition-related liabilities, Zeta’s position flips to a ~$66M net obligation.

What the comp table says. Zeta trades at the rich end of the martech cohort — roughly level with Klaviyo and above Braze, TTD and LiveRamp on EV/sales — despite being GAAP-thin, carrying a lower ~60% gross margin than Klaviyo (~75%) or TTD (~80%), and generating FCF that is entirely a function of adding back SBC. LiveRamp is the sobering anchor: a data-connectivity peer growing ~10% with real 13% EBITDA margins and genuine FCF trades at 1.7x sales — less than half Zeta’s multiple. The bull retort is Zeta grows ~2–3× faster; the bear retort is that ~half of Zeta’s reported growth is acquired and, once SBC is expensed, its “profitability” is thinner than RAMP’s. The market is paying a growth-and-AI-narrative premium the quality metrics do not independently justify. [Interpretation.]

Embedded expectations — reverse-DCF. Backing out what ~$5.33B of EV requires (10-yr FCFF, 10% discount, 3% terminal — a ~2.0-beta name arguably warrants a higher rate, which would only raise the bar) yields, roughly, combinations such as: ~15% revenue CAGR and ~15% terminal SBC-inclusive FCF margin; or ~12% CAGR and ~20% terminal true-FCF margin; or ~20% CAGR and ~10% terminal true-FCF margin. The common thread is stark: whatever the growth path, the price requires Zeta’s true, SBC-inclusive FCF margin to expand from ~0–1% today to ~12–20% — a transformation that has not begun. The market is not merely paying for visible growth; it is paying for a future margin structure that today exists only in the adjusted-EBITDA presentation. [Interpretation; assumptions explicit.]

Scenarios (implied fair EV — no price target).

• Bear (~$2.5–3.5B, ~2x sales): organic growth decelerates to ~8–10% as the roll-up laps and LiveIntent fades; SBC stays sticky; Culper concerns prove partly valid. Re-rates toward the LiveRamp/Braze zone. Material downside from ~$5.3B.
• Base (~$4.5–5.5B, ~3–3.5x sales): revenue compounds ~13–15%, true-FCF margin climbs to ~12–15% as SBC falls toward ~10% of revenue, GAAP operating income keeps inflecting. Roughly the current price — approximately fair value.
• Bull (~$8–10B+, ~4.5–5.5x forward sales): Palantir/Snowflake AI-data story converts into enterprise adoption; ~20%+ growth holds; true-FCF margin marches to ~20%; the GAAP inflection sustains. Meaningful upside.

What the market is pricing correctly vs. incorrectly. Correctly: the ~30% reported top line, the genuine 2025 GAAP operating-income inflection, and the multi-year decline in the SBC ratio. Optimistically: the durability of organic growth (headline flattered by acquisitions), the 0%→15%+ true-FCF-margin expansion baked into the reverse-DCF, and the near-complete dismissal of Culper’s data/revenue-quality thesis. The tape also appears to under-price the ~2.0 beta. Read: a fully-priced-to-modestly-rich growth story whose multiple already discounts a margin transformation Zeta has yet to demonstrate. (No recommendation; no target.)

11. Variant Perception

Consensus. Sell-side is broadly constructive: Zeta is framed as an AI-powered marketing-and-data cloud growing ~30% with a widening data moat, and the June-2026 Palantir partnership is read as validation of an AI-data-collaboration TAM. Bullish targets (e.g., BofA ~$28) rest on a re-rating toward AI-software peers, arguing Zeta is cheap versus AI-software comps on sales while inflecting to GAAP profitability. The consensus is momentum-and-narrative-driven — the stock is trusted again because it recovered from Culper and because Palantir re-anchored the AI story.

Strongest bull case. A genuine ~30% grower that just crossed into GAAP operating profitability with a declining SBC ratio, ~$165M reported FCF, and a roughly net-cash balance sheet — a scaling platform that self-funds rather than raising capital. Its consumer-intent data set plus Palantir/Snowflake open an enterprise data-collaboration and agentic-marketing adjacency that could sustain 20%+ growth for years; at ~3.4x forward sales it looks inexpensive versus Klaviyo (~3.8x) and the AI-software cohort, with clear operating leverage as SBC normalizes.

Strongest bear case. A roll-up whose reported growth overstates organic momentum (LiveIntent/Marigold inflate the +30%), on a lower ~60% gross margin than true software peers, whose entire free cash flow is manufactured by adding back $178M of SBC (true FCF ≈ $0). Culper’s 2024 data/revenue-quality thesis was never independently refuted, only outvoted by a recovering stock. The competitive position is contested from above (walled gardens/Amazon) and beside (TTD, Klaviyo, the marketing clouds), with no financially-visible moat (ROIC ~0.8%, gross margins trail peers). Insider distribution via trusts and heavy dilution round out a quality-of-earnings picture the adjusted metrics obscure. At ~4x sales for a GAAP-thin, SBC-funded, partly-acquired story, the risk/reward is asymmetric to the downside.

The 3–5 assumptions that matter most. (1) Organic growth — mid-teens+ compounding, or ~half-acquired? (2) SBC / true-FCF margin — does SBC fall toward ~10% of revenue so real FCF margins reach the mid-teens the price requires? (3) Data-sourcing / revenue quality — Culper noise or signal; is the opted-in data durable and compliant? (4) Competitive durability — can Zeta hold share against clouds/TTD/walled-gardens without margin-destroying price competition? (5) AI optionality — does Palantir/Snowflake convert into paid enterprise revenue, or stay a headline?

Falsification. Bull falsified if: a clean organic disclosure shows sub-teens organic; SBC/revenue stops falling or true FCF stays ~0 another year; or independent corroboration of Culper emerges. Bear falsified if: Zeta discloses durable ~20% organic growth with expanding GAAP operating margins and true-FCF margins visibly climbing into the teens; the buyback exceeds SBC (net share count falls); and Palantir/Snowflake produce disclosed, ramping enterprise revenue.

Factor-positioning read (the tape as evidence). In factor space Zeta is a high-beta (~1.96), small-cap, high-idiosyncratic-vol software name — Market is the dominant loading (R² ~0.35; ~65% of moves are stock-specific), with a strong negative low-volatility loading (behaves like the opposite of a defensive), a SmallSize tilt (+1.07), and — tellingly — Growth ~0, Value ~−0.07, and Momentum zeroed out. Despite the “AI-growth” narrative, the 2024 ATH → −70% drawdown → 2026 recovery has left it not loading as a clean growth or momentum name; its factor-similar peers are generic SMID-cloud-software and ETFs (XSW, CLOU, WCLD, Globant, Gartner, Q2, Paylocity, Pegasystems), not the DSP/adtech cohort — the model sees a high-beta software basket, not a differentiated adtech moat. Track record: strong absolute (y1 +44%, y3 +39% ann.) bought with a −70% max drawdown and a 0.28 five-year Sharpe. The recent path is a −10% six-month dip reversed by a violent three-month V-bounce (~+43% actual quarter) catalyzed by Palantir. What the tape is pricing: a sentiment-driven, narrative-led recovery in a stock that gets ~2× market beta — the Palantir rally is largely un-derisked and acutely macro/liquidity-sensitive. Consensus is arguably offsides in both directions depending on regime: too dismissive of downside beta if risk turns off, too anchored to the AI narrative if it does not convert.

12. Fact vs. Interpretation Table

13. Open Questions

1. What is the true, cleanly-disclosed organic growth rate — and how fast does it decelerate as LiveIntent/Marigold lap through 2026–27?
2. Does SBC actually normalize? The $254M→$428M jump in unrecognized SBC suggests the multi-year decline may be over — what is the forward SBC/revenue trajectory?
3. Was Culper signal or noise? Absent independent verification, the data-sourcing/revenue-quality allegations remain unresolved — is any portion of “integrated” or data-licensing revenue related-party or round-tripped?
4. Will the buyback ever exceed SBC so that net share count falls, or is it permanently a partial offset?
5. Does Palantir/Snowflake convert into disclosed, ramping enterprise revenue, or remain a narrative?
6. How exposed is the data model to the CA Delete Act / FTC data-broker enforcement, and what is the margin impact of stricter consent regimes?
7. How durable is the emerging >10% customer concentration, and what is the churn/renewal risk there?

14. What Must Be True

Bull thesis — what must be true, and its falsification test. Zeta must be a durable ~20%+ organic compounder whose SBC normalizes toward ~10% of revenue so that real, SBC-inclusive FCF margins expand into the mid-teens, converting the adjusted-EBITDA story into genuine owner cash flow; the data set must be durable, compliant, and differentiated enough to hold pricing against clouds/TTD/walled-gardens; and the AI-data adjacency (Palantir/Snowflake) must become paid enterprise revenue.

Falsification test: if, over the next 2–4 quarters, cleanly-disclosed organic growth prints in the low teens or below, or SBC/revenue fails to fall (true FCF stays ~0), or net share count keeps rising, the bull thesis is broken.

Bear thesis — what must be true, and its falsification test. Zeta must be a roll-up whose reported growth overstates a decelerating organic core, whose profitability is an SBC-add-back construction (true FCF ~0) on a below-peer gross margin and a below-cost-of-capital ROIC, with no financially-visible moat and a live regulatory/data-quality overhang — such that the ~4x-sales multiple de-rates toward the LiveRamp/Braze zone.

Falsification test: if Zeta discloses durable ~20% organic growth with expanding GAAP operating margins and true-FCF margins visibly climbing into the teens, the buyback exceeds SBC (net share count falls), and Palantir/Snowflake produce disclosed, ramping revenue, the bear thesis is broken.

APPENDIX A — Standard Diligence Questionnaire

Zeta Global Holdings Corp. (NASDAQ: ZETA) — as of 2026-07-10

Fact/Interpretation/Assumption labels where they matter.

General

What thoughtful questions have other investors asked? The dominant one is quality-of-earnings: how much of the ~30% growth is organic vs. acquired, and how “real” is the free cash flow given that SBC ($178M) exceeds it. Second is the Culper Research thesis (Nov-2024) on data sourcing (“consent farms”) and revenue quality — whether the “opted-in first-party data moat” is genuine or commodity data repackaged. Third is the founder-control / insider-selling / circular-comp governance set. Fourth, more recently, whether the Palantir/Snowflake AI-data partnerships are substance or narrative. (Interpretation.)

Cyclicality & Earnings Nature

• Cyclical high or low? Neither in the traditional sense; Zeta is a secular grower, not a classic cyclical. Modest political cyclicality (higher in presidential years — 2024; 2026 is a midterm partial tailwind). Reported margins are near a nominal high (first GAAP-positive op income) but that is add-back-driven and already reversed in Q1-26. (Fact/Interpretation.)
• External environment vs. internal actions? Growth is internally driven (sales execution + M&A) but valuation is highly external — a ~2.0-beta name whose price is macro/liquidity/sentiment-sensitive.
• Revenue stability? Consumption/usage model — billings ≈ revenue, thin backlog (RPO $376M; $227M NTM); NRR >110–115% provides base stability, but limited forward visibility.
• Market size / direction? Large, fast-growing, extremely fragmented (~14,000 martech tools); Zeta’s CDP/identity sub-segment grows ~mid-20s%+, but the profit pool is contested with walled gardens upstream. Domestic-led with international (UK/EU) exposure.

Business Quality & Competitive Moat

• Industry more or less competitive? More — AI lowers the barrier to shipping marketing copilots; incumbents bundle; walled gardens dominate the best data.
• Profitability (ROIC/ROE)? Poor. ROIC ~0.8% (below WACC ~10–11%); GAAP ROE negative; even cash-EBIT-ex-SBC returns are sub-WACC. (Fact.)
• Industry profitability / barriers? Low structural barriers at the point level; weak switching costs; the durable pricing power sits with Google/Meta/Amazon.
• Easily understood? Moderately — the “platform” label obscures that ~26% of revenue is integrated/media pass-through and growth is ~half acquired.
• Undermined by low-cost labor? Not directly; it is a data/software business, though offshore engineering is common industry-wide.
• Do brands matter / nature of competition? Competition is on data breadth, channel coverage, AI features, and sales relationships — not brand. Switching costs are real for embedded enterprise deployments (data integration, workflows) but modest at the module level.
• Customer switching costs? Moderate once ZMP is embedded as the system of record; the “One Zeta” consolidation pitch raises them — the genuine (if not wide) part of the moat.

Financial Condition & Balance Sheet

• Assets not fully recognized? The consumer data set/identity graph is largely internally-developed/harvested and not capitalized at “moat” value — but by the 10-K’s own description it is commodity-adjacent, so this is not hidden franchise value. (Interpretation.)
• Off-balance-sheet / contingent liabilities? Yes — $188.5M of acquisition-related liabilities (contingent consideration, earn-outs, Marigold seller notes); adjusting for these flips reported net cash ($122.7M) to a ~$66M net obligation. Operating leases modest. (Fact.)
• How conservative is the accounting? Mixed. Conservative on capex/software capitalization (light). Aggressive in presentation: Adjusted EBITDA is 64% SBC; diluted = basic (hides a ~47M-share overhang while loss-making). The contingent-consideration mark is non-cash and volatile.
• CapEx-hungry? No — ~2.6% of revenue all-in. Asset-light.

Capital Allocation & Management

• FCF generation and use? Reported FCF ~$165M, but net of $178M SBC it is ~−$13M. Uses: ~$120M buyback (a partial dilution offset + RSA-withholding vehicle), ~$90M acquisition cash, seller-note obligations. No dividend.
• Significant acquisitions? Yes — LiveIntent ($277M, 10/2024) and Marigold Enterprise ($302.8M, 11/2025, 69% goodwill, declining legacy-email assets), plus prior tuck-ins. Serial roll-up funded substantially in stock.
• Buying back shares? Yes ($120M FY25) but offset <half the dilution; net share count still rose.
• Issuing shares to insiders? Yes — heavy SBC ($178M/yr, ~13.6% of revenue), unrecognized SBC ballooning $254M→$428M; shares +52% in 2 years.
• Compensation policy? Cash bonuses at 200% of target on revenue + Adjusted EBITDA (which adds back SBC — circular); PSUs on 20-day-VWAP share-price hurdles (not returns on capital); $31.9M non-deductible officer comp under Section 162(m). Hedging/pledging prohibited; clawback exists. (Fact.)
• Motivations of management? Founder-controlled (dual-class, ~40% voting via ACI entities); incentives tilt toward share price and adjusted metrics, not per-share owner value. Insiders distribute via trust-routed gifts + Rule 144 (~$46.9M 2026 YTD); zero open-market buys. (Fact.)

Valuation & Market Data

• ADR / MLP / K-1? No — US C-corp, Class A common stock; no K-1.
• Dividend policy? None; restricted by the credit agreement.
• Profitability? GAAP thin-to-negative; “adjusted” profitability is add-back-driven; true SBC-inclusive FCF ~0.
• Net income vs. cash from operations diverging? Net loss −$31.5M vs. OCF +$198.9M — a wide gap driven by SBC ($178M), D&A ($72M), and the non-cash contingent-consideration mark ($37M). The divergence is expected for an SBC-heavy model but is exactly why the cash figure flatters.

Risks & Downside

• What would cause the stock to decline? Organic-growth deceleration; SBC re-acceleration / continued dilution; a validated data/revenue-quality finding; privacy-regulation tightening; multiple compression toward peers; a risk-off drawdown amplified by the ~2.0 beta.
• Catastrophic loss risk? Low absent a fraud/restatement finding — the business has real revenue, cash, and a roughly net-cash balance sheet.
• Total loss? Unlikely in the base case; the tail is a regulatory/legal finding that impairs the data model.

Recent News & Events

• Environment changed recently? Yes — the Palantir partnership (6/23/26) and Snowflake/OSI initiative re-ignited the AI-data narrative and drove a ~51% recovery off the March-2026 low. Sentiment-relevant; not yet in the numbers.
• Significant acquisitions? Marigold Enterprise closed 11/24/2025.
• Accounting-policy changes? None material beyond acquisition-accounting mechanics.
• Recent changes — markets/facilities/management? Founder/CFO unchanged; continued M&A-led expansion; Athena agentic-AI product launched (2025); disclosure shifting toward the “super-scaled” cohort only.

APPENDIX B — Source Appendix

Zeta Global Holdings Corp. (NASDAQ: ZETA) — as of 2026-07-10

Primary sources first. All financial figures reconciled to SEC filings where possible; third-party aggregators (ROIC.ai, AZI, FactorsToday) used for computed ratios, price series, and factor data and reconciled to filings. Management commentary treated as hypothesis, validated against filings and external evidence.

Primary — SEC filings (EDGAR CIK 0001851003)

• Form 10-K, FY2025 (filed 2026-02-25) — MD&A, revenue disaggregation (Direct 74% / Integrated 26%), Adjusted-EBITDA reconciliation, SBC note, other-expense note (change in FV of acquisition-related liabilities $36.7M), acquisition notes (LiveIntent, Marigold; Note 7 pro-forma), balance sheet, risk factors, seasonality. https://www.sec.gov/Archives/edgar/data/1851003/000119312526...zeta-20251231.htm
• Form 10-Q, Q1-2026 (filed 2026-05-01) — Q1-26 revenue $396.3M, operating loss −$18.8M, SBC $53.0M, unrecognized SBC $428.1M, customer concentration (Note 10). https://www.sec.gov/Archives/edgar/data/1851003/000119312526199156/zeta-20260331.htm
• Forms 10-K FY2021–FY2024 — multi-year revenue, margin, SBC, and share-count history.
• DEF 14A (2026 proxy, filed 2026-04-24) — CD&A (200%-of-target bonus on revenue + Adjusted EBITDA; PSU VWAP hurdles; “Marigold Deal Bonus”; (m) $31.9M), dual-class voting (Class B 10 votes; ACI Investment Partners ~40% voting power), beneficial ownership.
• Forms 4 / 144 (2025–2026 corpus, 188 Form 4s + 101 Form 144s) — insider transaction read: zero open-market purchases; CEO Steinberg ~4.7M shares and CFO Greiner ~2.2M shares gifted (code G) to family/estate trusts; ~$46.9M of Rule 144 sales in 2026 YTD.
• Form 8-K corpus (2024–2026) — $550M credit facility (8/2024); LiveIntent close (10/2024); Nov-2024 cluster (Culper report window, $242M equity raise at $23.50, initial $100M buyback); buyback raised to $200M (7/2025); Marigold agreement/close (9–11/2025); FY25 and Q1-26 earnings; 6/16/26 8-K = annual-meeting voting results (Item 5.07).

Primary — earnings calls / transcripts

• ROIC.ai transcript tools: Q1-2026 (2026-04-30), Q4-2025 (2026-02-24), Q3-2025, Q2-2025 earnings calls — management framing of guidance, organic-vs-acquired growth (ex-Marigold/political), ARPU/NRR, cost-of-revenue mix (social channel), Athena.
• Q2-2023 earnings-call transcript — disclosure of a pre-IPO SEC subpoena re: a data vendor Zeta says it stopped using in 2020.

Third-party — quantitative

• ROIC.ai — income statement, balance sheet, cash flow, profitability/credit/liquidity ratios, enterprise value ($4.86B YE25), valuation multiples (accessed 2026-07-10).
• AZI — 5-year adjusted daily price/OHLCV series (IPO 2021-06-10 → 2026-07-09), EMAs, beta; valuation_index own-history percentiles (P/S ~3.49x = 78th; P/B 5.54x = 16th; composite 47th); news feed.
• FactorsToday — stock-loadings (beta ~1.96; SmallSize +1.07; LowVol −0.58; Growth ~0; Momentum zeroed; R² ~0.35), leaderboard (y5 return +22% ann., y5 max DD −70%, y5 Sharpe 0.28; m3 V-bounce), stock-info, related-stocks, specific-vol (accessed 2026-07-10).

Secondary — external

• Culper Research short report (2024-11-13) — “consent farm” / data-sourcing and revenue-quality allegations; and Zeta’s investor-relations rebuttal (Nov-2024).
• Securities class-action filings (Rosen; Levi & Korsinsky; Hagens Berman; Bragar Eagel investigation, 2024–2025).
• BofA Securities note (maintains Buy; PT raised to $28, 2026-06-24) — cited as consensus data point, not as valuation input.
• Benzinga / trade-press coverage of the Palantir partnership (2026-06-23) and Snowflake/OSI marketing-data initiative (2026-06-29).
• chiefmartec 2025 martech landscape (~14,000 tools) — industry fragmentation.
• Peer filings/IR: The Trade Desk (TTD), LiveRamp (RAMP), Klaviyo (KVYO), Braze (BRZE) — comparable revenue, gross margin, EV/sales.

Note: management commentary and third-party analyst targets are treated as signal/hypothesis, not evidence; no analyst target is adopted as a price target. Figures reconciled to primary filings where they drive a verdict.