Ecommerce · Deep dive

ZyG

Ex-ironSource founders and Israeli Unit 81 AI operators betting that 60+ specialised agents on a unified data layer can do — cheaper and without buying the brand — what Thrasio raised $3B to try.

emerging

The question that decides it: Does an AI agent stack close the gap on the two things that killed the aggregator era — sustained brand equity and predictable CAC — or does it just make Thrasio's mistakes cheaper?

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

ZyG is a Tel Aviv-based agentic OS for DTC ecommerce, built by the founders of ironSource — the mobile-ad platform that IPO’d on the NYSE at an $11.1B market cap in 2021 and merged into Unity in late 2022 — reunited with three AI leads from Israel’s Unit 81. It exited stealth in March 2026 with a $58M seed and, on 5 May 2026, added a $60M Series A led by Accel at $500M post-money, taking total funding to $118M in under a year. The pitch: 60+ specialised AI agents on a unified data layer take a founder’s product from a “ZyG Score” pre-launch prediction through store, brand, creative, paid acquisition, CRO, retention and logistics — while ZyG collects a rev share instead of buying the brand. Everything about that echoes the Amazon-aggregator boom of 2020-2022, which ended badly. Whether the AI wrapper changes the underlying economics is the whole question.

Founding story

The lineage matters more than the product deck. Bar-Zeev ran ironSource as CEO for 13+ years, took it public via SPAC on the NYSE in June 2021 at an $11.1B market cap (the largest Israeli SPAC to that point), then negotiated the November 2022 all-stock sale to Unity at $4.4B — a haircut from the 2022 ad-tech collapse, but still a rare Israeli exit at that scale (Globes, CTech, 2022). He stayed on as President of Unity’s Grow segment until January 2024, when he “mutually agreed” to leave amid Unity’s restructuring (SEC 8-K, 10 Jan 2024), and kept his Unity board seat until February 2026 — weeks before ZyG stepped out of stealth.

Around him he reassembled a slice of the ironSource bench — Omer Kaplan (CEO), Assaf Ben Ami (CFO/COO), Nadav Ashkenazy and Daniel Shinar — and added the differentiating half: Dr. Eyal Amitt, Omri Steinmetz and Guy Tsur, three AI operators from Unit 81, the IDF’s technical sister to 8200 (the pipeline that fed Check Point, Palo Alto, Wiz and hundreds more). The framing: ironSource was a game-industry operating layer — user acquisition, monetisation, growth tools — and DTC ecommerce is the same problem waiting for the same treatment. ZyG was incorporated in January 2025 and ran quietly for ~14 months before the March 2026 launch.

How it works

The loop starts before a product exists at scale. A brand brings a product; ZyG runs it through an “agentic Product-Market Fit test” — validation agents scrape signals, seed small live campaigns, and produce a ZyG Score predicting its ability to grow into a “massive, profitable DTC brand.” That score is the gate. Below the threshold, no partnership; above it, ZyG signs and takes over the operational stack.

From there the platform runs the brand’s digital layer end-to-end. More than 60 specialised agents share a single data spine — agents for brand identity, store build, creative generation (static, video, UGC-style), paid acquisition across Meta/TikTok/Google, organic and SEO/AEO, influencer sourcing, CRO, lifecycle/retention, and last-mile logistics. The design point is that all agents read and write to the same unified customer, catalogue and creative graph, so a creative iteration on TikTok updates the CRO agent’s landing-page test and the LTV model on the same cohort within the same tick. ZyG also runs three of its own brands internally as live sandboxes for new agents before they touch a partner brand (Retail Technology Innovation Hub, May 2026) — a real operational detail, but also how Thrasio’s ops team ended up rationalising its portfolio. The difference: ZyG’s sandboxes exist to train agents, not to make the margin.

Product and business overview

The product is one platform sold as a partnership, not a subscription. Named modules track the loop: ZyG Score (pre-launch prediction), Brand Studio (identity, positioning, storefront), Creative (assets across formats), Acquisition (paid social/search), Growth (organic, SEO/AEO, influencer), Retention (CRM, lifecycle, LTV), Ops (logistics), and wrapping them, Growth Capital — cohort-based, predictive-LTV underwritten financing. The customer is a DTC founder or small brand with organic traction and a product ZyG believes will score; only a subset get partnered. Brands keep 100% of revenue and IP; ZyG earns on a consumption/pay-as-you-grow basis. This is the point ZyG hammers to differentiate itself from the aggregators.

Business model and pricing

ZyG has not published a price sheet, which is itself informative. The disclosed model has three parts. First, a consumption fee — “pay as you grow” — that scales with the brand’s revenue; ZyG earns nothing until the brand grows. Second, an implicit rev share embedded in that fee, though the specific take rate is not public. Third, an optional growth-capital line whose underwriting mechanic is the interesting bit: rather than lending against historical revenue like a Wayflyer or Clearco, ZyG claims to underwrite against cohort-based predictive LTV built from its own agent-generated signal, letting earlier-stage brands access larger cheques than a revenue-based financier would extend.

Elegant on paper. In practice it collapses two risks into one balance sheet: ZyG’s confidence in its LTV model is also its credit-risk model. If the model is wrong on a cohort, ZyG is short both the marketing spend it deployed and the capital it advanced against the same cohort. Aggregators at least owned the asset they were financing. ZyG is closer to a marketing agency that has decided to lend on its own performance forecasts.

Traction over time

There is no disclosed count of partner brands, no disclosed GMV, no disclosed take rate. The single quantitative traction number in the public record is founder guidance in the same news cycle as the Series A.

Market analysis

Global DTC ecommerce is ~$296B in 2025, projected to reach $319.6B in 2026 at ~7.8% CAGR (Swell, Ringly, 2026). The narrower agentic-commerce category is where ZyG lives and where the numbers get speculative: Grand View pegs agentic commerce at $5.7B in 2025 rising to $65.5B by 2033 (35.7% CAGR); Mordor sizes agentic AI in retail and ecommerce at $60.4B in 2026 growing to $218.4B by 2031.

Two structural forces underpin the pitch. DTC CAC has risen ~222% over eight years, the average brand loses $29 per new customer, and only ~28% of first-time buyers ever come back (Ringly, 2026) — something has to give. And AI ad-buying, creative generation and lifecycle automation genuinely work well enough now to displace agency hours. The counterpoints are equally structural: much of DTC’s incremental growth has migrated to TikTok Shop ($15.8B US GMV in 2025, +108% YoY), which is not primarily a brand-site channel; and the tools ZyG bundles are increasingly available a la carte or free inside Shopify.

Competitive intel

Thrasio (post-bankruptcy) is the shadow. Peak >$10B, >$3B raised, Chapter 11 Feb 2024, emerged mid-2024 with $495M of debt eliminated and a smaller portfolio of “brands that actually make money” (CNBC, 2024). ZyG’s answer — “we don’t buy the brand” — is real, but every operating problem Thrasio hit lands on ZyG’s rev-share P&L too. Razor Group / Perch is the same story: Razor absorbed Perch mid-2024, combined value now less than the ~$2B they raised separately (Modern Retail, 2024-25); Berlin Brands, SellerX, Heyday and Benitago collectively raised billions and are shrunk, absorbed or gone.

Shopify Magic + Sidekick + Meta Advantage+ are free and bundled; they set the price floor at $0 for much of what ZyG’s agents do. Klaviyo AI, Attentive, Rebuy own the retention/CRM boxes on ZyG’s org chart today, with real revenue and DTC-operator trust. AdCreative.ai, Pencil, Icon, Arcads are the point-solution creative layer at $39-$119/month starting tiers (DTC Skills, 2026) — a founder who buys piece by piece keeps 100% of upside for four figures a month, not a rev share.

TikTok Shop is the harder threat because it is a rival channel definition, not a rival platform: if discovery and purchase both live inside TikTok, the DTC-site brand ZyG is optimising is a smaller share of the customer relationship. Wayflyer, Clearco, Ampla compete with the financing leg using cohort/LTV underwriting without demanding operational control.

History and evolution

• Jun 2021: ironSource lists on NYSE via SPAC at $11.1B market cap.
• Jul 2022: Unity announces all-stock acquisition at $4.4B — a 74% premium but a >60% haircut to IPO cap.
• Nov 2022: Merger closes; Unity holds ~73.5%, ironSource shareholders ~26.5%.
• Jan 2024: Bar-Zeev departs Unity’s Grow role in the restructuring; retains board seat.
• Jan 2025: ZyG incorporated in Tel Aviv; ~14 months of stealth build.
• Feb 2026: Bar-Zeev steps off Unity board.
• 5 Mar 2026: Public launch; $58M seed with Bessemer, Viola, Lightspeed.
• 5 May 2026: $60M Series A led by Accel at $500M; Felix Capital joins; total $118M in five months.

A fast, well-capitalised push with reputationally heavy syndication. The stumble that matters — if the model doesn’t work — is still years out.

What people say

The case for. The founder pattern has moved the market. Trade press (TFN, Ventureburn, TheNextWeb, Crowdfund Insider, all May 2026) frames ZyG as ironSource’s growth playbook translated from mobile gaming into DTC, with an AI stack built by Unit 81 operators. Accel’s Sonali De Rycker took the board seat, and Bessemer and Lightspeed followed on from seed to Series A within two months — the ecosystem reads that as a signal the round was priced against real early data. Analysts credit the architectural choice: run your own brands as agent sandboxes first, and share one data layer across all agents rather than orchestrating a Zapier of separate SaaS. And the pay-as-you-grow rev share addresses the loudest DTC-founder complaint about aggregators — the loss of IP and upside.

The complaints. Consistent across specialist coverage and LinkedIn DTC-operator commentary through spring 2026. First, this looks like a Thrasio in AI clothing: same claim, same category (Meta/TikTok/Google-fed DTC), same underwriting problem (predicting cohort LTV and CAC in a volatile ad market), same likely failure mode — an over-hired ops layer against an under-performing portfolio. Aggregators raised >$16B collectively and the survivors are worth a fraction of that; that history is <18 months old. Second, the moat is unclear. Every module has a well-funded incumbent, and Shopify Magic plus Meta Advantage+ set a $0 floor on much of the automation. “Sixty agents” is a systems argument, not obviously a product moat. Third, the financing leg is risky: underwriting your own performance forecasts collapses model and credit risk into one balance sheet, and pre-launch LTV prediction is unproven at scale. Fourth, disclosure: no partner-brand count, no GMV, no take rate, only founder revenue guidance in a launch cycle — normal for stealth-plus-five-months, but it is how the aggregator era began, too. Fifth, channel drift: TikTok Shop and creator-led discovery are moving GMV out of the brand-site funnel ZyG is optimising.

Outlook: the open question

ZyG resolves on whether an agent stack changes the underlying economics of a DTC brand portfolio, or whether it just makes Thrasio’s mistakes cheaper. For it to break ZyG’s way, three things have to be true. First, the ZyG Score has to actually predict LTV on unseen products well enough to earn its role as a gating filter — not just look plausible in retrospect. Second, the unified-data agent architecture has to deliver a real compounding advantage over a founder assembling Klaviyo + AdCreative + Shopify Magic + Wayflyer + a fractional agency for a few thousand dollars a month; “sixty agents” is a claim, not a moat. Third, and hardest, the cohort-based growth capital has to underwrite well enough that losses stay small when the model is wrong — because unlike aggregators, ZyG doesn’t own the asset it is financing, and unlike a lender, it is also on the hook for the marketing spend that generated the cohort.

The other way is easier to picture. TikTok Shop and creator-led commerce keep migrating GMV out of the brand-site funnel. Bundled free automation from Shopify and Meta compresses what ZyG’s agents can charge for. The Series A vintage of DTC brands turns out to have the same CAC and retention economics as the last, and the growth-capital book takes losses. In that world, ZyG becomes a very well-funded operational agency with an unusually complicated revenue model. The pedigree is rare — a CEO who ran a public $11B company through a full cycle, an Accel partner on the board, three Unit 81 operators on the AI — which is why the round cleared at $500M on five months of public existence. But every dollar of the $118M is priced against the assertion that this DTC vintage is different because there is AI in it. The next twelve months — disclosed partner counts, cohort performance, a first material loss on the growth-capital book, and the composition of the next raise — will tell you whether that is true.

How a challenger would attack it

Unbundle the rev share. ZyG’s exposed flank is its own pricing logic: the platform bundles capabilities whose price floor is being set at $0 by Shopify Magic, Sidekick and Meta Advantage+, and at $39-$119 a month by the point-solution creative layer — then charges a revenue-linked consumption fee plus an undisclosed take rate for the bundle. A challenger ships the unified-data-layer architecture — the one genuinely novel claim — as transparent flat-fee software: agents on the founder’s own Shopify and Klaviyo stack, no gating ZyG Score, no operational takeover, founder keeps 100% of upside. Every founder ZyG rejects below its score threshold is the challenger’s lead list. The second vector is the balance sheet: ZyG collapses model risk and credit risk into one book by underwriting growth capital against its own LTV forecasts — a Wayflyer-style financier that partners with, rather than operates, its borrowers takes the financing leg without the sandbox conflict of ZyG running three house brands on the same data spine. Third, go where the funnel went: ZyG optimises the DTC-site loop while TikTok Shop ($15.8B US GMV, +108%) moves discovery and checkout inside the feed — an agent stack built creator-first attacks the channel ZyG’s ironSource-era playbook wasn’t designed for.

Same playbook, new buyer

The playbook — predictive scoring, shared-data agents, performance-linked fees, capital underwritten by your own model — is not DTC-specific; ZyG chose the vintage with the worst recent history. The cleaner ports: app and game studios, where the ironSource founders’ actual domain expertise lives and where user-acquisition economics are better instrumented than DTC’s $29-loss-per-customer funnel; Amazon-native sellers, where the aggregator collapse left thousands of profitable brands with no operating partner and where Thrasio’s shrunken estate proves demand for operations without ownership; and B2B/SMB commerce, where nobody has run the score-then-operate motion at all. Geographically, ZyG is optimising the US social-ad funnel from Tel Aviv; Southeast Asian and Latin American DTC markets have the growth without the incumbent tool stack — no Klaviyo trust, no entrenched agencies — making a bundled agentic operator more defensible there than against America’s a-la-carte alternatives. ZyG won’t chase these soon: its $500M valuation was priced on the agentic-DTC story its investors bought, its ZyG Score is trained on brand-site cohort data, and with under 100 people and three house-brand sandboxes tuned to one funnel, every vertical port means retraining the entire data spine.

Sources and further reading

• PR Newswire — “ZyG announces $60M Series A Round Led by Accel” (5 May 2026)
• Calcalist / CTech — “ZyG raises $60 million Series A at $500 million valuation just a year after launch” (May 2026)
• Calcalist / CTech — “ironSource founders raise $58 million in Seed funding for AI e-commerce startup ZyG” (Mar 2026)
• PR Newswire — “ZyG Launches an Agentic Operating System for eCom Scale” (5 Mar 2026)
• Tech Funding News — “ZyG grabs $60M from Accel for its agentic eCommerce platform” (May 2026)
• Retail Technology Innovation Hub — “DTC platform ZyG announces blockbuster $60 million Series A” (5 May 2026)
• Globes — “ironSource completes merger with Unity” (Nov 2022)
• SEC — Unity Software 8-K on Tomer Bar-Zeev departure (10 Jan 2024)
• CNBC — “Top Amazon aggregator Thrasio files for bankruptcy” (28 Feb 2024)
• Modern Retail — “Amazon Briefing: After a disastrous two years, here’s how aggregators are valuing brands now” (2025)
• Ecommerce Fastlane — “TikTok Shop In 2026: The $15B+ DTC Opportunity” (2026)
• Grand View Research — “Agentic Commerce Market Size & Share Report, 2026-2033”