Field notes from the founders
ZyG vs Signal over Noise: hire the chef, or own the kitchen?
November 4, 2026
In March, ZyG came out of stealth with a $58m seed round. By May it had added a $60m Series A led by Accel, at a reported $500m valuation. That's $118m to build what it calls "an Operating System for eCom scale", from the people who founded ironSource. Nobody in agentic e-commerce is making more noise right now.The pitch lands because the pain is real, and it's worse than ZyG lets on.Complexity in e-commerce doesn't scale with revenue. Running a $10m brand is almost as complicated as running a $100m one. You need the same tooling, the same tracking and the same ad accounts, run by someone who knows what they're doing. You still have to do LTV properly, with real cohorts and contribution margin, which is much harder than most dashboards make it look. And you still have to wrestle with attribution, which, if we're being straight about it, is an unsolvable problem: you're trying to prove causation from a pile of correlations and a pixel that's been squinting since Apple introduced App Tracking Transparency.The $100m brand has a data team to soak all that up. The $10m brand has a founder, a freelancer and a Google Sheet called LTV_final_v3. For an early-stage founder it's worse still. Building a Shopify site is easy. Scaling an e-commerce business is bloody hard, and it takes deep technical expertise that most founders never signed up for.What they signed up for was the product: the formula, the fabric, the recipe. For a lot of passionate founders the stack is a necessary evil that eats the time, energy and money they'd rather spend on the thing they care about. Offer those founders a team of agents to run the whole digital layer, which is roughly what ZyG means by "a single platform that replaces the fragmented tools, agencies, and manual workflows", and they'll be thrilled. They'd be right to be. It's a good idea, and ZyG has the money and the pedigree to do it properly.So the interesting question is what happens next. The business gets more complex and more custom, you want more control, and you'd rather not hand over quite so much money. Who's holding the pen then?A cooking analogy helps here.ZyG is a talented chef who turns up with their own kitchen. You hand over the ingredients (your product, your customers, your ad budget) and they cook, using their agents, their data layer and their models. You keep the name over the door, and ZyG is explicit that partners keep "full control of the brand IP". You get fed, quite possibly very well.Signal over Noise works the other way round. You keep your kitchen: Shopify, Klaviyo, Recharge, Meta, Google and whatever else is already there. Our agents cook in it alongside you, with every ticket, every decision and every move out on the pass where you can see it.You might well get the same dinner either way. The deal behind it is very different, and the difference shows up in three places.The brand IP is the sign over the door. The recipes are a separate thing: which creatives worked for which cohort, why LTV dipped in March, the pricing test that added two points of margin. That accumulated operating knowledge is what makes a D2C business work, and it may well be worth more than the logo.When execution runs inside someone else's system, that knowledge builds up inside someone else's system. ZyG's public material doesn't say what happens to the data, the models or the playbooks when a partnership ends. We haven't seen their contracts, so we'd suggest asking them directly. "If we part ways, what leaves with me?" is a fair question for any vendor, us included.Our answer is short. Your data is your data. It's mastered in your own systems, and you can query it in plain English from Slack or our app whenever you like. Every number we produce, CAC and LTV included, is rebuilt on one definition and cited back to its source. If we part ways, your data doesn't come with us, and your Shopify store and Klaviyo flows carry on exactly as they were.ZyG describes a "connected system of specialized AI agents" running on proprietary models for LTV, attribution, churn, pricing and inventory. Clever people built it, and it may be excellent. From the brand's side of the pass, though, it's a closed system. You see the plate and very little of the method.That matters more than it sounds, because confident wrong is far more dangerous than obviously wrong. An attribution model that over-credits paid social will happily scale your budget off a cliff, and if you can't see its inputs you have no way to argue with its outputs. You'll just notice, eventually, that the plates are getting smaller.Our agents do the work too. We think that's what operators want, and we expect them to take on more of it over time. The difference is that they work in the open. Each move comes with its evidence, its expected effect and a one-click reversal. Each action is logged with what was done, when, and on whose authority. The conversation about every decision sits where your team already talks, in the app, Slack or Teams. You keep full visibility and full control, and your team learns from every call the agents make, which compounds in a way that simply receiving outcomes can't.ZyG charges what it calls a "pay-as-you-grow" consumption fee, and several reports describe it as a revenue share. The rates aren't published. ZyG's case is that incentives are aligned, since it only wins when you do, and that the fee largely replaces tool and agency costs you'd otherwise carry. It also offers financing and says it "takes on the majority of the financing and growth risk". Those are fair points, and carrying someone else's risk has real value, which we come back to below.The shape of the curve matters more than the rate. A fee pegged to growth gets bigger precisely as you get better at your job. Every percentage point of a $5m brand is $50k a year. At $20m it's $200k, for broadly the same software doing broadly the same thing. Your best year becomes your vendor's best year too.If the revenue-share reports are right, the chef takes a slice of every lunch you sell, and the busier the restaurant gets, the bigger the slice. It's all agreed up front, so calling it lunch money would be unfair. Think of it as a service charge you can't opt out of.Signal over Noise starts from $500 a month, scaled by company size, usage and seats. Your revenue doesn't come into it. Grow from $5m to $20m and our share of your P&L gets smaller. That's the deal we'd want if we were on your side of the table.Some brands should, and we'd happily point them there.If you're pre-product-market fit, with a promising product, not much data, no team and no appetite to build one, ZyG is built for you. It starts by testing whether a product can scale at all ("clear go/no-go signals", in its words), then runs and helps finance the growth if the answer is yes. When you don't have a kitchen yet, a chef who brings one along with the budget is exactly what you want.We're the wrong choice at that stage, and it's better to say so here than on a sales call. Our rough floor is about $1m in annual revenue, and the reason is statistical. Below that size the samples are usually too small to separate signal from noise (the name was a clue). A retention cohort of 200 customers will tell you a story, but you shouldn't bet the ad budget on it. We earn our keep by rebuilding CAC and LTV from your own history and sharpening decisions across a stack that has grown complicated enough to need it. With six weeks of orders, three tools and no repeat customers, there's little history to rebuild and not much stack to untangle. The most useful thing you can do then is find out whether anyone wants the thing.Once you've found fit, the trade flips. You have a team, a stack, cohorts you can read and opinions about your own numbers. You'd rather keep the kitchen and make better decisions in it, faster, with your name on them. That's also about the point where a fee that grows with you starts to sting.
The chef and the kitchen
1. Who owns the recipes
2. Black box or glass box
3. A fee that grows with you is a fee that grows
Who should pick ZyG?
| ZyG | Signal over Noise | |
|---|---|---|
| Model | Runs your digital growth for you on its own platform | Works inside the stack you already own |
| Your tools | Positioned to replace "fragmented tools, agencies, and manual workflows" | Keeps Shopify, Klaviyo, Recharge, Meta, Google and the rest |
| Your data | Exit terms not covered in public material | Mastered in your systems, queryable in plain English, stays yours |
| Visibility | Proprietary agents and models | Every move shown with evidence, expected effect and source-cited numbers |
| Control | ZyG executes | Agents execute in the open, with a full action log and one-click reversal |
| Pricing | "Pay-as-you-grow" fee tied to growth (rates not published) | From $500/month, by size, usage and seats |
| Best for | Pre-PMF products that need validating, running and financing | Post-PMF D2C teams (roughly $1m+ revenue) who want to keep the pen |
Five questions to ask any growth partner (including us)
- If we part ways, what data, models and playbooks leave with me?
- Can I see the inputs behind a recommendation, or only the outcome?
- Can I see everything the agents have done, and undo it?
- What happens to the fee if my revenue doubles?
- Which of my existing tools stay, and which do you replace?