AI-powered targeting across email, WhatsApp, SMS and push, trained on your data, tuned to your customers, and measured in real revenue against a control group.
See the lift first. Pay only if it works.
of retention spend goes to customers who would have bought anyway, or never would
typical conversion when campaigns go to everyone
purchase cycles a customer must complete before most brands break even
Rising ad costs. Growing customer lists. And a marketing stack still blasting the same email, SMS and push to everyone, training customers to tune you out and quietly eroding your margin.
Every brand we meet has tried the same playbooks, rules in their ESP, RFM buckets, a few LLM prompts. They plateau at the same place, because none of it is actually built on your data.
Off-the-shelf tools and LLM prompts give every brand the same playbook. They've never seen your buyers, your catalog, or your margins.
The signal that decides who to reach, and on which channel, lives in your first-party data. Nobody else has it, and no foundation model will ever be trained on it.
A customer engagement engine trained on your business. It learns who to reach, when, and where, and gets sharper every campaign you run.
AI scores every customer on how likely they are to buy again in the next 30–90 days, and picks the channel they'll actually respond to, email, WhatsApp, SMS or push. Same send, more revenue.
Spot high-value customers drifting toward churn before your dashboards do, and attach a real revenue value to saving each one, so win-back budgets go where the money is.
Some customers buy because you messaged them. Others were going to buy anyway. We separate the two so you spend on customers your campaigns are really changing, not the ones inflating the report.
Not everyone needs 20% off. We match each customer to the lightest offer that will convert them, so discounts protect only the customers who need them and margin stays on the rest.
Hashed customer IDs and transactions. No names, no phone numbers. Customer identities never leave your systems.
A model trained on your buyers, your catalog, your patterns. You get a ranked list of who to message, when, and with what offer.
Your next campaign runs against a randomized holdout. We report incremental revenue, win or lose, in plain dollars.
Every campaign runs against a randomized control group and lands in a live dashboard that tracks treatment vs control revenue, day by day. No more debating open rates. You see, in dollars, exactly how much extra revenue the campaign made over people you didn't message.
Our founders built and shipped an end-to-end targeting stack for one of the largest consumer bases in the world, from ranking every customer, to picking the best offer for each one, to a live decisioning system measured against control every single month.
Every customer scored on likelihood to buy, upgrade or churn, refreshed monthly, validated on unseen future behavior.
Best-offer selection across a large catalog, balancing conversion probability against margin.
A real-time decisioning system with an always-on control group measuring incremental revenue month after month.
Client identity and results confidential. Methodology walkthrough on request.
Most tools show you segments and stop. Marginal is accountable for one number: extra revenue versus a control group. Every engagement includes a randomized holdout, a pre-agreed success metric, and a written readout, win or lose.
The engine underneath is our own proprietary methodology, refined over years of shipping targeting systems for some of the largest consumer bases in the world. You get the outcome, not a science lecture.
A sample Customer Revenue Audit: value-ranked segments, a 30-day campaign calendar, and ready-to-send message copy.
One free audit. One measured campaign against a control group. A dashboard that tells you, in dollars, exactly how much extra revenue we made you.
No seat licenses. No dashboards you'll never open. Pricing aligned to measured lift.
Marginal Labs is a founder-led practice built around one thing: customer targeting that survives a control-group test. Our founders come from BCG, Google, IIT and ISB, and have spent 4+ years shipping targeting and personalization systems for some of the largest consumer bases in the world. Delivery is AI-native: senior judgment, leveraged by modern tooling instead of a bench of juniors learning on your data. Every engagement is delivered by the person whose name is on the methodology.
Keep them, they're the pipes. Marginal is the brain that decides who goes into them. Your tool sends to segments someone eyeballed; we give it a ranked list built from your actual purchase patterns.
You share hashed IDs and transaction history only, no names, emails, or phone numbers. Scored lists come back keyed to your IDs, so only you can map them to real people.
A randomized control group is agreed before the campaign. We compare treated vs held-out customers on revenue over 30 days and publish the readout, including if the lift is smaller than hoped.
Roughly 12+ months of order history and 50,000+ customers is where modeling reliably beats rules of thumb. Smaller bases still get value from the audit, we'll tell you honestly if you're too early.
A data export, a 30-minute call, and permission to run one targeted campaign through your existing tools. If it works, we talk pricing. If it doesn't, you pay nothing.
Founders with backgrounds spanning BCG, Google, IIT and ISB, and 4+ years building targeting and personalization systems for consumer bases of 100M+, now bringing the same methods to consumer brands.
See the lift first. Pay only if it works.
Built by founders ex-BCG · Google · IIT · ISB, with 4+ years shipping targeting & personalization systems for consumer bases of 100M+ customers