About SegmentStream
The company, the principles, and how we work.
The work
An independent measurement engine, built for the brands paying for the ads.
Marketing measurement is closer to critical thinking than to marketing. It is the architecture of tracking, the discipline of data consolidation, and the math that turns ad spend into decisions a finance team can defend. It is also the processes and decision frameworks embedded into a marketing team that turn those decisions into moved budget.
We started SegmentStream in 2018. The category had drifted into a belief system, where teams trusted attribution models and MMM frameworks the way people trust horoscopes. Confident. Detailed. Mostly fiction. The brands paying for the ads deserved better.
Eight years on, the work has grown into the SegmentStream Measurement Engine: cross-channel attribution, identity stitching, incrementality experiments, marginal analytics, automated budget allocation, and AI agents that turn raw marketing data into answers a finance team can defend.
We built it for the brands paying for the ads. Independent of the platforms we measure, grounded in evidence, published in full, wired to act, delivered as a system, taught from first principles. That is the only side we work for.
Our principles
Independent of the platforms we measure.
We work for the advertiser. That is the only side we work for.
We do not co-market with ad platforms. No joint case studies. No subsidised client placements. No "strategic partnerships." No referral kickbacks. When you read a SegmentStream report, no ad platform has been in the room shaping the conclusion.
The same way an auditor's only client is the company whose books they audit, our only client is the advertiser. Our incentive is to count honestly, because the advertiser is the one paying us to count.
Evidence over modeling.
Deterministic before modeled. Always labeled.
A deterministic measurement is a fact. A model is a hypothesis. Until evidence backs it, we label it that way in the report.
Our default is deterministic data the customer owns. We label every modeled number as modeled. We run incrementality experiments when the answer matters. We report the confidence interval, not just the headline lift. When the data does not support a conclusion, we say so.
Platform-reported conversions get the same treatment. We collect them, but we never let them lead.
No black boxes.
Every model published. Every number auditable.
Every model we ship is one the customer's finance team can audit. We publish how attribution credit is assigned. We publish how identity stitching works. We publish how marginal analytics evaluates each next dollar of spend. We publish the geo-holdout methodology, the priors, the confidence intervals, and the parts where the method has known limits.
The SegmentStream Measurement Engine ships as nine open whitepapers, each one a full explanation of one method. If you cannot find the math behind a number we report, that is a bug. Flag it and we will publish it.
A measurement number you cannot defend is a measurement number you cannot use.
Built for action.
Measurement is only useful when it changes a decision.
A measurement report that ends in a slide deck and stays with the analyst does not change a business. Numbers get pulled, decks get written, the marketing lead nods, and the budget moves the way it was already going to move.
SegmentStream is wired to act on what it measures. Marginal analytics says where the next dollar should go. Automated budget allocation moves the budget against that signal, daily. Incrementality experiments validate that the move worked. The measurement, the recommendation, and the execution sit in one system.
Reports still exist, but they read as exhaust — what falls out of decisions made and validated, not what the team has to read to figure out what to do.
A system, not a tool.
Technology, process, expertise. All three, together.
Marketing measurement is technology, process, and expertise working as one system. The technology runs the math and the data infrastructure that feeds it. The process is how the company organizes its decisions around what the system surfaces. The expertise is the methodology knowledge that connects them. Skip any one and the system breaks at that layer.
The "just add AI" version of this skips the harder work. A chatbot on top of a dashboard does not change how a team answers questions or moves a budget. The operating system has to change — the process, the methodology, the people running it.
SegmentStream ships as a full system. The measurement engine handles the math. We help wire the measurement into how decisions actually get made each week. Our experts work alongside the team running the budget. The value comes from all three working together.
Education over selling.
Teach the discipline before selling the tool.
Marketing measurement is a discipline. It becomes useful when the buyer understands the math. So we teach the discipline in long form.
The nine measurement engine whitepapers explain each method from first principles. The 9.5-hour course on modern marketing measurement teaches the discipline end-to-end. Both are open. Both are the work a buyer needs to do before the platform becomes useful.
When a client buys SegmentStream after working through the methodology, they buy for the right reasons and they stay. So we write the textbook first.