From dashboards to decisions: rethinking publisher revenue with Burt’s Shay Brog
BY ROB BEELER + SHAY BROG, CEO OF BURT INTELLIGENCE
Publishers are under relentless revenue pressure, and most respond by squeezing more out of each channel. Direct, programmatic, commerce and subscriptions all have their own tools, reports and KPIs. Yet few publishers have the connected data and decision-making infrastructure to see how those channels affect one another, what’s likely to happen next and what to do about it. For ad ops and yield teams, that turns every tradeoff into guesswork.
In this Q&A, Rob Beeler joins Shay Brog, CEO of Burt Intelligence, to discuss why decisions get harder when teams work from separate views, why retrospective reporting sometimes falls short, and what publishers need to see across the business. Shay also explains what needs to be true before AI takes a more active role in revenue decisions, where direct revenue can benefit from programmatic-style discipline, and what he’d tell a publisher rebuilding for 2027.
Rob: Publishers are under enormous pressure to grow revenue while doing more with fewer resources. What does that pressure expose about the way most publishing businesses currently make revenue decisions?
Shay: I think it exposes a problem that was easier to live with when there was more margin for error. Publishers have a lot of data, but very little of it is organized in support of decision-making.
Most publishing businesses grew their revenue operations channel by channel. Direct has its systems and workflows. Programmatic has another set. Commerce. Audience. Each team can be quite sophisticated within its own domain, while the business as a whole still struggles to answer fairly basic questions. If we change pricing here, what happens to sell-through somewhere else? Are we actually short on inventory, or are we allocating it poorly?
When resources are plentiful, you can plug those gaps with people, spreadsheets and meetings. Under pressure, those gaps become much harder to paper over, often with teams looking at different versions of the truth. And that’s when you find out how data driven your organization really is. Everyone thinks they’re data driven until the data tells them to do something differently.
Rob: How did publishers get to a point where they can have sophisticated reporting across individual parts of the business and still struggle to see how those parts are affecting one another?
Shay: Because the technology stack reflects how the industry developed. Nobody designed this from Frankenstein scratch. It grew organically, for example, think how header bidding evolved as a response to some of Google’s product choices.
There was also a lot of venture capital pouring into adtech 10 to 20 years ago. Everyone was trying to get a bite of the apple, usually by solving one specific piece of an increasingly complex value chain, and in many cases they did. But the cumulative effect was an incredibly fragmented technology and data environment.
The problem is that revenue doesn’t respect those boundaries. A direct campaign affects inventory available to programmatic. Pricing affects sell-through. Optimizing any one of those things in isolation can have unintended consequences somewhere else.
So we created a world where two departments can look at their respective dashboards, both technically correct, and reach completely different conclusions about what the company should do. The industry got very good at optimizing the pieces. I believe the next challenge is optimizing the whole.
Rob: Publishers have spent years investing in retrospective reporting and calling it “business intelligence.” But does knowing what happened give you what you need to make good revenue decisions?
Shay: Not by itself. At Burt, we tend to think about this as a progression from descriptive to predictive to prescriptive. Descriptive analytics tells you what happened yesterday. That is where most traditional business intelligence lives, and it gives you a rearview mirror representation of the business as it was.
A few years ago, we acquired Inventale and introduced Burt Forecasting, which took us further into predictive analytics. How much inventory will I have? Which campaigns are likely to run into trouble? But you can have a remarkably accurate forecast and still make a bad decision.
The next step is how you move from what is to what ought. Given what we know, and what we believe is likely to happen, what should we do? Now you’re introducing objectives, tradeoffs, constraints and judgment. Should we sell this inventory now, or preserve it for something more valuable later?
Recent advances in AI are making it possible to start tackling that third, prescriptive layer in ways that weren’t practical before. That, to me, is where Decision Intelligence at Burt gets really interesting, as it helps people reason about what to do next.
Rob: When revenue is under pressure (so, all the time), which decisions have you seen become hardest to make when teams are working from separate views of the business?
Shay: The hardest decisions tend to be the ones where the costs and benefits show up in different parts of the business. Imagine the commerce team discovers that using recirculation widgets to push more traffic toward e-commerce content produces meaningful incremental affiliate revenue. Looked at through that team’s data, it’s an obviously good decision.
But that traffic came from somewhere. Maybe the commerce pages carry less advertising, or none at all. Direct campaigns may start falling behind on delivery. Is selling another 50 mattresses worth the make-good you will now owe a tier-one advertiser?
The commerce team hasn’t done anything particularly irrational. They’re optimizing for the KPIs they’re measured on. The problem is that the cost of that optimization is an externality; it shows up somewhere else, in someone else’s data.
Every team is ostensibly managing different things, but underneath it all they’re competing for the same finite resources: audience attention, inventory and time. You can make a completely rational decision within one part of the business that is irrational for the business as a whole.
Rob: So what do publishers need to see across the business to understand how each channel performed and how those channels are influencing one another?
Shay: They don’t need a bigger dashboard with every KPI from every part of the company crammed onto it. That might give you more visibility, but it doesn’t necessarily give you more understanding. What publishers need first is a common model of the business underneath those different views, because ultimately those channels are all making choices about the same consumer. That means connecting traffic, inventory, demand, pricing and revenue, with consistent definitions of advertisers, products and audiences across systems.
RPS (Revenue per Session) is one interesting way to think about this, if you can get it to mean all the revenue associated with that session.
A paywall is, in one sense, a kind of house ad. Economically, it’s competing with direct campaigns, programmatic ads and recirculation widgets for the same finite resource, which is the consumer’s attention.
And even RPS is only a proxy, because then you have to introduce time. Sometimes the economically rational decision is to sacrifice revenue in this session because it increases the LTV (Lifetime Value) of the relationship.
So understanding how the channels influence one another means getting beyond channel revenue and toward the incremental value, opportunity cost and time horizon of each choice you make about the consumer.
Rob: When you work with publishers that have started connecting those pieces more effectively, what changes first in the way their teams operate or make decisions?
Shay: The first thing that changes is actually pretty mundane: it becomes much cheaper to ask a question.
In a fragmented environment, there’s a transaction cost associated with curiosity. Answering a question means pulling data from three systems, reconciling definitions and building a spreadsheet full of VLOOKUPs from hell. So people learn, quite reasonably, not to ask very many questions. Once you connect and govern that data, the cost of asking the second and third question starts approaching zero.
Recently, a sales leader at one of our customers wanted to know whether a custom creative from a third-party vendor was generating enough revenue to justify its development and serving costs. Instead of asking someone to build an analysis, they could just ask the question. It’s not an earth-shattering question. That’s sort of the point.
But the more important change is cultural. When the cost of asking questions is low, people become more empirical. They stop treating data as something precious they periodically receive and start treating it as something they can interrogate.
Rob: We’ve all heard the pitch, right? AI will make your team “faster.” Of course, that means bad data and assumptions will scale right along with everything else. Before a publisher lets AI run the show with its revenue decisions, what needs to be true about the data it’s working from?
Shay: I wrote recently that a lot of the friction we’ve historically had around data was accidentally doing some governance work for us. If answering a question required knowing which system to log into and having the right permissions to the right datasets, there were natural barriers between the data and the person trying to use it. AI is very good at removing that friction, but you don’t get to remove the governance with it. You have to replace accidental governance with intentional governance.
An AI system needs context. It needs to understand that two differently named advertisers are actually the same entity, and which data is authoritative when two systems disagree. Then there’s authority and auditability. A salesperson asking a question and an autonomous agent changing campaign settings shouldn’t necessarily inherit the same permissions. Bad information can already lead to expensive decisions. An autonomous agent can act on it before anyone even notices.
So before I’d let AI run the show, I’d urge publishers to ensure they have governed and normalized data, consistent definitions, business context, provenance, permissions and very explicit boundaries around autonomy.
Rob: Programmatic gets a lot of the industry’s attention when we talk about data and automation. Where have you seen publishers make the biggest gains by bringing that same discipline to direct revenue?
Shay: “Programmatic” really just means “using software.” One of my favorite examples actually runs in the opposite direction. We spent years building financial reconciliation and discrepancy reporting for direct advertising. Then one of our more sophisticated programmatic customers essentially asked why we didn’t do this for them too.
In some respects, direct is the more interesting computational problem because the inventory is reserved. With programmatic guaranteed or direct, you’re making a commitment today against inventory that may not exist for weeks or months.
What’s always struck me as strange is that we’ll use sophisticated algorithms to decide what to do with a single programmatic impression worth a fraction of a cent, while somebody can get an RFP for a $500,000 direct campaign and still make consequential decisions using spreadsheets, email and institutional memory.
Forecasting is an obvious place where we’ve seen publishers make gains. Sellers can explore reliable forecasts themselves, so revops teams can focus on the difficult exceptions, pricing strategy and managing yield. Campaign delivery is another, as spotting problems early gives you optionality. Then there’s everything after the sale, from reconciliation and billing to make-goods and proof of delivery.
Rob: If a publisher were rebuilding its revenue operations strategy from scratch for 2027, what would you tell it to stop doing, start doing and never compromise on?
Shay: Stop organizing the business around the systems that happen to produce the data. Your business doesn’t naturally divide itself into GAM data, SSP data, CRM data, commerce data and subscription data. Those are artifacts of how the technology stack developed. Build a common model of the business and a commercial graph that connects those things to one another.
Start designing around decisions rather than reports. Before you build another dashboard, ask what somebody is supposed to do differently after looking at it. That gets you much closer to Decision Intelligence than starting with a list of KPIs.
And never compromise on your ability to use and govern your own data. I think this will become much more consequential in an AI world. A publisher’s intelligence about its own business shouldn’t be trapped inside the individual systems where the data originated.
Ready to turn revenue data into decisions?
Explore how Burt can help your teams connect their data and turn insight into faster, better revenue decisions by booking a demo today.
👉Want to stay in the loop with what’s happening in the Beeler.Tech community? Subscribe to our newsletter. If you’re interested in attending or sponsoring a future event, you can explore our upcoming events here.