David Yaffe and the work that should disappear

David Yaffe began a December 2024 conversation about data infrastructure with a small explanation for his calendar: a new baby. Here was the co-founder of a company devoted to getting information where it belonged on time, explaining the difficulties of getting himself into a meeting. There are scheduling problems even a streaming platform cannot solve.

He was more sanguine about the years spent building Estuary. Supporting both historical batches of information and continuously arriving changes had required a substantial investment before customers began arriving in earnest. He described roughly four years devoted to the underlying infrastructure. The choice had made the engineering harder. It also meant a customer could bring more than one kind of problem.

“But we chose a hard thing and it takes a long time to get it right.”

David Yaffe, December 2024

That patience sits beside a career with several changes of setting: Boeing, advertising technology, Google, Arbor, LiveRamp, and now Estuary. The jobs involve different machinery. A recurring concern is the amount of effort needed to make that machinery useful to somebody else.

The Boeing route to a bidding screen

Yaffe’s training was in mechanical engineering: a bachelor’s degree from Binghamton University, completed in 2005, followed by a master’s at Columbia in 2006. He worked as an aeronautical engineer at Boeing before entering the data business. Aircraft precede advertising in this particular résumé.

By 2008, he had joined Invite Media. Its business brought him into a world where information had to arrive quickly enough to affect an advertising decision. He became director of platform, then led product for Invite Media and DoubleClick Bid Manager after Google acquired Invite. This was product management conducted close to demanding infrastructure, where a useful answer delivered late could lose its purpose.

A former colleague, Scott Becker, described a combination of technical, mathematical, commercial, and interpersonal abilities. Becker credited Yaffe with understanding the industry, developing profitable features, and helping customers use the platform. He also called him “genuine, thoughtful, and funny.” The last quality is welcome in any profession involving both software and meetings.

The company inside the company

In 2014, Yaffe co-founded Arbor and became its CEO. The proposition was commercial: publishers and app developers held information that could be useful to marketers. Arbor would help them understand and monetize those assets. Yaffe and his fellow founders had come from Google, with a close view of how valuable a direct relationship with users could become.

LiveRamp acquired Arbor in 2016. Yaffe went on to serve as a general manager and COO at LiveRamp. Moving from founder to an operating role inside an acquiring company brought a broader set of responsibilities, but the recurring difficulty beneath the business remained familiar: information lived in separate systems, and keeping those systems in agreement consumed people and time.

At Arbor and later LiveRamp, the team built Gazette, a streaming broker that stored information as ordinary files in cloud storage. That design mattered because existing records could be read without putting the same pressure on the live systems serving current work. Gazette was eventually open-sourced. A piece of internal infrastructure had acquired a possible life outside its original employer.

Yaffe and Johnny Graettinger founded Estuary in 2019. Their experience supplied the starting point, though an effective internal tool was only part of a sellable service. Customers needed connections, transformations, and a manageable way to operate the whole thing. A capable broker could still leave a considerable amount of assembly on the customer’s desk.

Their ambition was to handle information arriving on different schedules without making a business maintain a separate world for each. A weekly report can wait. An application responding to a customer may have much less time. Estuary’s founders wanted both workloads to share infrastructure, while preserving the flexibility engineers needed to build around it.

Two weeks, with a long runway

Yaffe’s account of raising Arbor’s first round has an appealing number attached: about two weeks. In a 2020 interview, he explained the circumstances that made it possible. There had already been roughly a hundred conversations with marketers and publishers, a working product, and initial customers receiving value.

There were also relationships. Entrepreneurs he knew could introduce him to investors, and those introductions were organized to arrive together. Some of the people making introductions were investing themselves. The compressed fundraising window came after the work that made the company understandable and the introductions credible.

His stated preference was to raise when he understood what he would do with the money and felt confident in the idea. Read that alongside Estuary’s lengthy infrastructure development and the apparent contradiction becomes less mysterious. Different parts of company building move at different speeds. Investors can make a decision in a fortnight. Reliable software declines to respect the calendar.

The data executive who distrusts the scoreboard

Yaffe’s writing takes an interesting turn when it reaches management. A person selling better access to data might be expected to prescribe another dashboard. Instead, his 2023 essay on objectives and key results worries about what happens when measurements become the main instrument of leadership.

His argument is about incentives. Attach personal advancement to a selected number and employees have a reason to improve the number, including when it stops representing the organization’s actual purpose. A target can become a defensible performance review while the underlying business problem survives quite comfortably.

“Setting goals that solely track top-level metrics is a sign of bad leadership.”

David Yaffe, on managing with OKRs

He argues for a clear company purpose that helps employees make decisions, measurements that reveal organizational performance, and accountability informed by several perspectives. A manager announcing a larger revenue target has supplied a destination. People still need enough understanding to decide how to get there and which compromises make sense.

This is a useful complication in his public persona. Yaffe wants information to circulate more freely, but he also gives attention to the incentives surrounding it. Faster delivery cannot repair a poorly chosen objective. A company can make a remarkably efficient pipeline for numbers that encourage everyone to do the wrong thing.

Let somebody else ask the question

His argument for data democracy begins with ordinary frustrations. A marketer needs information for a customer story. An analyst has produced a useful report that colleagues cannot access. The relevant material exists somewhere inside the company, but getting it to the person who can use it becomes a small administrative expedition.

Yaffe proposes searchable catalogs, clear definitions of incoming information, integration between systems, and tools that let people derive further useful datasets. Access must work within governance and privacy requirements. The point is to give nontechnical colleagues room to apply their own knowledge, without sending every new question into an engineering queue.

There is an organizational idea inside that technical proposal. The employee who understands a local problem should be able to investigate it. Procurement, sales, and marketing each bring knowledge an infrastructure team cannot supply on command. Making their information discoverable gives that knowledge something to work with.

A price on attention

Yaffe has also written about why enterprises should concentrate on their own businesses rather than spend their days inventing internal software. His examples include banks and airlines, organizations with demanding work already on the agenda. Buying a service becomes more attractive when the service actually removes an operating burden.

At Estuary, that promise has required detailed, sometimes repetitive engineering. In his 2024 discussion, he described investing in a common framework for connectors, the software that talks to different systems. Building to shared specifications takes effort at the outset. His rationale was that it reduces the continuing work of maintaining unrelated implementations.

The commercial decisions follow the same concern for what happens after adoption. In a 2026 BUILDERS conversation, he described pricing that left customers room to expand, including a price reduction intended to encourage larger uses. He also gave a wonderfully mundane warning sign for enterprise sales: a shared Slack channel that attracts only a few messages and then falls silent. Enthusiasm has to survive contact with somebody’s working week.

The October 2025 financing gave the company resources to expand engineering, product, and its commercial teams. Yaffe and Graettinger framed their direction around letting customers choose the timing and deployment arrangements their work required. The same integration problem now served analytics, operational software, and AI.

An agent’s first day

By August 2026, Yaffe was discussing how agents should use business systems. His description was practical: documentation clear enough for a newcomer, understandable data, and permissions limited to a particular task for a limited time. Estuary had put effort into documented command-line tools and access controls. An agent needs a way to work out what it is looking at before it can do useful work.

The topic sounds newer than the concerns beneath it. Information needs definitions. Systems need connections. Access needs boundaries. Somebody has to make those details coherent. Yaffe’s career has carried him from aircraft engineering into advertising decisions, acquisitions, and company operations, giving him several views of what happens when machinery meets a business.

His ambition at Estuary leaves customers with a pleasantly ordinary possibility: the data arrives, the systems agree, and the team gets on with its day. For a founder who has spent years building the connections, that quiet outcome is a considerable amount of work.