Somewhere in every mid-sized company there is a tracking request for which nobody can account. Marketing filed it, assumed engineering shipped it, and built a quarter of reporting on the assumption. Engineering saw a vague ticket with no business context, parked it behind a release, and assumed marketing had moved on. Nobody lied, nobody dropped the ball in any way a postmortem would catch, and yet the number does not exist. Josh Silverbauer, Head of Analytics and Chief Revenue Officer at Intrepid Digital, has spent more than a decade in those rooms, occasionally bringing his love of music into the world of analytics. His diagnosis is blunt: the failure is linguistic, not technical. “That gap is where most measurement work goes to die,” he says. The cost lands downstream, in budgets allocated against data that was never captured and decisions defended with numbers that do not exist.
Lead With The Decision, Not The Tag
The default request from a marketing team is a specification: fire an event here, name it this, pass these parameters. It feels professional. It is also the fastest way to get a build that technically satisfies the brief and answers nothing. Silverbauer’s point is that an event request stripped of its purpose arrives on an engineer’s desk as busywork. “When marketers ask for an event, engineers hear a task with no context,” he says. Without the why, there is nothing for an engineer to push back on, improve, or flag as impossible given how the site is built.
Flip the order and the dynamic changes. Lead with the decision the data is meant to inform, what will be done differently once the number exists, and the conversation stops being a handoff and starts being a design problem two people are solving. “When marketers explain what they will do differently once they see that number, engineers can suggest a clean way to capture it,” Silverbauer says. “Start with the business question, and implementation usually improves on the way in.” That last clause carries more weight than it seems. Engineers know the codebase. Given the objective, they will frequently propose a cleaner, more durable method than the one requested by marketing. Specifying the tag instead of the decision forecloses that entirely, and the organization pays for the privilege of being prescriptive.
Put The Constraints On The Table Early
The second failure is sequencing. Most teams discover their constraints after the brief is written, the timeline is promised, and someone has already presented the forthcoming dashboard to a stakeholder. Then the data layer turns out not to support the attribute. Or the consent configuration blocks the tag for a meaningful slice of traffic. Or the script costs enough page performance that it will not survive review. Or the release cycle puts the whole thing six weeks out. “Data layers, consent, page performance, and release cycles all shape what is realistic,” Silverbauer says. None of those are edge cases. They are the standing conditions of any modern measurement stack, and they are invisible to anyone who does not work inside it.
The remedy is cheap, which is what makes the usual practice so hard to defend. Silverbauer frames it as a trade with a clear exchange rate: “Marketers who understand those constraints write better briefs. Engineers who understand the revenue and stakes prioritize it differently. 10 minutes of shared context save months of rework.” Note that the obligation runs both ways. Marketers need the technical boundaries. Engineers need the commercial stakes, because a ticket attached to a revenue consequence gets ranked differently in a sprint than a ticket that reads like a favor. Teams that skip this step are not saving time. They are deferring a conversation that gets significantly more expensive the later it happens, usually surfacing as a rebuild rather than a revision.
Present Findings As A Story, Not A Dataset
Measurement work that survives implementation can still fail at the final step, and this is the one most analytics teams underestimate. A clean dashboard delivered without interpretation is a bill, not an asset. Silverbauer’s rule is to show the friction and land on a decision you want made. The structure matters: the data points to a problem, the problem has a cost, and there is a specific action on the table. Absent that shape, the audience nods, the deck gets filed, and the behavior stays exactly as it was.
His warning is the kind that should make analytics leaders uncomfortable, because it applies retroactively to a great deal of work already shipped. “If no one understands what the tracking had impact on, it will not change anything,” he says. That reframes the whole value proposition of a measurement program. The deliverable is not the tag, the event, or the dashboard. It is a changed decision. Everything upstream exists to produce that, and work that does not reach it has consumed engineering hours, marketing budget, and organizational patience for nothing. The implication for how analytics teams justify themselves is uncomfortable but clarifying. Count decisions changed, not events tracked.
What makes Silverbauer’s argument practical rather than aspirational is his insistence that this is learnable. Fluency across the developer and marketer divide is not a personality type that a company either hires or does without. It is a skill, and the entry cost is low. Start by learning one thing about how the other team works. One constraint, one priority, one reason the other side says no. In organizations where that curiosity is normal, the black hole closes and tracking requests start producing answers rather than mysteries.
Connect with Josh Silverbauer on LinkedIn, where he breaks this down every week, sometimes with a song.
