Published June 15
Poem interpretation, the process of unpacking a poem's deeper meaning, themes, and emotional impact by examining its structure, word choice, and literary devices. In school, it always felt like educated guessing with the confidence of expertise. The willow tree is grief. The willow tree is resilience. The willow tree is the poet's complicated relationship with her mother. Every interpretation is carefully argued. The poet, who possibly just liked willows, is not available for comment.
Now replace the poem with a campaign. Replace the students with MMM, attribution vendors and platform dashboards. Ask what the numbers mean and how the campaign performed…
To increase the difficulty level, the data about to be interpreted usually comes from a variety of sources. The platforms, Google, Meta, Amazon and the likes, selling the media, measuring the outcomes, and defining what counts as success. Basically delivering, pre-digested interpretations. The open web vendors, offering transparency, independent signals, and numerous dashboards. All these numbers, ready for a DIY interpretation.
Regardless of the data source, a recent CIMM study found that across every single measurement domain, advertisers described themselves as only "slightly confident" in their conclusions. Not confident, not even moderately confident. Slightly. After decades of investment in data infrastructure, identity graphs, MMM, attention metrics, and now AI. Slightly.
A lack of data volume certainly isn't the problem. The poem is longer than ever. There's more of it than anyone can possibly read. The problem is that no two readers agree on what it means, occasionally landing at completely conflicting conclusions, each of them with very good reasons for their interpretation.
The walled gardens at least offer consistency. One entity controls the media, the measurement, and the definition of success. You may not love what you see, but everyone in the room got the same number. Which is simultaneously a solution and a larger problem. When the platform selling the media is also the one telling you the media worked, the conflict of interest isn't a footnote. Consistent interpretation, yes. Disinterested interpretation, no. Or as Winston Churchill never said: "Do not trust any statistics you did not fake yourself."
The open web offers a different solution: transparency, independent signals and third-party validation. And it's real, as far as it goes. The challenge is that independence produces signals from so many sources, which makes interpretation harder rather than easier. Multiple methodologies, multiple answers to the same question, and no clear hierarchy for which should lead. The natural reaction from advertisers isn't we need more dashboards or another report. It's we need a single version of reality. Turns out transparency is the 'easy' part; agreement about what the transparent numbers mean is considerably harder.
Confidence itself holds where signals are blunt and boardroom-friendly: ROAS, CPA, sales lift. The metrics that don't require a methodology appendix to defend. Beyond that it starts to erode. Take attention. It replaced viewability because viewability wasn't enough, which is true. But viewability was a beautifully blunt instrument: 50% of pixels visible for one continuous second. The MRC drew a line, the industry stood behind it, and everyone moved on. Attention is more meaningful, and the IAB and MRC published guidelines for it. Sixty pages of them. But those guidelines cover how to measure attention, not what the numbers mean or which methodology to trust. They explicitly note that different approaches "may lead to different conclusions." More science, more rigour, same interpretive problem. Each tool is doing something real. The problem is that doing something real and being agreed upon are two different things.
What makes this more than a tooling problem is that the confidence gap is as much human and organizational as it is technical. Advertisers have learned to function within the fragmentation, which is not quite the same thing as solving it. Most have built workflows, developed internal hierarchies of trust, learned which numbers to lead with in which rooms. The problem isn't that they can't function within the system, it's that functioning within it and being confident about it are two different things.
Kelly Abcarian, CSO, Matter More Media, writing in The Current, frames it from the supply side: two campaigns with identical reach can produce entirely different outcomes depending on environment, audience strategy, brand economics, creative. The measurement exists but what the measurement means is a different conversation.
That's a conversation that still requires human judgment. Someone has to decide which signal leads, which methodology to trust, which number goes in the deck. Someone has to sit with the ambiguity and make a call anyway.
This someone is now joined by a something. With agentic AI entering the room, the industry is moving, quickly, toward systems that don't just interpret signals but act on them. Agents that reallocate budget, suppress audiences, swap creative mid-flight, without pausing to ask whether the inputs are coherent. The slightly confident data doesn't become more reliable when an agent is running on it. The interpretation problem doesn't disappear. The human who was at least aware of the uncertainty does.
The most honest thing the industry has said in years is: slightly confident. That may not sound like much, but in a room full of performed certainty, it is at least an admission of uncertainty. The Germans have a proverb for it: Selbsterkenntnis ist der erste Schritt zur Besserung (self-awareness is the first step to improvement). In measurement, that translates into clearer interpretation, better decisions, and ultimately more rational allocation of ad spend.
This op-ed represents the views and opinions of the author and not of The Current, a division of The Trade Desk, or The Trade Desk. The appearance of the op-ed on The Current does not constitute an endorsement by The Current or The Trade Desk.
