August 2020. Biden was up big in our polls. Up big in everyone's polls.

I got a call from someone on the ground who said they just didn't think it was right where they were. I didn't either. But we had called every state correctly in 2016 with the same methodology, so I wasn't ready to throw out the method. I just knew something was off.

What we eventually figured out was that we had the right amount of data — we just didn't have the right kind.

Here's a simple example. We had the correct quota of rural respondents. But when we looked closer, too many of them were coming from rural areas that were more educated, higher income, higher on healthcare indices than average.

The right number of rural people. The wrong rural people. Not actually representative of rural America.

And when a variable like that correlates with how people vote, you have a problem traditional polling methods won't catch.

So we went and got more data.

We pulled in outside sources — county-level databases, economic indices, demographic layers that go well beyond the standard age, race, and gender buckets. We looked for what was actually predictive of the outcome. Then we reweighted.

It was slow. But it worked.

Our late October polls ended up somewhere between 2nd and 5th most accurate, depending on who was doing the ranking.

But it wasn't something you could do at scale. That's really why this company exists.

The core insight wasn't really about polling.

It was about what happens when you bring more data together — the right data, the predictive data. It doesn't just improve one answer. It expands the entire surface area available for intelligence and insight.

Every new source you add doesn't only tell you something new — it illuminates relationships you couldn't see before, surfaces variables you didn't know to look for, and corrects assumptions that had been quietly baked in.

We think about this a lot in terms of what we call lift.

In traditional research, you start with a question and go find data to answer it. Lift flips that. When you've assembled enough of the right data sources — demographic, behavioral, geographic, economic, attitudinal — the data starts telling you things you didn't think to ask.

Patterns emerge. Correlations appear. The intelligence compounds.

That matters more now than it ever has.

We are entering an era where reasoning is no longer constrained by the human brain. AI can reason across data sets, at scale, continuously. Which means the limiting factor isn't horsepower anymore. It's surface area.

The enterprise with the most, best reasoning surface area wins. Not because they're smarter, but because they're working with a more complete picture of reality.

Wick does a lot of things. But it all starts from the same place: go get more data, find what's predictive, and let the intelligence follow.

Six years later, we think we're close to solving the original polling problem too. That'll just be the first proof point.

David Burrell

David Burrell is co-founder of Wick and writes about opinion research, audience intelligence, AI readiness, and the proprietary knowledge layer behind better enterprise AI.