The Attribution Crisis is Actually a Coordination Problem

Everyone in ecommerce is having a collective panic attack about attribution. Meta’s telling you one number, Google’s telling you another, your Shopify dashboard has a third opinion, and your CFO is wondering why you can’t just “figure it out” like it’s a simple arithmetic problem.

The prevailing wisdom is that this is a data problem. iOS 14 broke everything. Cookie deprecation is making it worse. If we could just get better tracking, better pixels, better first-party data, we’d solve this.

But I think we’re diagnosing the wrong disease.

The attribution crisis isn’t a data problem. It’s a coordination problem. And if you understand why, you’ll understand something fundamental about how markets actually work. And also why most of economics is quietly, catastrophically wrong about digital advertising.

Hayek and Your Facebook Ads

Friedrich Hayek won the Nobel Prize in Economic Sciences in 1974 (NobelPrize.org) for pointing out something that should have been obvious but somehow wasn’t: markets work because they coordinate distributed knowledge, not because anyone has perfect information.

His classic essay “The Use of Knowledge in Society” (1945) is an essay everyone references and nobody reads. The core insight: the “economic problem” isn’t how to allocate resources when you have all the information. It’s how to coordinate action when knowledge is scattered across thousands of people who each know only a tiny fraction of the whole.

Prices aren’t just numbers. They’re signals that coordinate behavior across people who will never meet, never communicate, and often want completely opposite things.

Now think about ecommerce attribution.

Your customer saw an Instagram ad (knowledge your media buyer has). They Googled your brand (knowledge Google has). They read a Reddit thread (knowledge you’ll never have). They saw your billboard during their commute (knowledge that exists nowhere). They mentioned you to their friend (knowledge that also exists nowhere). They finally bought after getting a discount code in an email (knowledge your ESP has).

Which touchpoint “caused” the purchase?

This is like asking which specific price signal “caused” market equilibrium. The question itself reveals a fundamental misunderstanding of how distributed coordination systems actually work.

The Central Planning Fantasy

Here’s the uncomfortable part: most attribution models are just digital central planning.

Multi-touch attribution says: “If we could just gather ALL the data from ALL the touchpoints, we could correctly allocate credit and optimize our spending perfectly.”

This is the exact same logic that said: “If we could just gather ALL the information about supply and demand, we could plan an economy better than the market can.”

It failed spectacularly in 1989. It’s failing quietly right now in your marketing dashboard.

The reason is the same one Hayek identified: the knowledge required is too dispersed, too contextual, too temporally bound to ever be captured by a central system.

Your customer’s decision wasn’t caused by your Facebook ad OR your email OR your SEO. It was caused by a complex coordination of signals, contexts, timing, and incentives that emerged from all those touchpoints plus hundreds of others you’ll never see.

The “attribution problem” is just the digital marketing version of the “calculation problem” that killed central planning. We’re trying to centrally optimize something that only works because it’s decentralized.

Why This Actually Matters

“Fascinating theory, but I still need to know where to spend my money.” - Every CMO reading this, reasonably. Fair. But understanding this as a coordination problem instead of a data problem changes everything.

If it’s a data problem: You need better tracking. More pixels. First-party data. Customer data platforms. Attribution modeling software. Data warehouses. Engineers to maintain all of it.

You invest millions in infrastructure to capture more data.

And you end up with seventeen dashboards that all disagree with each other.

If it’s a coordination problem: You need better signals. Clearer feedback loops. Faster iteration. Teams that actually talk to each other. Systems that connect action to outcome without requiring a PhD to interpret.

You invest in organizational infrastructure.

And you end up with teams that can actually ship things and learn from what happens.

One approach tries to model reality perfectly. The other tries to coordinate effectively within reality’s constraints.

Only one of these actually works.

The Real Cost

This misdiagnosis is expensive in ways that don’t show up on a P&L.

I’ve watched companies spend millions on attribution tools, customer data platforms, data scientists, consulting engagements, and machine learning models. All in pursuit of the perfect attribution model.

They end up with more data than ever before and less clarity about what to do.

The data scientist builds a beautiful model. Marketing doesn’t trust it because it contradicts their last-click data. Finance doesn’t trust it because it’s “too complicated.” The CEO doesn’t trust it because the numbers changed when they ran it again last week.

Meanwhile, the companies that are winning have mostly given up on perfect attribution. They’ve focused on coordination: getting marketing, product, and operations aligned on clear signals, fast feedback, and lots of small experiments.

They don’t know exactly which touchpoint caused which sale. But they know what they shipped, when they shipped it, whether it worked, and what to try next.

That’s not perfect information. But it’s good enough to make decisions and learn fast.

Which is all markets ever needed anyway.

Coase and Your Slack Threads

Ronald Coase (NobelPrize.org) (another Nobel winner, this is turning into a theme) pointed out something else that matters here: firms exist because sometimes coordinating internally is cheaper than coordinating through market mechanisms.

But here’s the thing: most ecommerce companies are organized like internal markets, then act surprised when they can’t coordinate.

Your marketing team “buys” development time with tickets and priorities. Your product team “negotiates” roadmap with engineering through sprints and standups. Your merchandising team operates in a completely different economy with different currency (margin dollars vs. marketing dollars vs. engineering hours).

All these internal markets have massive transaction costs. Status meetings. Slack threads (so many Slack threads). Email chains. Conflicting priorities. The endless back-and-forth of “did this ship yet?” and “what’s the status?” and “can someone update the doc?”

The attribution crisis just makes these coordination costs visible.

When you can’t track attribution across channels, you also can’t coordinate across teams. The same information problems that break your marketing dashboard also break your ability to actually work together.

This is why companies spend millions on attribution tools and still can’t answer “did our campaign launch on time?”

Wrong problem. Right problem costs less but requires actual organizational change, which is harder than buying software.

What Actually Works (And Why It’s Annoying)

The companies I’ve seen navigate this well have done something that sounds trivial but apparently isn’t: they’ve stopped trying to attribute everything and started coordinating everything.

They don’t ask “which channel drove this sale?”

They ask:

• Did our campaign launch when we said it would?

• Do all the teams know what’s live?

• Can we connect this specific change to this specific outcome?

• What did we learn and how fast can we iterate?

• 
  

They don’t build more complex attribution models. They build simpler coordination systems.

They don’t try to centrally optimize budget allocation with perfect information. They create feedback loops that let teams self-optimize with good enough information.

It’s Hayek’s solution to the knowledge problem, applied to marketing: distributed decision-making coordinated by clear signals, not central planning coordinated by perfect data. (My tailor doesn’t have a CRM. He has a notebook. He knows exactly what’s in production, who needs a fitting, and when everything will be done. Your $50M ecommerce company has seventeen systems and nobody knows what shipped yesterday. Think about that.)

The Uncomfortable Implication

If this is right (and I think it is, but I’m open to being wrong - I collect data on being wrong, it’s surprisingly useful), then most of what we think we know about marketing optimization is wrong.

Not “needs some adjustments” wrong. Fundamentally, structurally, philosophically wrong.

The entire discipline of marketing mix modeling, multi-touch attribution, media optimization - it’s all built on the assumption that we can measure, model, and optimize based on complete information.

But if Hayek was right (and the Nobel committee thought he was), this is impossible in principle, not just hard in practice.

You can’t build a perfect attribution model for the same reason you can’t centrally plan an economy: the knowledge you need is too distributed, too contextual, too dependent on timing and circumstances and things you’ll never observe to ever be fully captured.

The best you can do is create systems that coordinate effectively despite incomplete information.

Which is, inconveniently, much harder than buying more software.

Why I Think About This

I’m building something right now that tries to solve this. Not with better attribution (we’ve established that’s impossible), but with better coordination. Getting teams aligned, making work visible, connecting action to outcome in ways that actually help people make decisions.

But honestly, the more I work on it, the more I realize the problem isn’t technical. It’s philosophical.

We keep trying to solve coordination problems with data problems. We keep being surprised when it doesn’t work.

The brands that win won’t have the best attribution models. They’ll coordinate fastest, iterate fastest, learn fastest.

Just like markets don’t work because anyone has perfect information - they work because prices coordinate distributed knowledge - ecommerce won’t work because you have perfect attribution.

It’ll work because your teams can coordinate effectively despite imperfect information.

The Thing That Keeps Me Up

Here’s what bothers me: AI is going to make this so much worse before it makes it better.

More data. More models. More “insights.” More confident predictions based on correlations we don’t understand.

But if the fundamental problem is coordination, not information, then we’re just building more sophisticated ways to be wrong.

The solution isn’t better tracking. It’s better teamwork. The crisis isn’t technical. It’s organizational. The problem isn’t that we can’t measure everything. It’s that we’re trying to.

And unlike the calculation problem that doomed central planning, this one is actually solvable. You just have to stop trying to centrally plan your marketing and start coordinating it instead.

Which is less exciting than buying new software, but also more likely to work.

###

Yes, I just cited three Nobel Prize winners in an article about Facebook ads. No, I’m not sorry. If we’re going to be wrong about something, we might as well understand why the smart people were right.