Context, Learning Curves, and the Real Economics of Martech

Martech as learning infrastructure

Scott Brinker has recently been writing about what he calls “context-as-a-service.” At first glance, the phrase might sound like another entry in the long list of technology abstractions that appear every few years. However, the more I thought about it, the more it struck me that what Brinker is describing can be interpreted in economic terms.

The real economics of martech is the economics of organizational learning.

Companies compete in part by moving up the learning curve faster than their competitors. Kenneth Arrow’s classic work on learning by doing argued that organizations improve through accumulated experience solving real problems over time. Friedrich Hayek made a complementary argument: economic advantage often comes from the ability to use dispersed knowledge effectively. Firms that learn faster and coordinate what they know more effectively reduce costs, make better decisions, and become harder to compete against.

Marketing technology, data systems, and analytics platforms can therefore be interpreted as learning infrastructure. They are the mechanisms through which firms observe customers, interpret signals, and adjust actions—designing customer journeys, triggering workflows, running A/B tests, personalizing experiences, and selecting the right channel at the right moment. The value of these systems does not lie primarily in their features, but in their ability to help an organization learn faster than the market around it.

Context is not knowledge

Seen from this perspective, Brinker’s argument about context becomes easier to understand.

Context is simply the relevant state of the world at a particular moment: who the customer is, what they have done recently, what they might need next, what constraints apply, and what signals suggest about intent. Software systems increasingly exist to capture and organize that context so that decisions can be made in real time.

Knowledge, however, is not the same thing as context. Context is raw situational data. Knowledge is what happens when an organization repeatedly interprets that context and turns it into reliable action. Context is therefore an input to learning, not learning itself. Over time, organizations that convert context into action more effectively accumulate experience faster than their competitors. This is the logic behind the experience curve popularized by the Boston Consulting Group: firms that learn faster move down the learning curve more quickly, often strengthening their competitive position in the process.

The problem is fragmentation

This distinction matters because many organizations assume that accumulating more data or more tools automatically improves performance. Sometimes it does. Additional tools can add needed capability, specialization, or scale. But beyond a certain point—often relative to team size, coordination capacity, and the organization’s ability to integrate systems—tool proliferation can dilute learning rather than strengthen it.

When context is distributed across many systems, the organization’s learning becomes distributed as well. Different tools contain partial views of the customer. Different teams operate from slightly different definitions of reality, creating friction in how they coordinate and prioritize work. Disconnected applications introduce handoffs, reconciliation work, and delays as signals move between platforms. Under those conditions, organizations struggle to convert raw context into coherent knowledge.

Learning compounds

Learning compounds only when signals accumulate within systems that can interpret them consistently and act on them repeatedly.

If context is fragmented across dozens of tools, the learning of the firm does not compound. It dissipates.

In this sense, the value of martech rationalization is often misunderstood. The goal is not aesthetic simplicity or vendor consolidation for its own sake. The real objective should be to reduce coordination costs and signal latency so that context can be interpreted and acted upon more quickly. When that happens, learning begins to compound inside the firm instead of being slowed, stalled, or allowing competitors to pull further ahead because fragmentation and coordination friction prevent the firm from learning fast enough.

New solutions, upgrades, and the build-versus-buy question

Organizations can approach this problem in more than one way. They may implement new solutions designed to consolidate context, or they may upgrade and better integrate the systems they already have. Either path can involve building internally, buying from vendors, or some combination of the two.

Buying software, in this sense, is not merely purchasing functionality. It is purchasing access to accumulated knowledge about how certain problems are typically solved. Vendors embed workflows, assumptions, models, and operational patterns derived from many customers. Firms that adopt these tools can therefore move further up the learning curve more quickly than if they started from scratch.

However, this advantage is not permanent. Vendors themselves must continue learning. When a platform stops evolving with the market, its encoded knowledge becomes stale. Organizations then migrate to tools that better reflect current practices and signals. In that sense, switching technologies is often less about features than about aligning with where the industry’s learning frontier currently sits.

AI changes the economics

Artificial intelligence changes the economics of this process.

AI dramatically reduces the cost and time required to build new software capabilities. Features that once took years to develop can now appear in weeks, whether they are created by vendors or assembled internally by firms themselves. As a result, the number of tools will likely continue to expand, and businesses will remain tempted to respond to new opportunities or new complexity by adding yet another product to the stack.

That instinct is understandable. New tools can add real capability. They can solve immediate problems, enable new forms of automation, or make sophisticated practices more accessible than they once were. But more tools, or even newer tools, will not be sufficient if the result is simply more fragmentation. Unless those tools help consolidate context, reduce friction, and strengthen the firm’s ability to capture and reuse what it learns, they may add software without materially improving knowledge.

This is why the firms that benefit most from AI will not necessarily be those that deploy the most tools. They will be those that use AI, new systems, and better integration to organize context in ways that allow learning to accumulate and compound.

The economic conclusion

In economic terms, the martech problem is ultimately a compounding learning problem.

Context provides the raw material. Technology organizes the signals. But competitive advantage ultimately belongs to the organizations that convert those signals into knowledge faster than everyone else.