Marketing Insights
Two approaches to finding what works in marketing. Leader A tries three content formats, measures what gets engagement, doubles down on the winner, and iterates. Leader B researches competitor content
Two approaches to finding what works in marketing. Leader A tries three content formats, measures what gets engagement, doubles down on the winner, and iterates. Leader B researches competitor content strategies, analyses the market for underserved niches, develops a hypothesis about what an untapped segment needs, and designs a controlled experiment to test it.
Both approaches can work. But they find different things, at different speeds, with different confidence levels. Most marketing leaders don't realize they're almost exclusively using one of them.
When I studied how 13 B2B SaaS marketing leaders actually search for what works, the split was stark: 12 of 13 primarily relied on experiential search. They learned by doing, iterating from feedback, converging on what produces results. Only 5 of 13 sustained cognitive search. These leaders used hypothesis-driven analysis, comparative reasoning, and deliberate exploration of unfamiliar territory. The split wasn't random. It mapped systematically onto professional background. The most consequential discovery is what the research literature calls failure myopia. This is a systematic blind spot where your learning system evaluates new capabilities during their worst-performance period and kills them before they can mature. Your experiential learning isn't just incomplete. It's actively misleading you about what would work if you gave it time.
Understanding this split changes how you learn, how you experiment, and how you avoid the hidden traps that each search mode produces.
Key Concepts in This Article
Cognitive Search: Drawing on Gavetti and Levinthal's (2000) dual-process model of organisational search, cognitive search is a forward-looking learning mode in which leaders use mental models, hypotheses, and analytical frameworks to identify promising approaches before committing resources. In this research, sustained cognitive search was observed in only 5 of 13 participants — predominantly corporate transplants who carried analytical infrastructure from institutional environments.
Experiential Search: The complementary mode in Gavetti and Levinthal's (2000) framework — a backward-looking learning mode in which leaders infer what works from the results of previous actions: trying, observing, adjusting, iterating. The dominant mode across 12 of 13 participants in this research. Produces rapid tacit knowledge but is bounded by what you attempt.
Search Myopia: The systematic blind spots that experiential search produces. Takes three forms: temporal myopia (favouring short feedback loops over long ones), spatial myopia (favouring proximate learning over distant learning), and failure myopia (treating early poor performance in new capabilities as evidence of inherent inferiority). Drawn from Levinthal and March (1993) and empirically visible across the research data.
The Tactical-Strategic Inversion: This research programme's conditional synthesis of the Kahneman-Gigerenzer debate. Experiential search guided by pattern recognition is ecologically rational for tactical decisions (content format, campaign optimisation) but produces systematic bias for strategic decisions (capability architecture, brand investment). The same leader can demonstrate genuine tactical expertise and strategic blind spots simultaneously.
Experiential search is how most constrained marketing leaders actually learn. You try something, observe what happens, adjust, and try again. It's fast, grounded in reality, and produces the kind of tacit knowledge that no analysis can replicate. When a startup-native leader described their approach (doing search engine optimisation, managing the website, working within zero budget), they weren't describing a strategy. They were describing a search process. Each activity produced feedback that shaped the next activity.
The data showed this pattern across nearly every participant. One startup-native leader re-mined outbound campaign data and discovered that roughly a dozen people had dropped through the cracks in the pipeline. These were prospects who had been hand-raisers during an outbound sequence but fell off between expressing interest and booking a sales call. Using an existing HubSpot licence, the leader built a four-piece re-engagement email sequence in two days. Within a day of launching, two prospects booked calls and one progressed to deal negotiation. The finding came from reviewing actual results, not from a hypothesis. Another startup-native leader tracked a hierarchical indicator chain: website activity, opportunities, revenue. These three metrics functioned as an experiential learning system, with each signal shaping the next action.
Experiential search has genuine strengths. It grounds decisions in reality rather than theory. It produces rapid feedback loops that allow fast correction. It also generates tacit knowledge. This is the kind of intuitive pattern recognition that experienced practitioners describe as "gut feel," but it's actually the accumulated residue of thousands of micro-experiments. But experiential search has a structural limitation that most practitioners never notice. You only learn from what you try. You never learn from what you don't attempt. And under constraint, what you don't attempt is everything that doesn't produce immediate, measurable results.
One solo founder illustrated this boundary neatly. When evaluating an affiliate marketing platform at $12,000 per year, the founder passed. Not because the budget was unavailable, but because they had no prior experience in affiliate marketing and couldn't estimate the return. As the founder put it: within paid lead generation and e-commerce, they could make reasonable assumptions based on experience. With affiliate marketing, they didn't know how to recruit affiliates, whether manual recruitment would work, or what realistic conversion rates looked like. The experiential learning system had nothing to draw on, so the opportunity was invisible. It wasn't rejected on evidence, but abandoned for lack of it. This is the experiential search boundary in action. Cognitive heterogeneity in what leaders can even evaluate is shaped by what they've previously tried.
Cognitive search works differently. Instead of trying things and observing results, you observe the market, form a hypothesis about what might work, and design a test to evaluate it. It's slower, more analytical, and requires infrastructure that most constrained leaders don't have. But it finds things that experiential search systematically misses.
Only 5 of 13 participants sustained cognitive search. Four were corporate transplants. One was a startup-native exception. The professional background pattern is significant. Corporate environments develop cognitive search capabilities through institutional infrastructure: market research teams, competitive intelligence functions, structured planning processes. When leaders move to startup environments, some carry these capabilities with them.
One corporate-transplant leader applied explicit Pareto reasoning to structure their marketing priorities. The goal was doing as little as possible that would bring the biggest impact, identifying the three activities that would deliver 80% of what was needed. The same leader imported an agile marketing framework from a technical background (build, measure, learn). This is a cognitive search methodology applied to marketing execution. This leader also drew on specific external references to design their approach: a 2010 Cisco article on content marketing and a New Balance case study about brand salience loss when a CEO over-indexed on demand capture. These weren't random inputs. They were analogical reasoning from prior contexts, the hallmark of cognitive search. Another corporate-transplant leader explicitly relied on historical pattern matching: primary growth channels were mostly decided based on past experience data.
The most sophisticated cognitive search in the dataset combined multiple analytical inputs. One startup-native leader developed a structured framework. This framework mapped company type to persona to pain point to competitor to channel habitat. It represented a cognitive map of the market used to identify untested opportunities. This was the exception that proved the rule: the one startup-native leader who sustained cognitive search was also the one who built the most explicit analytical infrastructure.
But cognitive search capability doesn't guarantee cognitive search practice. One corporate-transplant leader (analytically sharp, with clear views on ICP profiling, competitive positioning, and thought leadership strategy) described wanting to extract deep domain knowledge from a senior colleague. This person had decades of customer-facing experience. "Every time I had a conversation with him, I came out with so many ideas," the leader said. They envisioned an offsite to identify the company's most distinctive expertise: the moments where "the coin drops" with customers. But short-term commercial pressure made it impossible. The forecast had been pulled down four times, and leadership needed pipeline activity by the following week. The cognitive search capability was real, but the structural conditions overrode it. This is itself evidence that search mode is environmentally shaped, not just an individual trait.
The data suggests that experiential search and cognitive search are not alternatives you choose between. They're complementary capabilities, and the leaders who demonstrate both produce the most innovative approaches.
Experiential search works best in environments with rapid feedback, novel markets with no precedent to analyse, and resource-constrained contexts where speed of learning matters more than breadth. When a startup-native leader produced an entire campaign from two 45-minute founder conversations using AI tools (landing page copy, brand colours, logo, ad banners, ad copy), all without a designer or coder, at a cost that produced qualified leads at $5–6 per lead on Instagram, that was experiential search operating at maximum velocity. The capability was built by doing, not by analysing.
Cognitive search works best in established markets with analysable data, decisions with high downside risk, and contexts where prior knowledge can be transferred.
The critical capability isn't mastering one mode. It's knowing which mode to use when, and being able to switch. The leaders who relied exclusively on experiential search converged quickly on what worked. But they converged on local optima. The leaders who incorporated cognitive search explored more broadly. Sometimes this came at the cost of speed. The ideal is what the research literature describes as an initial cognitive search to identify promising regions, followed by experiential search to fine-tune within those regions (Gavetti & Rivkin, 2007).
There's a deeper pattern here, and it's the most practically useful finding in the research. Your gut feeling is reliable for tactical decisions. Which content format to test? Which ad copy to run? How to optimize a campaign that's already working? You've had enough repetitions at that level to develop genuine pattern recognition. But your gut is unreliable for strategic decisions. Whether to invest in brand? How to allocate between activation and long-term capability building? When to enter a new channel category? Nobody gets enough repetition at the strategic level to develop reliable intuition. The environment changes too slowly, the feedback is too ambiguous, and the sample of personal experience is too small. This is why the same leaders who make excellent tactical judgements fall into the activation trap at the strategic level. This is also why knowing about the bias doesn't help overcome it.
To prevent confusion, it's worth being explicit about what these concepts are not. Practitioners will pattern-match to familiar ideas and miss the distinctive contribution.
It's not A/B testing. A/B testing is a method. Experiential search is a learning mode. You can run A/B tests as part of experiential search, but experiential search also includes informal observation, pattern recognition from accumulated experience, and the tacit knowledge built through years of "doing marketing." The search mode concept is about how you learn, not which tools you use to measure.
It's not "data-driven marketing." Cognitive search doesn't mean looking at dashboards. Dashboard analysis is typically backward-looking. It interprets what already happened. This makes it experiential search with better instrumentation. Cognitive search means forming hypotheses about untested territory. What would happen if we targeted a different segment? If we entered a different channel? If we restructured our capability architecture? It's forward-looking analysis of what you haven't tried, not backward-looking analysis of what you have.
It's not System 1 versus System 2\. The mapping is tempting but misleading. Kahneman's System 1/System 2 describes individual cognitive processing speed. Search mode describes an organisational learning orientation that operates over weeks and months, not milliseconds. Experiential search can involve slow, deliberate analysis of results. Cognitive search can produce fast, intuitive hypotheses. The distinction is about what generates the learning input (prior action or prior reasoning), not about how fast you think.
It's not the lean startup's build-measure-learn. Build-measure-learn is a specific method for product iteration that operates within the experiential search paradigm. It tells you to learn from what you build. Cognitive search asks what you should build before you build it. It asks whether the learning you'll get from building is actually the learning you need. The lean startup assumes that doing is the best way to learn. This research shows that for strategic decisions, doing is often the worst way to learn because feedback is too slow and too noisy.
It's not simply "exploration versus exploitation." March's (1991) framework describes a resource allocation tension. Search mode describes a cognitive orientation. You can explore (try new things) using either experiential search (just try it and see) or cognitive search (hypothesise about what's worth trying, then test). The explore/exploit balance tells you how much novelty to pursue. Search mode determines how you pursue it. It also determines whether you can even see the novel options that exist outside your experiential boundary.
Here's where the search mode distinction becomes practically consequential. Each mode produces its own form of blindness.
Experiential search naturally favours immediate feedback over distant signals. You learn what worked this week, this month, this quarter. You don't learn what would have worked over 18 months because you never ran an 18-month experiment. One startup-native leader captured this vividly. As a company grows and should start investing more in brand, the leaders don't do it. They're addicted to the here and now. The "addiction" is a search myopia. The experiential learning system produces valid short-term knowledge that systematically excludes long-term knowledge.
Experiential search favours local market learning over broader patterns. You learn what works with your current customers, in your current market, through your current channels. You don't learn what's happening in adjacent markets, with different customer segments, through channels you haven't tried. One startup-native leader's information environment was explicitly local. They relied exclusively on internal performance data, because external data was "too cumbersome and expensive to access." The spatial boundary of experiential search was also the spatial boundary of strategic imagination. This leader wasn't irrational. External market data from firms like Gartner or Forrester can cost $50,000–60,000 per report for a small organisation. The constraint is real, but the consequence is that learning becomes bounded by what you can observe from where you already stand.
This is the subtlest and most damaging. Experiential search learns from what you tried that failed. It never learns from what you didn't try. New marketing capabilities (content marketing, thought leadership, community building) typically perform poorly initially before improving through learning. Experiential search evaluates them during the poor-performance period and concludes they're inferior to established activities.
One startup-native leader gave a textbook example of this dynamic. The company launched competitor ads, a new capability they hadn't previously used. The ads didn't deliver leads within a week. The team decided it was "not successful" and killed it. But when reflecting on the experience, the leader recognised that they had "jumped to conclusions too quick." There were multiple factors to examine: ad copy, conversion path, landing page. Before concluding the channel didn't work, you'd want to test these. The awareness came after the fact. In the moment, the experiential learning system treated one week of data as a definitive verdict. The capability was never built. This is metacognitive awareness without metacognitive override: knowing you have the bias doesn't prevent it from operating.
The combined effect of these three myopias is convergence on local optima. You develop a marketing approach that works adequately within its narrow parameters but never discovers the globally superior alternatives that exist outside the experiential learning boundary. This is the search-level mechanism that produces the activation trap.
The distinction between cognitive and experiential search draws on four theoretical streams that together explain how search mode shapes capability development.
Gavetti and Levinthal's (2000) dual-process model of organisational search provides the foundational architecture. Cognitive search is forward-looking: leaders use mental models to identify promising regions of the solution space before committing resources. Experiential search is backward-looking: leaders infer what works from the results of previous actions. The optimal strategy (initial cognitive search followed by experiential refinement) outperforms either mode in isolation (Gavetti & Rivkin, 2007).
March's (1991) exploration-exploitation framework explains why experiential search dominates under constraint. Exploitation produces reliable, near-term returns; exploration produces uncertain, delayed returns. Under resource constraint, the opportunity cost of exploration rises. The activation trap is, in March's terms, a premature shift from exploration to exploitation. The firm converges on demand-generation approaches before adequately exploring brand-building alternatives.
Levinthal and March's (1993) myopia of learning specifies three systematic distortions that experiential learning produces: temporal myopia (favouring short feedback loops over long ones), spatial myopia (favouring proximate activities over distant ones), and failure myopia (treating early failures as evidence of inherent inferiority rather than as the learning cost of capability development). All three are empirically visible in the research data.
Gigerenzer's (2008) ecological rationality framework provides the counterpoint. In environments where feedback is fast, cues are valid, and the decision-maker has adequate experience, heuristic search outperforms analytical search. The early-stage B2B SaaS environment meets these conditions for tactical decisions but not for strategic decisions. This explains why experienced leaders show tactical expertise and strategic bias simultaneously.
The micro-foundational decomposition of the research data identified 84 distinct cognitive process types across 13 participants, distributed across five domains: means assessment (12 types), problem formulation (18 types), constraint reasoning (15 types), decision-making under uncertainty (22 types), and prototype action (17 types). This cognitive process inventory provides the empirical basis for understanding how search modes operate in practice.
The dominance of experiential search. Twelve of 13 participants primarily rely on experiential search as their dominant learning mode. This manifests as feedback interpretation (the richest evidence cluster in the dataset), adaptive reasoning, and iterative adjustment. The experiential search pattern is consistent across cohorts. Both startup natives and corporate transplants learn primarily by doing.
The contingency of cognitive search. Only 5 of 13 sustain cognitive search alongside experiential search: 4 corporate transplants (P\_4, P\_6, P\_7, P\_9) and 1 startup-native exception (P\_10). The classification of P\_9 merits a note: while P\_9 explicitly observes that organisational politics often override data-driven analysis, they nonetheless describe analytical approaches to market assessment and competitive positioning that qualify as cognitive search capability. The fact that structural conditions sometimes override that capability is itself evidence for the structural causation thesis rather than evidence against the capability's existence. The startup-native exception (P\_10) is analytically significant. This participant built explicit cognitive search infrastructure (the company-persona-pain-competitor-habitat framework) without institutional support, suggesting that cognitive search capability can be deliberately constructed under constraint.
Search mode and innovation. The data is consistent with Gavetti and Levinthal's prediction that cognitive search produces more innovative approaches. The participant demonstrating the broadest cognitive search (P\_10) also described the most innovative marketing approach in the dataset. This approach was an automation-first, AI-augmented system that differed fundamentally from any other participant's approach. The participants with the narrowest search (exclusively experiential, no cognitive component) described the most conventional approaches.
Cohort patterns in search mode. The startup-native cohort shows stronger experiential search capabilities: rapid calibration, realistic constraint assessment, high adaptability. The corporate-transplant cohort shows stronger cognitive search capabilities: systematic analysis, structured hypothesis-testing, analogical reasoning from prior contexts. These are not quality differences. They reflect adaptation to different professional environments.
The research programme's most consequential theoretical contribution to the search mode literature is the conditional synthesis (what I've termed the tactical-strategic inversion), developed from the Kahneman-Gigerenzer debate. The Gigerenzer tradition argues that simple heuristics are ecologically rational when the environment supports them. The Kahneman tradition argues that heuristics produce systematic biases that degrade decision quality.
The data supports a conditional resolution. At the tactical level (content format selection, campaign optimisation, channel tuning), experiential search guided by pattern recognition is ecologically rational. Feedback is fast enough, cues are valid enough, and experience is accumulated quickly enough that heuristic judgement outperforms elaborate analysis.
At the strategic level (capability architecture, activation-versus-brand allocation, long-term investment planning), experiential search produces systematic bias. Feedback is too slow, cues are too ambiguous, and the sample of personal experience is too small to calibrate reliable intuition.
The Kahneman-Klein (2009) convergence provides the diagnostic: reliable intuitive expertise requires a high-validity environment (stable, learnable regularities) and adequate learning opportunity (prolonged practice with rapid, accurate feedback). Tactical marketing meets both conditions. Strategic marketing meets neither.
Within the broader Effectual Orchestration framework, search mode determines the quality of the means audit and the creativity of the bundling phase. Experiential search produces thorough knowledge of available means. Leaders who learn by doing have intimate knowledge of what their current resources can accomplish. Cognitive search produces novel means configurations. Leaders who form hypotheses about untested combinations discover capabilities that experiential search would never reveal.
The effectual logic that drives the EO model (means-driven, affordable-loss, experimental) is essentially an experiential search strategy optimised for extreme uncertainty. But the conditional synthesis adds a warning: effectual logic is ecologically rational for tactical execution but may be insufficient for strategic capability development. The means-driven approach naturally favours available resources over needed resources, current capabilities over missing capabilities, and measurable outcomes over unmeasurable outcomes. These preferences produce strategic drift: a gradual accumulation of tactical optimisations that fails to produce coherent capability architecture.
The prescription is not to abandon experiential search but to supplement it with deliberate cognitive search at the strategic level. This means building even minimal analytical infrastructure: periodic competitive analysis, structured reflection on capability gaps, deliberate hypothesis-testing about untested approaches, and (perhaps most importantly) diverse information inputs that break the algorithmic closure described in Network-Mediated Capability Building.
If you're a startup-native leader, your experiential search capabilities are probably strong. You learn fast from what you try. The risk is that you only learn from what you try. Under constraint, what you try is limited to the familiar and the measurable. The most valuable intervention isn't more experimentation. It's building minimal cognitive search infrastructure: a monthly competitive review, a quarterly reflection on what you haven't tried, a conversation with someone whose information diet is different from yours.
If you're a corporate-transplant leader, your cognitive search capabilities probably survived the transition to startup life. The risk is that your mental models (the analytical frameworks you carry from corporate experience) may not apply to your current context. The pattern recognition that makes you strategically sophisticated in familiar environments can produce overconfidence in unfamiliar ones. The most impactful intervention isn't more analysis. It's faster experiential feedback: shorter experiment cycles, more direct customer exposure, less time between hypothesis and test.
And regardless of background, watch for the three myopias. If all your learning comes from this quarter's results, you have temporal myopia. If all your learning comes from your current customers, you have spatial myopia. If you've never run an experiment long enough for a new capability to mature past its initial poor-performance period, you have failure myopia. The activation trap isn't just a resource allocation problem. It's a search problem. You're trapped not because you can't afford to invest differently, but because your learning system never shows you that different investments would work.
Next in this series: The Structural Causation Thesis — Why Marketing "Biases" Are Environmentally Produced.
This post is part of a 10-part foundation series exploring how marketing capabilities emerge under constraint. The concepts draw on an ongoing research programme involving 13 in-depth interviews with B2B SaaS marketing leaders, analysed through the lens of effectuation theory, resource orchestration, and cognitive micro-foundations. Browse all pillars.
References
Gavetti, G. and Levinthal, D.A. (2000) 'Looking Forward and Looking Backward: Cognitive and Experiential Search', Administrative Science Quarterly, 45(1), pp. 113–137.
Gavetti, G. and Rivkin, J.W. (2007) 'On the Origin of Strategy: Action and Cognition over Time', Organization Science, 18(3), pp. 420–439.
Gigerenzer, G. (2008) 'Why Heuristics Work', Perspectives on Psychological Science, 3(1), pp. 20–29.
Kahneman, D. and Klein, G. (2009) 'Conditions for Intuitive Expertise: A Failure to Disagree', American Psychologist, 64(6), pp. 515–526.
Levinthal, D.A. and March, J.G. (1993) 'The Myopia of Learning', Strategic Management Journal, 14(S2), pp. 95–112.
March, J.G. (1991) 'Exploration and Exploitation in Organizational Learning', Organization Science, 2(1), pp. 71–87.
