Series: Thinking in Systems for Digital Marketers | Article 5 of 6
This series applies the core concepts of Donella Meadows’ Thinking in Systems: A Primer to the realities of digital marketing. Each article draws directly from the book and translates its frameworks into practical tools for mid-level marketers. If you’re just joining us, we recommend starting with Article 1.
“Everything we think we know about the world is a model. And our models are simpler than the world.” — Donella Meadows, Thinking in Systems
You’ve been there. A campaign that worked beautifully last quarter collapses this one — same creative, same audience, same budget. A channel you wrote off six months ago suddenly starts driving your best leads. A competitor you’d barely noticed captures a meaningful slice of your market seemingly overnight. A strategy your entire team agreed on produces results that nobody predicted and nobody can fully explain.
The instinct is to look for what changed. What did we do differently? What did the platform do? What did the market do? The search for a single cause to explain a single surprising effect feels like the right diagnostic. It almost never is.
Marketing surprises are not random. They are what complex systems do when the people running them mistake their mental model of the system for the system itself. Donella Meadows dedicated an entire chapter of Thinking in Systems to explaining why systems consistently produce outcomes that surprise even the most experienced people managing them — and her diagnosis is both uncomfortable and clarifying.
The surprises are not the system’s fault. They are the gap between how the system actually works and how we think it works. And that gap has four recurring shapes.
Surprise #1: Your mental model of the campaign is not the campaign
Meadows opens her chapter on system surprises with a statement that should stop every marketer cold:
“We are surprised because our models — our mental maps of reality — are always simpler than the reality they represent. The map is not the territory.”
Every campaign plan, every funnel diagram, every attribution model you use is a simplification. It captures some of the system’s behavior and ignores the rest. This is not a flaw in your thinking — it is an unavoidable feature of how humans process complexity. We cannot hold the full system in our minds, so we build models that are good enough to act on.
The problem is not the simplification. The problem is forgetting that the simplification is happening.
When a campaign underperforms, most marketers interrogate their execution against their mental model: did we target correctly? Did we spend enough? Did the creative match the brief? These are questions about whether the execution matched the plan. They are almost never questions about whether the plan captured how the system actually works.
The deeper diagnostic — the one Meadows would ask — is: what did our model leave out? What relationship, what delay, what feedback loop was real in the system but absent from our planning assumptions?
In practice, this means asking questions that feel uncomfortable precisely because they challenge the framework rather than the execution:
- Did we assume a linear relationship between budget and results in a system that is nonlinear?
- Did we assume our attribution model was showing us cause and effect when it was showing us correlation?
- Did we assume the audience would behave this quarter the way it behaved last quarter?
- Did we assume platform behavior would be stable when platform behavior is a variable?
Most marketing surprises are not surprises at all when you look back at them. They were predictable outcomes of dynamics that were real in the system but missing from the model. The question is not why the system surprised you. It is what your model failed to include.
Surprise #2: You underestimated the delay between cause and effect
We covered time delays in Article 3 in the context of pipeline management. But delays are so central to why systems surprise marketers that Meadows returns to them repeatedly — and with increasing urgency:
“The world often surprises us because of delays between causes and their effects. We think we see what is happening right now, but we are actually seeing the consequences of what happened weeks or months ago.”
This is not just a pipeline problem. It is a universal feature of marketing systems — and it produces surprises in both directions.
When a strategy is working, the delay between cause and effect means the results you’re seeing today were produced by decisions made weeks ago. Marketers frequently attribute current performance to recent actions — the campaign launched last week, the creative refreshed last month — when the actual driver was a brand investment, a channel mix decision, or an audience development effort made a quarter earlier. The wrong cause gets credited. Future decisions get built on a false understanding of what actually worked.
When a strategy is failing, the delay means the damage accumulates invisibly before it surfaces. A declining organic presence, a weakening email list, a narrowing paid audience — these erosions happen gradually and show up suddenly. By the time the surprise hits the dashboard, the system has been signaling the problem for weeks.
Meadows identifies a specific pattern she calls policy resistance — where a manager responds to a system problem with an intervention, the intervention takes time to work, and the manager — not seeing immediate results — applies more of the intervention before the first dose has had its effect. In marketing, this looks like increasing budget on a failing campaign before the algorithm has had time to respond, or launching a new creative before the previous one has completed its learning phase.
“The delayed response leads to overreaction, which leads to overcorrection, which leads to oscillation. The system swings between two states not because it is broken — but because its managers are reacting to old information.”
The discipline required here is one of the hardest in marketing: acting on a model of the system’s future state rather than its current dashboard. Investing in inflows before the pipeline shows depletion. Refreshing creative before fatigue is confirmed. Developing new channels before existing ones plateau. None of these feel urgent in the moment. All of them prevent the surprises that feel catastrophic in retrospect.
Surprise #3: You assumed linearity in a nonlinear system
Most marketing planning assumes that relationships between variables are roughly linear. Double the budget, roughly double the results. Improve the conversion rate by 10%, revenue grows by 10%. More impressions, more awareness. These assumptions are intuitive, easy to model, and frequently wrong.
Meadows is direct about this:
“Nonlinearities are all around us. They are the rule, not the exception. Yet our minds are poorly equipped to deal with them, and our models almost always assume linear relationships where none exist.”
Marketing systems are saturated with nonlinearity. Audience targeting exhibits nonlinear returns — the first layer of targeting refinement dramatically improves efficiency; beyond a certain point, further refinement starves delivery and raises costs. Budget scaling is nonlinear — a campaign that performs at $500 per day may not perform proportionally at $5,000 per day, because the audience pool and auction dynamics are fundamentally different at different spending levels. Brand awareness is nonlinear — the relationship between awareness investment and purchase intent is not a straight line. It builds slowly, then tips.
The practical consequence of nonlinearity is that results from one context do not transfer linearly to another context. A campaign that performed well at a small scale does not simply perform proportionally at a large scale. A channel that works for acquiring customers at low volume may exhibit completely different economics at high volume. A message that resonates with an early audience may not resonate with the broader market.
This is why “we’ll just scale what’s working” is one of the most dangerous strategies in digital marketing. Scaling is not multiplication. It is moving into a different part of the system’s behavior curve — one where the relationships between your inputs and outputs may be fundamentally different from the relationships you observed at the scale where you validated your assumptions.
Before scaling anything, the right question is not “did this work?” It is “at what scale did this work, and what might change about the system’s behavior at the scale we’re planning to reach?”
Surprise #4: You mistook a short-term fix for a solution
The fourth and perhaps most insidious source of marketing surprises is what Meadows calls the shifting of the burden — a systems archetype in which a symptomatic solution addresses the visible problem while leaving the underlying structure intact. The symptom improves. The system relaxes. The underlying problem quietly worsens. Eventually, the symptom returns — larger, more resistant, more expensive to treat.
“The most pernicious form of system trap is one in which a short-term solution that relieves a symptom also weakens the system’s ability to address the fundamental problem. The better the symptomatic solution works, the less pressure there is to find a real one.”
In marketing, this pattern is everywhere.
A brand with a weakening value proposition runs aggressive promotional campaigns. The promotions drive short-term conversion. The pressure to fix the underlying positioning problem dissipates. The next quarter, the promotions need to be deeper to produce the same lift. The reliance on discounting grows. The brand’s price integrity erodes. The value proposition weakens further. The surprise — “why is our margin collapsing?” — is the predictable outcome of a burden that kept shifting.
A company with declining organic reach doubles its paid media spend. Paid fills the gap. The urgency to invest in content, SEO, and brand building fades. The organic foundation continues to erode. Paid costs rise as competition increases. The company becomes progressively more dependent on paid acquisition at progressively worse economics. The surprise — “why is our customer acquisition cost rising every year?” — is the cumulative effect of repeatedly choosing the symptomatic fix over the structural one.
A marketing team that misses its targets hires more people. Headcount rises. The pressure to examine whether the strategy, the channel mix, or the targeting model is fundamentally sound never materializes. The team gets larger without getting more effective. The surprise — “why aren’t more resources producing better results?” — is the outcome of treating a strategic problem as a resource problem.
Meadows’ prescription is uncomfortable because it requires resisting the symptomatic solution even when it works in the short term:
“The way out of the burden-shifting trap is to find ways to strengthen the fundamental solution — even at the cost of short-term pain — and to weaken, not strengthen, the symptomatic solution over time.”
In marketing terms: fix the positioning, not just the promotion. Build the organic foundation, not just the paid layer. Improve the strategy, not just the headcount. These are harder, slower, and less immediately satisfying than the symptomatic fix. They are also the only interventions that change the system’s behavior at the level where the surprise is actually being generated.
A diagnostic for the next time marketing surprises you
When something unexpected happens — a campaign that shouldn’t have failed, a channel that shouldn’t have worked, a result that nobody predicted — run through these four questions before reaching for a single-cause explanation.
1. What did my model leave out? Identify the simplifications in your planning assumptions. What relationships, delays, or feedback loops were real in the system but absent from your model? The surprise almost always lives in the gap between the map and the territory.
2. Am I reacting to current data or to a delay? Ask whether the result you’re seeing now is the consequence of a decision made recently — or a decision made weeks or months ago. Identify the delays in your system and make sure you’re attributing causes to the right time horizon.
3. Where did I assume linearity that the system is nonlinear? Challenge the proportionality assumptions in your plans. Does the relationship you observed at one scale hold at another? Does the channel mix that worked at one budget level work the same way at a different one?
4. Am I treating a symptom or solving the problem? Ask whether your proposed response addresses the fundamental structure producing the problem — or whether it relieves the pressure without fixing the source. If it relieves pressure without fixing the source, ask what you are making it easier to avoid.
Meadows closes her chapter on system surprises with a passage that reads, in retrospect, like a quiet challenge to every person who has ever been blindsided by a result they should have seen coming:
“We can’t avoid being surprised by systems. But we can reduce our surprise by being humble about what we don’t know, by building better models, by watching for delays, and by resisting the temptation of the fix that feels good right now.”
Marketing surprises will not stop. The system is too complex, the variables too many, the delays too long for perfect prediction. But surprise is not the problem. Surprise without learning is the problem — responding to unexpected outcomes by adjusting execution rather than interrogating assumptions, by fixing symptoms rather than structures, by reacting to the dashboard rather than understanding the system behind it.
The marketers who get surprised less often are not smarter. They are more honest about what their models leave out.
Next in the series: Article 6 — Leverage points: the one change in your marketing system that moves everything else
This series is based on Donella Meadows’ book Thinking in Systems: A Primer (Chelsea Green Publishing, 2008), interpreted through a digital marketing lens.
