The Hidden Cost of Optimization: How Over-Fitted Algorithms Destroy Long-Term Brand Equity

TL;DR: Every percentage point improvement in short-term conversion metrics may be borrowing against brand equity you cannot see on any dashboard. Goodhart's Law applied to marketing means that optimizing for CPA or ROAS as a target systematically erodes the brand associations that create long-term pricing power -- and the damage compounds invisibly for years before surfacing as declining organic demand and rising acquisition costs.

---

The Algorithm That Ate Your Brand

We need to talk about a corporate disease that looks, on every dashboard, like perfect health.

Your performance marketing team is running a machine. Every quarter, cost-per-acquisition drops. Click-through rates climb. Return on ad spend holds steady or improves. The algorithm — whether it lives inside Meta, Google, or your own bidding stack — learns, refines, converges. The numbers are beautiful.

And underneath those numbers, your brand is dying.

Not dramatically. Not in a way that triggers an alert or a board-level conversation. Slowly. Like osteoporosis. The bones look fine on the outside until one day they don't, and by then the structural damage has been compounding for years.

This essay is about that compounding. About the specific mechanism by which marketing algorithms, doing exactly what they were designed to do, systematically erode the thing that makes a company worth paying attention to in the first place. It is about what happens when a measure becomes a target.

Goodhart's Law and the Marketing Machine

In 1975, Charles Goodhart, a British economist advising the Bank of England, observed something about monetary policy targets that has since become one of the most cited principles in social science: "When a measure becomes a target, it ceases to be a good measure." Expressed more formally:

$$\text{If } x \text{ is a good proxy for } y, \text{ then optimizing } x \text{ directly causes } \frac{\partial \, \text{corr}(x, y)}{\partial t} \lt 0$$

This principle, now known as Goodhart's law, applies with terrifying precision to modern marketing.

Goodhart was talking about money supply metrics. The Bank of England had been targeting M3 — a broad measure of money supply — as a proxy for inflation control. The moment M3 became the official target, banks and financial institutions changed their behavior to game the metric. M3 stopped reflecting what it had originally measured.

This principle applies with terrifying precision to modern marketing.

Consider what happens when we tell an algorithm to minimize cost-per-click or maximize return on ad spend. The algorithm does not understand brand building. It does not understand customer lifetime value beyond the attribution window we give it. It does not understand that a 22-year-old encountering your brand for the first time while browsing Instagram might be worth $14,000 over the next decade, even though she didn't click your ad today.

The algorithm understands one thing: the metric we told it to target.

So it targets that metric. Ruthlessly. Brilliantly. It finds the people most likely to click. It finds the creative most likely to generate an immediate response. It finds the placement most likely to produce a measurable conversion within the lookback window.

And in doing so, it systematically ignores — or actively avoids — every activity that builds long-term brand equity.

This is not a failure of the algorithm. It is the algorithm succeeding at exactly what we asked. The failure is ours. We asked the wrong question.

The Survivorship Bias in Click Optimization

Here is the mechanism, and it is more specific than most marketers realize.

When a bidding algorithm runs for several months, it begins to build a model of "who converts." It analyzes thousands of signals — demographics, browsing history, device type, time of day, purchase history — and converges on a profile of the ideal target.

That profile, almost invariably, describes someone who already knows your brand.

Think about it. Who is most likely to click on your ad? Someone who searched for your brand name. Someone who visited your site last week. Someone who abandoned a cart. Someone who follows you on social media. These are not new customers. These are people your brand already won. The algorithm is claiming credit for demand that already existed.

This is survivorship bias in real time. The algorithm never sees the counterfactual — the customer who would have converted anyway without the ad, or the future customer who needed a brand impression today to convert three years from now. It only sees what happened within its attribution window. And so it allocates more and more budget toward people who need the least convincing.

The chart tells the story. Over two years of algorithm-driven optimization, the share of budget reaching genuinely new audiences collapses from 62% to 11%. The algorithm is not expanding your market. It is mining your existing market with increasing efficiency — and calling it growth.

We have a name for a system that extracts from existing reserves without replenishing them. We call it depletion. In natural resource economics, the term is "mining the asset." It produces excellent short-term returns. The long-term outcome is predictable.

Binet and Field: The Evidence We Keep Ignoring

In 2013, Les Binet and Peter Field published "The Long and the Short of It," an analysis of the IPA (Institute of Practitioners in Advertising) Databank — the largest database of advertising effectiveness case studies in the world. Their dataset covered over 900 campaigns across decades.

Their central finding has been validated repeatedly in the years since, and the marketing industry has spent that entire time finding reasons not to act on it.

The finding: campaigns that balance brand building with short-term activation produce substantially better business results over time than campaigns focused on either approach alone. The approximate ratio: 60% brand, 40% activation.

This is not a philosophical claim. It is an empirical result drawn from the largest controlled dataset in advertising history.

The specific mechanisms Binet and Field identified:

Brand campaigns work by creating and refreshing memory structures — mental associations between your brand and buying situations. They operate on broad audiences. They are emotional rather than rational. Their effects are slow to build and slow to decay. They reduce price sensitivity, increase willingness to pay, and create "mental availability" — what Ehrenberg-Bass researchers call category entry points (the probability that your brand comes to mind in a buying situation).

Activation campaigns work by converting existing demand. They target people who are already in-market. They are rational, specific, and short-lived. They trigger immediate behavioral responses. Their effects appear quickly and decay quickly.

The crucial insight: activation without brand building is like harvesting without planting. You can do it for a while. The yields look great. And then the field is empty.

The data shows a clear peak in business effects at approximately the 60/40 brand-to-activation ratio. The exact ratio varies by category — Binet and Field found that categories with longer purchase cycles benefit from higher brand shares, while impulse-purchase categories can shift slightly toward activation. But the directional finding is consistent: most companies spend too much on activation and too little on brand. Quantifying the optimal split requires the kind of rigorous brand-performance portfolio optimization that borrows from Modern Portfolio Theory.

And the gap has widened since 2013. The rise of programmatic advertising, automated bidding, and attribution-obsessed marketing cultures -- what we might call the attention economics of cognitive load in advertising applied at industrial scale -- has pushed allocation further and further toward measurable, short-term activation. Many companies now run at 80/20 or 90/10 in favor of activation. Some have eliminated brand spending entirely.

They are eating their seed corn.

The Adidas Confession

In October 2019, Simon Peel, Adidas's global media director, made an admission that sent tremors through the marketing industry. Speaking at an Effie Awards event, he stated publicly that Adidas had over-invested in performance marketing at the expense of brand building — and that the company had the data to prove it was a mistake.

The specifics: Adidas had been allocating approximately 77% of its media budget to performance (what they called "digital performance") and just 23% to brand. Internal econometric modeling revealed that brand activity was driving 65% of their sales across wholesale, retail, and e-commerce. Performance marketing was driving just 35%.

Read that again. The channel receiving 77% of the budget was generating 35% of the sales. The channel receiving 23% of the budget was generating 65% of the sales.

This is what Goodhart's Law looks like at a $20 billion company. The performance campaigns looked efficient. They had measurable, attributable returns. They showed up beautifully in dashboards. And they were systematically cannibalizing the brand investment that was actually driving the majority of demand.

Peel was blunt about the cause: over-reliance on last-click attribution. The attribution model rewarded the final touchpoint — the performance ad that someone clicked before buying. It gave zero credit to the brand campaign that had, over the preceding months and years, created the awareness and consideration that made the click possible.

Adidas was not unique. They were simply the first major company honest enough to say it publicly. Their econometric work showed something that academic researchers like Binet and Field had been documenting for years: brand building is the primary demand generator, and performance marketing is the demand harvester. When you starve the generator to feed the harvester, you get a system that looks efficient right up until the moment it collapses.

The Brand Decay Curve

What happens when brand investment stops? The answer is not intuitive, and the non-intuitiveness is precisely what makes the problem so dangerous.

Brand equity does not decline linearly. It follows a decay curve that looks, mathematically, much like radioactive half-life — but with a critical twist. The brand decay function can be expressed as:

$$B(t) = B_0 \cdot e^{-\lambda t}$$

where $B_0$ is the initial brand equity, $\lambda$ is the decay constant (typically 0.03--0.06 per month for consumer brands), and $t$ is time in months since investment ceased. The half-life $t_{1/2} = \frac{\ln 2}{\lambda}$ ranges from 12 to 24 months depending on category and initial brand strength.

In the early months after brand investment stops, almost nothing visible happens. Awareness stays high. Consideration stays stable. Sales continue.

This is the trap. The CFO cuts brand budget in Q1. Q2 results look fine. Q3 results look fine. Everyone congratulates themselves on the efficiency gain. The savings flow to the bottom line or get reallocated to performance marketing, which shows immediate, attributable returns.

Then, sometime between month 12 and month 24, the decay becomes visible. Unaided awareness begins to drop. Consideration declines. New customer acquisition costs start rising — slowly at first, then accelerating. The brand has been living on residual equity, and the reserves are now depleted.

Three lines. Three stories. Brand equity (dark blue) begins declining immediately but the decline is masked for the first two quarters because sales volume (medium blue) lags behind. By the time sales decline is undeniable at month 12-15, brand equity has already lost 30-40% of its value. Meanwhile, new customer acquisition cost (red) begins its exponential climb.

The asymmetry between destruction and reconstruction is brutal. Millward Brown's BrandZ data shows that it takes approximately 3-5 years of consistent investment to build a brand position that can be destroyed in 18-24 months of neglect. The decay is faster than the build. This is not a symmetric game.

We call this the Brand Decay Asymmetry: the rate of brand equity destruction exceeds the rate of brand equity creation by a factor of approximately 2-3x. Building takes years. Destruction takes quarters. And the destruction is invisible on any standard marketing dashboard until it is already advanced.

Why CAC Rises When Brand Investment Falls

The relationship between brand investment and customer acquisition cost is one of the most important — and least discussed — dynamics in marketing economics. Here is the mechanism.

When your brand is strong, several things happen that reduce acquisition costs:

People search for you by name. Branded search is the cheapest traffic you can buy. When brand awareness is high, a meaningful percentage of your customers come to you directly — they type your name, they search your brand, they go to your website. This traffic costs pennies. Performance marketing does not create this behavior. Brand marketing creates it.

People trust your ads. Ad engagement rates are higher for brands people recognize. A Facebook ad from a known brand converts at 2-5x the rate of an identical ad from an unknown brand. This is not because the creative is better. It is because the brand has pre-existing memory structures that make the ad feel familiar rather than foreign.

People accept your price. Strong brands command price premiums. Weak brands compete on price. When you compete on price, your performance marketing must work harder — lower margins mean lower allowable CAC, which means more aggressive bidding, which means higher costs per result.

People refer others. Word of mouth is free. Strong brands generate more of it. Every referral is a customer you did not have to acquire through paid channels.

When brand investment declines, each of these effects weakens. Branded search volume drops. Ad engagement declines. Price sensitivity increases. Referrals slow. And the entire load of customer acquisition shifts to paid performance channels — which, remember, are now more expensive because the brand isn't doing the pre-selling work.

This creates a vicious cycle that we call the Brand-CAC Death Spiral.

The chart shows three five-year trajectories for customer acquisition cost (in USD). The balanced approach (60/40 brand-to-activation) shows CAC declining slightly over time as brand effects compound. The performance-heavy approach (20/80) shows CAC rising steadily as brand equity depletes. The performance-only approach (0/100) shows CAC nearly quadrupling in five years.

Year 1 tells the wrong story. In the first year, the performance-only approach has the lowest CAC. This is why CFOs love it. This is why boards demand it. It looks like the most efficient option — because it is harvesting brand equity that was built with previous investment. By Year 3, the lines have crossed. By Year 5, the performance-only approach costs nearly 4x the balanced approach to acquire a customer.

The companies that cut brand budgets in response to rising CAC are, almost without exception, making the problem worse. They are applying more performance pressure to a system that is failing precisely because of insufficient brand pressure. It is the marketing equivalent of treating dehydration by running faster.

The Optimization Trap Framework

We propose a framework for understanding how companies fall into — and can escape from — the cycle of over-optimization. We call it the Optimization Trap, and it has five stages.

Stage 1: The Efficiency Mandate. A company decides to make marketing "more efficient" or "more accountable." This usually comes from finance or from a new CMO under pressure to show ROI. The mandate is not wrong in itself. Accountability is good. The problem is the definition of "efficiency" — it is almost always defined in terms of short-term, attributable returns.

Stage 2: The Attribution Mirage. The company implements attribution modeling — usually last-click or some variant of multi-touch attribution. Immediately, performance channels look more efficient than brand channels. This is because performance channels are designed to capture demand at the moment of conversion. They are, by definition, closer to the point of attribution. Brand channels, which created the demand in the first place, are invisible in the model.

Stage 3: The Reallocation. Based on attribution data, budget shifts from brand to performance. The immediate results are positive. ROAS improves. CAC drops (temporarily). The attribution model confirms the decision. Everyone is pleased. The algorithm begins to optimize.

Stage 4: The Silent Erosion. Over 12-24 months, brand equity declines. The effects are invisible in standard dashboards. Unaided awareness drops but nobody is measuring it. Branded search volume declines but it is buried in aggregate search data. Consideration among non-customers weakens but nobody is surveying them. The algorithm continues to optimize, targeting an ever-narrower audience of existing customers and high-intent prospects.

Stage 5: The Cliff. Performance metrics begin to deteriorate. CAC rises. ROAS declines. The audience is exhausted. The algorithm has nowhere to go — it has already found everyone who was easy to convert. The company responds by increasing performance spend, which accelerates the spiral. By this point, brand equity has been depleted to the point where rebuilding requires 3-5 years and significant investment.

Most companies that contact us are in Stage 4. They can feel something is wrong but can't identify it because their measurement systems were designed to track exactly the metrics that look fine until Stage 5 hits. The irony is perfect: the more sophisticated your performance measurement, the more invisible the brand problem becomes.

Goodhart Dynamics in Automated Bidding

Let us get specific about how automated bidding systems — the engines of Google Ads, Meta Ads, and every major programmatic platform — create Goodhart dynamics at scale.

An automated bidding system has a simple objective function: given a budget, maximize some target metric (conversions, revenue, clicks) by deciding how much to bid for each impression opportunity. The system processes millions of signals per second and learns, through reinforcement, which combinations of audience, placement, creative, and timing produce the best results.

The learning process itself introduces three specific forms of Goodhart corruption.

Corruption 1: Audience Narrowing. The algorithm discovers that certain audience segments convert at higher rates. It allocates more budget to these segments. These segments — invariably — are people who already know your brand. The algorithm then reports improving efficiency metrics, because it is spending less on the broad, new-audience impressions that had lower immediate conversion rates but were generating future demand. The metric improves. The reality worsens.

Corruption 2: Creative Convergence. The algorithm tests multiple creative variants and allocates budget to winners. The "winning" creative is the one that generates the most immediate responses — which means direct-response creative with clear calls to action, price messaging, and urgency signals. Brand-building creative — emotional, narrative, awareness-focused — tests poorly in short attribution windows and gets suppressed. Over time, every ad in your account looks like a direct-response ad. Your brand voice disappears from paid media entirely. This convergence accelerates creative fatigue, as a narrower creative portfolio exhausts its novelty faster.

Corruption 3: Placement Arbitrage. The algorithm finds that certain placements convert better. These are almost always lower-funnel placements — search results pages, retargeting inventory, product listing ads. Upper-funnel placements — video, display on premium publishers, connected TV — show weaker short-term returns and get deprioritized. The algorithm moves your presence from places where brands are built to places where demand is captured.

Each of these corruptions is rational at the individual decision level. Each improves the targeted metric. And collectively, they produce a marketing program that is superbly efficient at converting existing demand and completely incapable of generating new demand.

This is overfitting in the machine learning sense. The algorithm has over-fitted to the characteristics of people who convert within the measurement window. It has no model of — and no incentive to learn about — the much larger population of people who might convert in the future if exposed to the brand today.

Econometric Evidence: The Brand Halo Effect

If brand effects are real, they should be measurable outside the narrow window of attribution models. They are. The field of marketing mix modeling (MMM) and econometric analysis has produced consistent evidence of what researchers call the "brand halo effect."

The brand halo effect refers to the incremental lift that brand investment provides to all other marketing channels. When brand awareness is high, every performance channel works better — higher click-through rates, better conversion rates, lower cost per acquisition. The brand acts as a force multiplier.

Ebiquity, one of the world's largest independent media investment analysts, published analysis in 2022 showing that brand advertising generates approximately 58% of total marketing-driven revenue when long-term effects are included, compared to just 18% when only short-term effects are measured. The gap — 40 percentage points — is the brand halo that standard attribution models miss entirely.

Analytic Partners' ROI Genome project, which has analyzed over $1 trillion in marketing spend across 750+ brands, found similar results. Their 2023 report showed that brands which cut spending on upper-funnel activity experienced a median 18% increase in cost-per-acquisition for lower-funnel activity within 12 months. The brand had been subsidizing performance marketing's efficiency, and when the subsidy disappeared, the true cost became visible.

The numbers paint a consistent picture. Every performance metric degrades substantially when brand investment is withdrawn. Paid search click-through rates drop by more than half. Social media acquisition costs nearly double. Display conversion rates fall by two-thirds. This is the hidden subsidy that performance marketing has been receiving from brand investment — and that disappears when the brand budget is cut.

One of the most telling findings comes from a 2021 Google and Ipsos study that tracked branded search volume as a function of brand media spend. They found a consistent elasticity: for every 10% reduction in brand media spend, branded search volume declined by approximately 6% within 12 months. Since branded search is the highest-converting, lowest-cost acquisition channel for most companies, this single effect can account for a significant portion of the CAC increase that follows brand cuts.

Short-Term vs Long-Term Revenue Modeling

The fundamental problem with how most companies evaluate marketing investment is the time horizon. Standard attribution models use 7-day, 14-day, or 28-day lookback windows. Some advanced implementations extend to 90 days. Almost none model beyond a year.

But brand effects operate on multi-year time horizons. The memory structures that brand advertising creates decay slowly — Ehrenberg-Bass Institute research suggests a half-life of approximately 2-4 years for well-established brand memories. A brand impression delivered today may influence a purchase decision three years from now. No attribution model will ever capture this.

When we extend the modeling horizon, the economics change dramatically.

Consider a simplified example. A company spends $100 on a performance campaign. Within 30 days, it generates$150 in revenue. ROAS: 1.5x. Good.

The same company spends $100 on a brand campaign. Within 30 days, it generates$40 in measurable revenue. ROAS: 0.4x. Bad, says the dashboard.

But extend the window. Over 12 months, that brand campaign generates an additional $60 in direct response (people who recall the brand and search for it later),$35 in reduced acquisition costs for performance campaigns (the halo effect), and $25 in price premium that would not exist without brand salience. Total 12-month return:$160. True ROAS: 1.6x.

Over 36 months, the brand campaign continues to compound. The memory structures it created generate an additional $80 in direct and indirect revenue. Total 36-month return:$240. True ROAS: 2.4x.

The performance campaign, by contrast, generated its $150 and then stopped. There is no compounding. No residual effect. The revenue was a one-time extraction, not an investment.

The crossover point — where cumulative brand returns exceed cumulative performance returns — occurs at approximately month 10-14, depending on the category and brand strength. Before that point, performance looks superior on every metric. After that point, brand looks superior on every metric.

Virtually every measurement system used in modern marketing is designed to read only the left side of this chart. That is not an accident. Attribution vendors, ad platforms, and performance agencies all have economic incentives to keep measurement windows short, because short windows make performance marketing look more efficient. The measurement system is itself a Goodhart victim.

Rebalancing: A Practical Architecture

Theory without implementation is decoration. Here is how we recommend companies rebalance their marketing investment to account for long-term brand effects.

Step 1: Measure what matters. Before changing any budget allocation, install measurement systems that capture brand effects. This means:

• Quarterly brand tracking studies (unaided awareness, consideration, preference, net promoter)
• Branded search volume tracking (indexed monthly)
• Share of search analysis (your brand's share of category search volume)
• Marketing mix modeling with at least 3 years of data
• Customer cohort analysis comparing acquisition costs by source over time

Step 2: Establish a brand investment floor. Based on Binet and Field's research and the econometric evidence, we recommend a minimum brand investment floor of 40% of total media spend. For companies currently at 80/20 performance-to-brand, this is a dramatic shift. Do not make it overnight. Phase it over 12-18 months. A reasonable trajectory:

• Quarter 1: Move from 80/20 to 70/30
• Quarter 2-3: Move from 70/30 to 60/40
• Quarter 4-6: Stabilize at 60/40 and measure

Step 3: Protect brand budget from quarterly optimization pressure. This is the hardest step. Brand budget will always look "inefficient" in short-term dashboards. The CFO will always want to reallocate it. The performance team will always claim they could use it better. You must treat brand investment the way you treat R&D — as a long-term investment that is evaluated on a long-term horizon, not a quarterly one.

Step 4: Reconfigure algorithm constraints. Within your performance channels, impose guardrails on automated bidding:

• Set minimum frequency caps on new audience segments (the algorithm will want to avoid these; force it)
• Maintain minimum creative diversity requirements (prevent convergence to pure direct-response)
• Allocate at least 20% of performance budget to prospecting campaigns with separate KPIs
• Use reach-based objectives for a portion of paid social (not just conversion objectives)

Step 5: Implement a dual-horizon reporting framework. Report two sets of numbers to leadership:

• Short-term (30-day) performance metrics: CPA, ROAS, conversion rate. These satisfy the quarterly accountability mandate.
• Long-term (12-month rolling) brand health metrics: unaided awareness, share of search, brand consideration, cohort CAC trends. These track whether you are building or depleting the asset.

The Uncomfortable Math

Let us close with a calculation most marketing teams never perform.

Imagine a company spending $10 million annually on marketing, allocated 80/20 to performance/brand. Performance CAC is$50. They acquire 160,000 customers per year. Brand awareness sits at 45%.

The CFO proposes cutting brand entirely and reallocating to performance. Year 1 results: the extra $2 million in performance spending, at$50 CAC, acquires 40,000 additional customers. Total: 200,000. Revenue jumps. The CFO is promoted.

Year 2: Brand equity has decayed. Unaided awareness drops to 38%. Branded search declines 15%. Performance CAC rises to $62 because ads are less effective without the brand halo. The$10 million in performance budget now acquires 161,000 customers. Fewer than the original balanced approach.

Year 3: Awareness at 31%. CAC at $81. Customers acquired: 123,000. The company has lost 37,000 customers per year compared to the original balanced scenario — and has no brand equity left to support a recovery.

Total customers acquired over three years:

• Balanced (60/40): approximately 495,000
• All-performance: approximately 484,000

The all-performance approach acquires fewer total customers over three years — despite spending 100% of budget on acquisition. And it leaves the company with a depleted brand that will take years and millions to rebuild.

This is the hidden cost of over-optimization. It is not visible in any single quarter. It is not captured by any standard attribution model. It compounds silently, like interest on a debt you didn't know you had. And by the time it appears in the numbers, the damage is measured in years, not quarters.

Goodhart warned us in 1975. Binet and Field showed us the data in 2013. Adidas confessed in 2019. The evidence is not ambiguous. The only question is whether we will act on it before the algorithm finishes eating the brand — or after.

We know which choice most companies will make. The algorithm is too convincing. The dashboard is too clean. The quarter is too short.

That is the trap.

---

References

1. Goodhart, C.A.E. (1975). "Problems of Monetary Management: The UK Experience." Papers in Monetary Economics, Reserve Bank of Australia.

2. Binet, L. & Field, P. (2013). The Long and the Short of It: Balancing Short and Long-Term Marketing Strategies. IPA Publications.

3. Binet, L. & Field, P. (2017). Effectiveness in Context. IPA Publications.

4. Sharp, B. (2010). How Brands Grow. Oxford University Press.

5. Peel, S. (2019). Presentation at the Effie Awards, October 2019. Reported in Marketing Week, "Adidas: We over-invested in digital advertising."

6. Ebiquity (2022). "Re-evaluating Media: What the Evidence Reveals About the True Worth of Media in the Customer Journey."

7. Analytic Partners (2023). ROI Genome Intelligence Report: The Brand Building Imperative.

8. Google & Ipsos (2021). "Search and Brand: Understanding the Relationship Between Brand Awareness and Branded Search."

9. Kantar Millward Brown (2022). BrandZ Global Top 100 Most Valuable Brands Report.

10. Romaniuk, J. & Sharp, B. (2022). How Brands Grow Part 2. Oxford University Press.

11. Ehrenberg-Bass Institute (2020). "Advertising's Effects on Sales: Decay Rates and Long-Term Impact." Working Paper.

12. Nielsen (2023). "Marketing Mix Modeling Meta-Analysis: Cross-Industry Brand Effectiveness Benchmarks."

13. Field, P. (2020). "The Crisis in Creative Effectiveness." IPA Publications.

14. Ritson, M. (2019). "Adidas proves the long and short of effectiveness theory." Marketing Week, October 2019.

15. Taleb, N.N. (2018). Skin in the Game. Random House. Chapter on the Lindy Effect and survivorship.