Happy Sunday, Everyone!

After a whirlwind spring conference season, I’m (finally) back for a few weeks. This past week, we rolled out a series of new initiatives to our team – and I found myself explaining Jevons’ Paradox. Afterwards, I saw many people found it helpful in thinking about the impacts of AI on marketing. This inspired me to start thinking about what other paradoxes, razors and mental models I use in my day-to-day, as well as how those might be helpful to you all.

To be brutally honest: marketing teams have never had more platforms, more data, more software, or more expectations. The result is a steady drumbeat of tactical noise: new channels, new formats, new optimizations, new tasks, new reports – all positioned as the one “true” path to growth. If you’ve worked in marketing for any length of time, you know what I’m talking about all too well.

This truth leads to a second one: what separates marketers + operators who actually drive real results from those who spin their wheels is not a magic mar-tech stack or a shiny new tool - it’s pattern recognition. The ability to spot a problematic system dynamic, recalibrate and respond before it wrecks the (initiative, quarter, campaign, whatever) is the closest thing remarkable marketers have to unfair advantage.

The 11 ideas that follow - part paradox, part razor, all proven over years - function as a pattern library. Each one opens with a quick primer, then moves to the specific ways each system dynamic hijacks marketing programs, and closes with a set of actionable recommendations.

This is a slightly different issue than most - it’s intended to be a bookmarked resource, not a read-it-once-and-never-again article. So, skim it once for context; revisit them whenever a metric or campaign moves in ways that do not make sense on the surface. You will find that most surprises are not surprises at all. They are simply old rules, wearing new clothes.

Let’s get to it.

1. Braess’s Paradox

The origin: German mathematician Dietrich Braess noticed traffic slowed after engineers added a new highway. Drivers, pursuing individual shortest routes, piled into the “shortcut.” Bottlenecks shifted, congestion worsened, and the entire network moved slower than before the massively-expensive “improvement.”

Why it haunts marketers:

Channel sprawl is our highway. Agencies (or in-house marketers) eager to scale add Reddit, CTV, and an affiliate network onto a Meta‑Google core. Each launch makes sense in isolation, but collective side effects creep in: fragmenting creative resources, obscuring impact, and hindering channel-specific mastery. The end result is a brand spending more than ever to reach the same people, while marketers are heads-down in production & analysts drown in attribution fog. The org congratulates itself on going “omnichannel”.... right up until the CEO/CFO observes CAC trend lines trending up, incrementality faltering and marketing spend going up-and-to-the-right.

What Braess’s Paradox looks like in the wild:

A DTC apparel brand layered Reddit, then Pinterest, then connected TV on top of Meta and Google over the course of ~6 months. Blended CPM rose 15% while CAC ticked up 24%. A strategic retreat to three channels plus heavier creative testing pushed CAC back below pre‑expansion levels within four months.

Early warning signs:

• Performance improves in no single channel even though blended spend rises.
  ​
• Teams complain about bandwidth for versioning and reporting.
  ​
• Attribution confidence drops because touch points multiply.

How To Respond:

1. Incremental reach audit: Pull overlap matrices; if a channel adds <5 % unique impressions and < portfolio ROAS, cut or consolidate.
   ​
2. One‑in‑one‑out rule: Any new media lane requires pausing the lowest decile channel for 30 days, forcing a real control test.
   ​
3. Depth before width: Build 5 net‑new concepts inside proven channels before proposing a brand‑new placement. Teams surface scale hiding in plain sight.

2. Simpson’s Paradox

The origin: In the early 1950s, statistician Edward H. Simpson showed that a trend appearing in an aggregated dataset can vanish (or even reverse) once you split the data into meaningful sub-groups. Classic example: a medical treatment that seems worse overall but saves lives in every single age bracket. The aggregate masks the truth.

Why it haunts marketers:

Marketers live on blended dashboards: “paid search ROAS is down” or “email is flat.” But purchase intent, device mix, and geography are rarely uniform. A slump in one sufficiently-large segment will drown out hidden pockets of over-performance. If you make a change based on the aggregate before understanding the underlying distribution, you risk torching the golden sub-cohort that’s producing outsized results.

What Simpson’s Paradox looks like in the wild:

A SaaS firm saw their conversion rate fall from 3.4% to 2.9% following a site-wide design update. Panic ensued. The CMO wanted to immediately pull back spend + revert website changes. A cohort cut told the real story: CVR on desktop was up 30%, but a site-speed bug on Android reduced mobile CVR to 1.% (yes, slow sites do kill conversion rate). Fixing the Android issue sent overall CVR to 3.7%.

Early warning signs:

• A single blended KPI drops while no individual campaign manager can “feel” the decline.
  ​
• Performance swings coincide with a new geo-target, creative, or device update.
  ​
• Stakeholders debate whether to kill a channel without agreeing on the segmentation that matters.

How to respond:

1. Mandatory cohort hygiene: Every weekly report surfaces at least two primary (device, geo) and one secondary (new vs. repeat) splits.
   ​
2. Variance flagging: Trigger an alert whenever aggregate KPIs move ±5 % but no cohort shifts in the same direction, which is a classic Simpson tell.
   ​
3. Segmented budgeting: Allocate budgets by high-performing sub-groups first; let laggard cohorts prove they deserve incremental dollars rather than hiding in the roll-up.

3. Goodhart’s Law

The origin: In 1975, economist Charles Goodhart warned the Bank of England that once a specific monetary indicator (“M3 money supply,” in that era) became an official target, banks would rearrange their books to hit it—rendering the metric useless as a signal. Anthropologist Marilyn Strathern later distilled the rule into its modern punchline: “When a measure becomes a target, it ceases to be a good measure.” The idea now sits at the center of everything from AI alignment to school-testing policy.

Why it haunts marketers:

Digital platforms are giant incentive engines. Tell Meta to maximize “purchases” and the algorithm exploits low-value flash-sale buyers; ask your email team for “opens” and they crank click-bait subject lines that drive plenty of opens (and plenty of unsubscribes); bonus an agency on “cost per install” and it floods the funnel with junk traffic from lower-quality ad networks. At first the KPI graph looks heroic, but downstream revenue, brand equity, or customer experience quietly erode. By the time the spreadsheet exposes the gap, you’re months - and potentially millions - behind forecast.

What Goodhart’s Law looks like in the wild:

Wells Fargo + the Account Opening Fraud. Throughout the 2000s, Wells Fargo tied branch-staff bonuses to the number of new accounts opened per customer. The metric (“products per household”) became the target - exactly as Goodhart warned. To hit impossible quotas, thousands of employees began creating CC, savings & deposit accounts without customer consent. By 2016 the practice had ballooned to more than two million phony accounts, triggering $185 million in fines, mass firings, and a $3 billion federal settlement. The headline KPI looked stellar on quarterly slides, yet the underlying behaviour torched shareholder value and brand trust (which Wells Fargo still has not fully recovered).

App-install inflation: A relatively well-known app celebrated a CPI drop from $11 to $4 after switching agencies and pivoting strategy. Several weeks later, day-30 retention crashed from 38% to 14%. Investigation revealed ads were running on low-quality reward apps. The “cheap” installs cost the company millions in wasted promo credits, not to mention entire cohorts of low-quality customers that failed to ever reach baseline LTV targets.

Early warning signs:

• One headline metric spikes while customer LTV, NPS, or gross margin drifts the opposite direction.
  ​
• Channel managers defend performance with platform screenshots yet can’t map wins to bankable cash flow.
  ​
• Customer-service tickets rise or brand search sentiment falls during a period of “record-setting” KPI gains.

How to respond:

1. Outcome ladder: Define a hierarchy: micro-conversion →pre-purchase indicator → purchase → 90-day LTV. Only automate on a lower-funnel proxy if its historical ratio to the final rung is stable. This happens all the time when Meta media buyers try to “scalp” cheap traffic, based on the underlying assumption that all clicks from Meta are equal. That’s rarely the case, and it ends up resulting in way too much budget being wasted.
   ​
2. Balancing metrics: Couple every target to a counter-weight: CPA with payback period, email opens with unsubscribe rate, CTR with bounce rate. Gaming one trips the other’s alarm. When I talk about Metric Duos + Trios, that’s why.
   ​
3. Metric rotation & review: Hold annual “KPI review” meetings that force teams to justify why each metric is still predictive of the desired business outcome. Before adding more KPIs to the database, have teams remove at least 50% of the ones being added (example: if the team wants to add 4 KPIs, then sunset 2 old KPIs).
   ​
4. Incentive alignment Link agency fees, internal bonuses, and platform optimization to incremental revenue or profit, never raw platform numbers
   ​
5. Lift & causal testing Strip a random 5-10% of audiences from the optimized tactic; if incremental lift is <20% of reported conversions, you’ve got a Goodhart casualty

Goodhart’s Law isn’t an argument against measurement - it’s a warning to respect the difference between a dashboard and reality. Keep the metric honest, or it will lie to you with perfect precision.

4. The Cobra Effect

The origin: Way back in colonial Delhi, British officials sought to address a growing problem: deaths due to cobra bites. Their solution was relatively simple, if (ultimately) perverse: the officials would pay a bounty for each dead cobra. Well (unsurprisingly), enterprising locals began breeding cobras in backyard pits, neatly collecting payments until the government scrapped the program. With revenue gone, breeders freed their “inventory” into the streets, which ended up raising, not lowering, the snake population. Economists now brand any policy that backfires through perverse incentives the Cobra Effect.

Why it haunts marketers:

Marketing runs on bounties: affiliate commissions, coupon codes, partner SPIFFs, influencer payouts, agency bonuses tied to a single metric. When the reward structure ignores true incrementality, partners quickly optimize for the bounty, not the business. Affiliates poach branded search traffic, coupon extensions pop offers at checkout, and agencies stuff cheap, low-quality impressions to hit CPM or CPI targets. Spend rises, margins thin, but topline growth stalls, leaving the brand owners/executives wondering why the “performance” budget behaves like a cost center.

What the Cobra Effect looks like in the wild:

Coupon cannibalization. A well-known wellness retailer opened a 15% affiliate commission on “first-time purchases.” Coupon plug-ins like Honey + Capital One Shopping surfaced a public “WELCOME15” code, cookie-stuffed at checkout, and claimed credit for users already in the CRM. Affiliate order share jumped 60%, but true new-customer revenue rose by less than 10% of that total (~5.5%). The brand was paying for customers who were going to purchase anyway, with the discount plugins simply sniping credit from other channels like Google, Meta + Organic.
​
​Leads without lives. This can also happen inside any sales booking organization (i.e. a mortgage company or dental/implant company), where an end-user (i.e., the mortgage originator or the dental office) is paying per lead. While this setup sounds logical, what (almost inevitably) happens is the intermediary booking calls from low-quality, low-propensity leads in order to hit quota/secure a payout, despite the fact that many of the calls booked have no intention of ever buying. This costs the end company twice: the first payout to the intermediary for the call, and the second (and more nefarious) because they end up over-staffed in the call center.

Early warning signs:

• Spend spikes in a pay-for-performance channel while blended revenue flat-lines.
  ​
• Surge in low-intent traffic: coupon, TM+coupon searches, or sub-7-second session durations.
  ​
• LTV of bounty-sourced customers trails organic by double digits.

How to respond:

1. Incrementality testing: Geo-split or time-boxed holdouts reveal true lift; axe partners under 1.2× incremental ROAS.
   ​
2. Smart commission tiers: Pay higher rates for new-to-brand conversions and scale steps with downstream LTV, not top-line gross revenue.
   ​
3. Cookie hygiene & negative scrubbing: Block payouts on last-click hijacks (brand + “coupon”) and scrub orders with leaked codes.
   ​
4. Quality gates: For lead programs, require secondary validation (credit check, double-opt-in, validation of intent) before triggering a payout. Monitor performance by booking channel, and if you see one intermediary with an outlier share of low-quality/junk leads, aggressively remove or remediate.

The bottom line: reward the outcome you actually want, measure the delta it produces, and audit ruthlessly. Otherwise, you’ll find yourself surrounded by well-fed cobras, wondering why the antidote costs a fortune.

5. Abilene Paradox

The origin: Management thinker Jerry B. Harvey coined the Abilene Paradox after a blistering 1974 Texas afternoon when his family collectively decided to drive 53 miles to Abilene for dinner, despite the fact that none of them actually wanted to go. Each person assumed the others desired the trip, so everyone consented to avoid rocking the boat. On the ride home, they discovered their unanimous “agreement” had been a chain of silent misreads. Lesson: group discomfort with open dissent can push teams toward options that nobody, individually, believes are good.

Why it haunts marketers:

Creative reviews, product decisions, channel selection, website re-designs, you name it - marketing is riddled with decisions that invite strong opinions but punish outliers who speak up. Agencies fear ruffling the CMO, junior analysts defer to “HiPPOs” (highest-paid person’s opinion), brand managers avoid friction with peers. The result is bland creative, half-measure media allocations, or watered-down positioning that satisfies no one and underperforms everyone. Because the failure is collective, no single stakeholder feels accountable, yet the business bears the cost.

We’ve all been in meetings where the Abilene Paradox shows up - everyone silently nodding along as (whoever) presents whatever. We’ve all sat through brainstorms where someone suggests something relatively inane + safe, and everyone else just agrees along – despite knowing that the idea is - at best - mediocre. If you want further proof that the Abilene paradox is real, look no further than Jaguar.

What the Abilene Paradox looks like in the wild:

A B2B SaaS company convened a relatively large group of stakeholders to choose a new homepage. The agency presented three versions: a bold, edgy message; a data-heavy proof block; and a “safe” cloud-blue stock imagery-plus-icons layout. Silence followed the reveal. Believing the silence signaled disapproval of the riskier options, the creative lead defaulted to the safe option. 6 weeks post-launch, CTR fell 28%, sales pipeline thinned, revenue started to fall – all while the sales team lead (who was in the meeting) complained the brand “looked like everyone else.”

Early warning signs:

• Meetings end in lightning-fast consensus and light discussion.
  ​
• Ideas with strong pros and cons receive no clarifying questions.
  ​
• Post-launch metrics underperform while internal chatter includes “We all knew this probably wouldn’t work.”

How to respond:

1. Red-team rotation: Designate one attendee per meeting to argue against every proposal, ensuring dissent is procedural, not personal. This mirrors the famous Israeli “10th man” approach, where, if 9 individuals all see the same data and come to the same conclusion, the 10th is duty-bound to disagree, no matter how unlikely, awkward or politically inconvenient.
   ​
2. Anonymous pre-vote: Collect blind “lose / iterate / approve” ratings on each deliverable/campaign/design – then share the aggregate feedback before beginning discussion. This provides individuals with some comfort/anonymity, but lays bare the “real” feelings on each concept before groupthink can take hold. It’s far more difficult to say that “everyone agrees” on Design 1 if half of the people have rated it “lose.”
   ​
3. Decider documentation: Write a brief after-action memo capturing objections, rationale, and owner sign-off. If results falter, you can go back and understand why the decision was made, identify the mistake (if there was one - it might just be bad luck!) and learn from it.
   ​
4. Psychological safety cues: Leaders model candor by openly challenging their own ideas first; the room then follows suit.

Applied rigorously, these habits surface true preferences early, prevent “polite mediocrity,” and keep your marketing bus headed somewhere everybody genuinely wants to go, profit included.

6. Chesterton’s Fence

The origin: In a 1929 essay, G. K. Chesterton imagines two reformers stumbling on a random fence in a field. The first insists on tearing it down; the wiser second replies: “If you don’t see the use of it, I certainly won’t let you clear it away. Go away and think. Then, when you can tell me what purpose the fence served, I might allow you to destroy it.” The parable warns that inherited constraints often solve problems we no longer notice—remove them blindly and the original threat returns with interest.

Why it haunts marketers:

CMOs rotate, agencies churn, tech stacks are updated. Each new leader is eager to purge “old clutter”, do something new + prove their value to the brand. Yet many marketing fences - an oddly timed promotional email, a legacy print drop, a clunky but gated signup flow - exist because they have some utility - whether that’s defending a fragile revenue stream or constraining a downstream cost (like sales resources). The problem is most of these fences are ugly, clunky or straight-up odd – so they seem like low-hanging fruit ripe for the picking. In reality, the opposite is true: removing the fence ends up costing far more, either in terms of lost sales, increased demands on sales resources, declining lead quality, you name it. Sometimes the old, ugly-looking thing actually serves a quite valuable purpose.

What Chesterton's Paradox looks like in the wild:

There are many examples of this, but my favorite one is a law firm with a downright hideous site. Each of the service pages were excessively long (we’re talking 5,000+ words), poorly designed, and pretty much broke every single “rule” of good marketing + good design. Despite all this, the firm was (and still is) quite successful - recovering hundreds of millions for clients each year. The new marketing director came in, “modernized” the site, and the partners quickly identified that both conversion rate AND case value were plummeting. It was only after some of the staff attorneys showed the “new” site to current clients that they identified the problem: those clients all viewed the “old” site as authentic, and they interpreted the firm’s lack of fancy marketing (relative to competitors) as an indicator that the firm prioritized client success over bells-and-whistles. In this case, the fence (ugly as it was) was what drove high-value, complex and highly desirable plaintiffs to retain this firm over many others (and, in particular, the others with far nicer sites).

Early warning signs:

• A cost-cut “quick win” coincides with an unexplained dip in retention or average order value.
  ​
• Veteran team members voice vague discomfort: “I’m not sure we should scrap that yet” but lack hard data. As Jeff Bezos has famously said, “When the data and the anecdotes disagree, go with the anecdotes.”
  ​
• New initiatives generate more work - more tickets, more workarounds, more sales resources, more returns - than the process they replaced.

How to respond:

1. Purpose interviews: Talk to the architects before sunsetting any long-standing tactic. Capture the original KPI it shielded.
   ​
2. Data archaeology: Pull historical dashboards; time-shift the KPI trend to confirm the fence’s cause-and-effect.
   ​
3. Parallel test, never hard stop: Run a 50/50 holdout for at least one (ideally, 2-3) sales cycles. If removal hurts net revenue or NPS, reinstate until a validated replacement exists.
   ​
4. Institutional memory Log every retired fence with its rationale, metrics, and test results in a living notion (or other repository), so the next regime doesn’t repeat your tuition bill.

7. Brandolini’s Law

The origin: On 11 January 2013, Italian software developer Alberto Brandolini fired off a now-famous tweet: “The amount of energy needed to refute bull**t is an order of magnitude bigger than to produce it.”* It formalized what journalists, scientists, and policymakers had long intuited: once a dubious claim slips into public discourse, debunking it demands exponentially more time, money, and attention than the seconds it took to create. Online virality and algorithmic echo chambers amplify the imbalance, giving bad data a head start your comms team may never fully close.

Why it haunts marketers:

Marketing thrives on bold promises, zero-to-hero case studies, and pithy stat lines. In the haste to ship a deck, a landing page, or a founder’s LinkedIn thread, teams round numbers, drop context, or lean on “directionally correct” anecdotes. The claim sails through creative, yet if it lands wrong (with regulators, fact-checking reporters, or savvy consumers), the blowback diverts entire quarters of budget and focus to damage control. Worse, the corrections rarely travel as far as the original exaggeration, so perception lags reality long after the lawyers sign off.

What Brandolini’s Law looks like in the wild:

Is it Really Tuna? A 2021 class-action lawsuit - along with an extensive writeup in The Washington Post - claimed lab tests found “no identifiable tuna DNA” in Subway’s sandwiches. Social and late-night media pounced, and within 48 hours the meme was global. Subway commissioned multiple accredited labs, rolled out a national TV rebuttal, and fought the suit for more than two years, running up hundreds of thousands in legal and PR costs. The case was ultimately dismissed with prejudice in 2023, but Google still auto-suggests “Is Subway tuna real?”.

Nutella Does NOT Cause Cancer. In 2016, an EFSA toxicology note about potential carcinogens in improperly refined palm oil became an Italian tabloid headline: “Nutella could give you cancer.” The story quickly went viral across FB and IG (particularly in Europe). Ferrero (makers of Nutella) responded with a multi-market blitz - prime-time TV spots, full-page newspaper ads, and scientific explainers - detailing its lower-temperature process and certified supply chain. Sales eventually rebounded, yet every January searches for “Nutella cancer” spike anew, proving that debunking misinformation demands orders of magnitude more effort than the seconds it takes to publish it.

Early warning signs:

• “We’ll clean up the footnotes later - just get the visuals approved.”
  ​
• Edits add certainty (“clinically proven”) without adding citations.
  ​
• Social replies ask for sources and receive silence or deflection.

How to respond:

1. Two-gate fact check: No external claim exists without (a) primary source citation on file and (b) legal and/or SME sign-off
   ​
2. Evidence library: Maintain a living database of validated stats, study PDFs, and permissions. Creatives pull from the library instead of Googling under deadline.
   ​
3. Myth-bust content hub: Publish long-form FAQs and whitepapers proactively; they become instant ammo when challengers emerge
   ​
4. Crisis rehearsal: Misinformation storms are crisis situations. The easiest solution is to treat them as such, and prepare accordingly. Identify major weak points (check out this article on Crisis Comms) or vulnerabilities, develop responses/messages, and ensure you have ready access to any primary source material necessary to refute problematic claims.

Brandolini’s Law reminds us that the cheapest word in a headline can balloon into the costliest line item in your budget. Precision and accuracy up front are not inherently bad or bureaucratic - in some cases, they’re the single-best ROAS investment you can make.

8. Streisand Effect

The origin: In 2003, Barbra Streisand sued an aerial-photography archive for posting a single frame of her Malibu estate. The image had logged only six downloads (and 2 of them from her own lawyers) before the lawsuit. News of the takedown attempt rocketed across the web, driving more than a million views in about a month, being reprinted countless times across a variety of publications. The incident crystallised a counter-intuitive truth first hinted at by political theorists like John Stuart Mill: attempts to silence information can amplify it instead. Behavioral economists later mapped the mechanism to reactance theory: people’s hard-wired pushback when they feel a freedom (to know, share, decide) is being curtailed.

Why it haunts marketers:

Brands run on perception. A single negative TikTok, an unflattering Glassdoor post, or a leaked price hike can feel existential - especially to owners/founders, legal or investor relations. The knee-jerk reaction: enforce a takedown, issue a cease-and-desist, bury the term with paid search.

The result is often textbook Streisand Effect: the controversy trends, online personalities and/or media outlets sniff a censorship angle, a few articles/posts are written, and Google’s “Top Stories” connects the critique to your brand keyword, and social sentiment charts resemble a cardiac event.

What the Streisand Effect looks like in the wild:

One of the better examples of the Streisand Effect concerns the LAPD and the Cola Corporation, way back in 2024 (yes, it’s been an eternity). The LAPD Foundation attempted to claim copyright of the letters “LAPD” after the Cola Corporation launched a shirt with “F*** the LAPD.” The claim wasn’t successful, and the shirts benefited from a truly insane amount of free publicity – all at the expense of the LAPD.

Early warning signs:

• An issue lives on small, siloed platforms but spikes each time the brand engages.
  ​
• Social monitoring shows neutral or low-reach chatter, yet legal or exec channels demand zero-tolerance removal.
  ​
• Earned-media reporters follow the brand’s takedown filings more closely than the original content.

How to respond:

1. Harm-Benefit Grid Map Incremental Reach × Incremental Harm. If potential amplification dwarfs the damage, pursue engagement over erasure.
   ​
2. Transparent Rebuttal Correct factual errors in public comments, link primary evidence, invite dialogue. You satisfy curious observers and starve the outrage economy.
   ​
3. Own the Narrative If the criticism is fair, acknowledge and outline fixes—turning a liability into proof of customer obsession. If it’s satire, lean in with self-aware humour; Colbert-bump beats Streisand-spike.
   ​
4. SEO & Content Buffer Publish authoritative explainers that rank organically; they dilute sensational headlines without triggering censorship alarms.
   ​
5. Crisis Fire-Drills Quarterly simulations pair PR, legal, and social teams to rehearse “ignore vs. engage vs. litigate” decision trees—clarity now prevents panic later.

Streisand reminds us that in the attention economy, force multiplies friction. Smart operators measure volume, velocity, and veracity before swinging the legal hammer. Sometimes the loudest silence is strategic transparency served at Internet speed.

9. Jevons’s Paradox

The origin: In 1865, British economist William Stanley Jevons observed that the steam engine’s soaring fuel efficiency didn’t lessen coal demand; it multiplied it. Cheaper input per horsepower made steam viable for more factories, mines, and railways, so overall coal consumption exploded. The lesson: lower unit cost can unlock entirely new demand curves, swamping the savings that sparked the surge.

Why it haunts marketers:

Every year, martech stacks promise to “do more with less”: AI varianting, one-click syndication, automated bidding. Production and distribution costs tumble, so marketers respond by unleashing volume - daily instead of weekly creatives, 20 A/B tests instead of a handful, 15 emails a month instead of 6. Audiences drown, auctions tighten, suppression lists crank up. The efficiency win flips into a reach-cost spiral that raises blended CPMs and erodes brand goodwill.

What Jevons’s Paradox looks like in the wild:

Meta Audience Expansion. A health snack brand with a well-defined, wildly niche product (super healthy, 5-ingredient treats) wanted to scale Meta. Their existing setup was fairly standard: a tight lookalike, a well-designed interest stack, and a remarketing ad set. The problem was reach - they were selling at/below targets, but were struggling to scale. Their agency’s solution was to turn on audience expansion. The immediate win was clear, as the broader audience unlocked more low-hanging fruit; the rebound was brutal, as the much larger audiences simply didn’t have the same buyer density as the well-honed initial audiences. The end result was higher spend, but lower conversion rates + a much higher CAC.

Generative creative A well-known ugly footwear retailer used AI copy and templates to mass-produce 1,500 Meta ads for a fraction of the cost of human-produced ads. When those ads were put in the account, CTR dropped while CPMs climbed. The underlying driver of those increases was simple: Meta was struggling to test that many creatives across that large of an audience, resulting in massive inefficiency. While quantity drives quality, there’s a happy medium (and 1,500+ new ads in a week isn’t it). The brand ended up restricting the number of assets added to the account to the best ~200, and found performance reverted to normal.

Early warning signs:

• Auction metrics inch up faster than post-campaign savings.
  ​
• Frequency or inbox volume rises while engagement drops.
  ​
• Marginal lift per additional creative, email, or keyword trends toward zero.

How to respond:

1. Throttle by marginal utility Set a hard ROI hurdle for each incremental impression or email batch; pause when lift < cost.
   ​
2. Diversify attention Reinvest some efficiency gains into under-fished channels—podcasts, partner newsletters—where auction density is low.
   ​
3. Govern creative volume Retire two assets for every ten new ones. This enforces natural selection and prevents frequency creep.
   ​
4. Feedback audits Quarterly cross-compare production savings to media cost changes and NPS; if CPM creep or brand fatigue outpaces savings, scale back.

Jevons reminds us that cheaper isn’t free. The smartest teams treat every efficiency gain as capital to be rationed, not an excuse to carpet-bomb audiences until the economics invert.

10. Paradox of Choice

The origin: Psychologist Barry Schwartz’s 2004 book The Paradox of Choice, back-stopped by Sheena Iyengar and Mark Lepper’s jam-table experiment (30 flavors vs. 6), showed that while people love variety in theory, too many options spike decision costs in practice.

Behavioral economics research since has captured the mechanism: choice overload elevates cognitive load, fuels decision fatigue (Baumeister), amplifies anticipated regret (Kahneman & Tversky), and nudges shoppers toward the status-quo (or abandonment) when mental transaction costs outweigh perceived utility.

Why it haunts marketers:

Growth teams equate assortment or feature depth with value creation: more SKUs, more ad versions, more pricing plans, more widget toggles, more options = better. Each addition looks harmless in isolation; collectively they crush both purchase intent and user satisfaction. The end result: slower decision times, higher abandonment rates and lower ROAS, as potential clients/customers weigh “what if” scenarios, revisit competitors, or browse other sites (i.e., forums, reviews, third-party write-ups). All the while, in-market campaigns keep spending - resulting in CAC creeping up because the bottleneck isn’t awareness - it’s too many options.

What the Paradox of Choice looks like in the wild:

• DTC apparel site launches 64 color/size variants on one hoodie. PDP bounce rate rises 18%; heat maps show cursor flutters over swatch grid. Revenue crumbles as users stop buying - there’s simply too many options, and they have no idea which to choose.
  ​
• SaaS tool offers “Starter, Growth, Pro, Premium, Enterprise, Custom.” Inbound SDR calls triple (“Which plan fits?”), yet self-serve conversions drop 27%. The end result - sales team busier than ever, but the time-to-revenue increases (self-serve is often the fastest path to revenue, and is often the most profitable).
  ​
• Email team packs three CTAs - a webinar registration, free trial & resource download into every newsletter. Click-to-open rate falls from 11% to 6%; heat map reveals no single link tops 2%.
  ​
• A PPC manager for a large brand continually adds new KWs to an account, resulting in upwards of 30 campaigns + 5,000 active KWs. Budget fragments, QS slips, and average CPC rises 12% versus a curated 800-KW account.

Early warning signs:

• High dwell time + high exit on decision pages.
  ​
• Customer-service chats start with “I’m confused about…”
  ​
• A/B tests where stripped-down variants beat richer ones despite identical pricing.

How to respond:

1. Assortment pruning: Adopt a management consulting style approach to your on-site/in-app options, SKUs, packages, etc.: remove the bottom 10% or 20% once per year. Option bloat slowly kills growth, so stay ahead of it by methodically pruning what isn’t performing.
   ​
2. Default nudges: Highlight a “Best for Most” plan; pre-select the popular size/finish to leverage default bias.
   ​
3. Guided selling: Quizzes, configurators, or comparison sliders translate complexity into single-path decisions, cutting cognitive load by an order of magnitude and leveraging the IKEA effect (i.e., the user has “built” their package/cart/assortment, so it appears more valuable to them).
   ​
4. Decoy restraint: One decoy (good-better-best anchor) boosts clarity; five decoys simply result in overload. Keep pricing options simple and streamlined: one option you want most to buy, one option you’re OK with some buying, and one option that no-one in their right mind should ever include. Anything else should be a call with sales.
   ​
5. Progressive disclosure: Surface essential info first: price, core benefit, differentiating features, then reveal specs on scroll or click, mirroring Fogg’s simplicity principle.

Choice feels like freedom; unmanaged, it becomes friction. It’s our role as marketers to curate, so every additional option/variant/package must support that goal, not invite more paralysis.

11. Tilt

The origin: In high-stakes poker, “tilt” describes the moment a player, rattled by a bad beat, starts playing like a rookie - chasing long shots, ignoring odds, burning bankroll. Neuroscience calls it amygdala hijack: stress hormones flood the prefrontal cortex, shrinking working memory and amplifying risk-seeking. Behavioral economists spot the cocktail of loss aversion (Kahneman & Tversky), sunk-cost fallacy, and overconfidence bias that follows. Whatever the label, the pattern is identical: the player is all emotion and no logic. Risk-reward goes out the window, and budgets usually get busted.

Why it haunts marketers:

Marketing is poker with bigger antes and more spectators. A CAC spike, a social-media roast, a CFO email at 1 a.m. - any hit to ego or targets can trigger tilt. We’ve all observed PPC managers removing cost/bid caps “to get some wins,” or the CMO green-lighting a set of unvetted TikTok creators, or the CEO authorizing a last-minute major discount in response to a competitor’s latest sale (or latest reported numbers). Dashboards turn crimson, agencies scramble, and the next meeting call feels like an interrogation, with everyone trying to figure out why everything went the wrong way. Ultimately, the P&L doesn’t care about the “why” - it only records the losses.

What Tilt looks like in the wild:

• Budgets on fire: A travel startup missed Q1 bookings by 12%. The CMO doubled Google budgets, forcing the PPC team to adjust caps up in order to spend (there’s not a sufficient volume of available customers at the current targets to absorb a 100% more budget). The end result was more customers acquired, but the CAC on those incremental customers was so high that acquiring them made little financial sense – the brand would’ve been better off simply not spending.
  ​
• Creative whiplash: A viral tweet roasted a brand’s tagline. Within 24 hours the copy team replaced every headline site-wide, all without QA. The fallout from that was horrific - broken links were everywhere, dozens of pages had layout/formatting issues, and (ironically) users on X piled on, roasting the brand again for the poor UI/UX.
  ​
• Price-war spiral: One competitor launches a 15% flash sale. In response, your client’s CMO counters with 20%; the competitor responds with 25% + free shipping. All the while, both of you are eroding margins for a category that was growing just fine at full price. The only winners here were ego + consumers.

Early warning signs:

• Slack threads heavy on exclamation marks, light on data.
  ​
• KPI meetings where the solution vaults straight to spend or price changes with no hypothesis doc.
  ​
• Leaders invoking “We can’t afford to look weak” or “Let’s make them pay” more than customer insight.

How to respond:

1. Cool-down Period: Any spend or pricing move >20 % triggers a mandatory 48-hour pause plus a written hypothesis (data, expected lift, guardrails). Teams call it the “seatbelt clause.”
   ​
2. Tilt Breakers: Nominate two peers (ideally one quantitative and one more creative) who hold veto power for 24 hours on emotionally charged decisions. Their only job: ask, “Is this strategy or spite?”
   ​
3. Pre-mortem Journaling: Before reacting, write the worst-case post-mortem headline (“We torched $2 MM chasing vanity installs”). Seeing the carnage in print often restores rational throttle.
   ​
4. Stress-test Dashboards: Pair any real-time metric with a lagging health KPI: ROAS vs. contribution margin, share of voice vs. brand favorability. If the quick-fix plan helps one but harms the other, back to the whiteboard.
   ​
5. Culture of Candor: Leaders publicly own their past tilt moments - “I green-lit that $700k panic promo” and detail the fallout. Vulnerability normalizes pause and inquiry over reflex. The more you can be honest about your failures as a leader, the more your team will be proactive and honest with you.

Personal hygiene for decision-makers:

• Physiology first: Even five minutes of box breathing or a walk lowers cortisol, widening the choice set your brain can perceive.
  ​
• Decision journal: Log the context, mood, data, and stake of every major call. Patterns of tilt (late nights, certain stakeholders) surface fast.
  ​
• Micro-advisory board: Keep a text thread of two external operators who can kill your FOMO with blunt math whenever you feel “urgent.”

Tilt doesn’t just nuke campaigns - it erodes the trust capital that lets teams take smart risks in the first place. The fix isn’t superhuman stoicism; it’s engineered friction that slows the emotional grenade before it detonates budget, brand, and morale. Great marketers don’t avoid tilt; they build systems that catch it, cage it, and convert the chaos into disciplined, data-backed iteration.

Putting the Models to Work In Your Day-To-Day

These ideas become leverage only when they leave the page and enter your day-to-day workflows. As you review briefs, data or client challenges, force yourself to think about what paradoxes can explain the result. Challenge your team to do the same - there’s a lot of comfort in complacency, but ultimately moving beyond it is what leads to outsized rewards.

Before any material change, run the checklist: Are we adding a channel without headroom; are we trusting a metric ready to betray us; are we tearing down a fence we do not understand; are we gambling on tilt? The habit takes minutes, but can save client relationships, budgets and your mental health.

At the end of the day, marketing is not a guessing game. It’s systems thinking that wears creative clothes. Paradoxes, razors, and mental models are the tools that can help you navigate through the ambiguity, craziness and noise to find the right signal for your (or your client’s) brand. Follow it with rigor and curiosity, and most surprises turn into calculated risks. Ignore it, and even the best-designed tactics sink under the weight of unseen forces.

Choose pattern recognition. It compounds faster than any tool in your stack.

Until next week,

Cheers,

Sam

​

Loving The Digital Download?
Share this Newsletter with a friend by visiting my public feed.

​

​

Follow Me on my Socials

​​​

​

1700 South Road, Baltimore, MD 21209 | 410-367-2700
​Unsubscribe | Manage Preferences​