When I was 19 and in college, I got a phone call that my dad was in the hospital. He couldn't keep food down. His blood pressure was so high you'd be surprised he didn't have a heart attack.

When I got there, they were running everything. Blood pressure, A1c, fasting glucose, kidney function, lipid panel, weight. They checked his eyes. They checked his feet. It was a cluster of measurements that, taken together, told a story no single number could tell on its own.

He was diagnosed with type 2 diabetes.

There's a reason doctors look at symptoms as a system before they make a diagnosis. Blood pressure alone could mean a dozen things. A1c alone could mean a dozen more. But blood pressure and A1c and fasting glucose and kidney function, measured together and tracked over time, narrow the diagnosis to something actionable. And even then, doctors still test before they prescribe. They don't guess. They don't skip to treatment because the first number looked bad. Prescription without diagnosis is malpractice.

Your business works this way too. If you've been running one long enough, you'll recognize a similar pattern.

When your cost to acquire a new customer starts rising, it could be an audience problem, an offer problem, or a bigger external trend that has nothing to do with your current tactics. A single metric can't tell you which one. But if your costs are going up and your cancellations are too, now you might have a product problem or a competitor problem.

And if you're not tracking the impact on contribution margin, you're going to get a stern chat with your CFO in the same tone my dad's GP gave him at his six-month checkup. Profitability can be down even when marketing is reporting higher ROAS than ever. Both can be true at the same time. That's the problem. Your metrics need to track profitability after all expenses are paid.

This happens with tech solutions and operations too. Digital transformation initiatives, martech implementations, platform migrations. They get purchased and configured at significant cost but end up with low adoption. Automation speeds up a process but results don't improve because the bottleneck is somewhere else. The symptom looks different in every department. The structural cause is the same: one metric, tracked in isolation, treated as the whole story. And a guess with a $3 million budget behind it is expensive malpractice.

In 2000, Harvard Business Review published a now-famous line: "The brutal fact is that about 70% of all change initiatives fail."¹ McKinsey's own transformation surveys back that up from the other direction: across 2012, 2014, 2016, and 2018, no more than 30% of transformations succeeded, and in their most recent survey, only 16% did.² And in 2025, MIT's NANDA initiative published findings that 95% of generative AI pilots delivered zero measurable impact on the P&L, despite $30 to $40 billion in enterprise investment.³

The failure rate is accelerating. And the reason is the same one that makes a bad doctor dangerous: single-symptom diagnosis leading to premature treatment. Teams build in isolation because that's how the work is divided. Each one measures the metrics assigned to its own function. And initiatives scale before the teams understand how their metrics impact the ones outside their dashboards.

Here's a case your team should study. These mistakes are more common than you think.

At Signs.com, the team believed 30% of their customers were repeat buyers. That number had been sitting in their reports for years. It came from Google Analytics (Google's web analytics platform, which tracks website visitor behavior), which was looking at a rolling month-long window of returning visitors.

Based on that assumption, they built an entire growth strategy around personal projects and gifts. They found new vendors. Designed new landing pages. Ran competitive pricing research. Launched 20 new products targeting people who were buying for personal projects and gifts, figuring that if 30% were coming back, even losing a little on the first purchase would pay off on the second.

It consumed the entire bandwidth of their marketing team for the better part of a year.

But here's the kicker: that 30% repeat purchase rate was wrong.

When we looked at past purchase data, the all-time repeat purchase rate was 8%. Google Analytics was measuring return visits in a 30-day window. The actual purchase history told a different story. At 22 percentage points, that gap isn't small. And most CRMs wouldn't have caught it either, because they aggregate single-point metrics the same way.

They'll tell you averages. They won't tell you that the average blends together a customer who buys every month and a customer who bought three years ago and never came back. Every forecast they had built on top of that number relied on false assumptions about acquisition costs, repeat customer behavior, and even the average order size. The math all looked great on paper. Then December 2017 hit and they posted a $100,000 loss on their P&L.

Those 20 product launches added almost no new revenue that year. Most of them lost time and money. And this was a company with enough search engine credibility to rank number one on Google for virtually any product they launched. Ranking first didn't save them because the underlying assumption about who was buying and why was wrong.

When we segmented the customers, the picture changed completely. The 92% who had only purchased once were those personal buyers and gift shoppers. The 8% who kept coming back were franchise owners buying signage every single month. The entire growth strategy had been built around the segment that would never return, while the segment that was actually driving repeat revenue had been invisible in the averages.

Only 17% of senior marketers say their firm tracks lifetime value well.⁴

Every business has a customer lifetime. Every business needs to understand what that lifetime is worth and how long it takes to recoup the cost of acquiring that customer. But customer lifetime value isn't the only place this thinking applies. Every initiative in a business should be evaluated against payback period or total cost of ownership to understand whether it's actually going to impact the bottom line. Often it doesn't. And when teams don't know this, initiatives are dead long before the kickoff meeting.

Most initiatives fail at inception. When cascading effects don’t get mapped ahead of implementation, execution gets the blame. But, sometimes, upstream, a few metrics out of place pushed a strategist to focus on something other than the root cause and took the rest of the chain with them.

They’re looking at the surface-level problem or opportunity because it is a quick-win or an emergency fire drill.

If we went by volume of orders, this customer segment made sense. But when we got outside the proxy measurements and into actual profitability by customer type, the story completely shifted. The growth strategy was wrong because it was built on a surface-level assumption, and no quality of execution could have fixed that. The real answer was three layers deeper.

Unfortunately, this happens in every organization as it grows, no matter how senior the team. Teams, at times out of necessity, build within a silo and don’t realize how things are going to connect, what other information they need, or even what metrics really matter when they are prioritizing the needs of their own department.

We’ve seen organizations with 56 planned initiatives, submitted to PMO from each business unit, where only 4 were investigating what drove costs up and retention down. The rest were consuming budget and bandwidth without moving revenue or profits, but instead were one-off reports for campaigns or projects that would have benefitted from a one-time structure so each new initiative could be compared from the same baseline. But, different teams had proposed the same project using different language and KPIs. Their metrics were never connected to the rest of the business.

This is a structural problem, not a people problem. When every department measures something different and the numbers only get reconciled in an emergency, initiatives get built on assumptions that were never validated. We’ve written extensively about where this waste accumulates and what it costs.

A metric can look right and still mislead you for a lot of reasons. Sometimes the underlying data in a source is broken. Sometimes the system is missing the full context of your business. Or other times, the metric itself is a proxy that doesn't actually track revenue performance.

We see this in OKRs where teams measure the number of things they'll get done but not how any of it impacts the bottom line. At that company, it was a mix of all of these.

This is the foundation of what we call Paired Metrics. A single metric in isolation tells you what happened. It can't tell you why, whether it matters, or what to do about it. Only the relationships between metrics can do that, and which relationships matter changes depending on where a customer is in their journey with you and what the market is doing at that moment. The pairing that explains a first purchase is not the pairing that explains a renewal. A system that measures two customer stages the same way will be confidently wrong about one or both of them.

At a publicly traded company we consulted for, organic traffic (unpaid search) dipped by 2% in a single quarter. Executives panicked. They had just spent over a million dollars restructuring their website. They circled the wagons around their website team, their strategy team, their marketing team, and their technology teams, trying to figure out if the numbers were accurate, what broke, and what they could do to fix it.

When we looked at the full picture, paid advertising had increased by 300% that same quarter, and total website traffic was up 800% year over year. The advertising was driving a massive wave of new visitors, and it had only slightly reduced the share coming from unpaid search. That 2% dip was noise. But because it was tracked as a single number without pairing it against the rest of the system, multiple teams spent weeks investigating a problem that didn't exist while a million-dollar initiative stalled.

Looking at connected metrics over time helps determine whether a change is significant. But even when it is significant, you still need to answer why. Often, answering that question requires qualitative data that links the behavior to the motivations of the market. What are customers actually saying in support tickets, cancellation forms, and reviews? What are they telling each other on Reddit? If you don't have this in your business, it's the single biggest opportunity to add context to your reports and get out of dashboard hell.

Most marketing teams are handed Return on Ad Spend as their North Star, and the platforms reporting it are the same ones selling the ads. Operations teams and IT teams, this is for you too: the next time your marketing team asks to build yet another ROAS dashboard, keep reading.

ROAS, which is calculated as revenue divided by advertising cost, can easily be manipulated by ad vendors that have the wrong incentives. Sometimes they'll reallocate budget toward remarketing audiences at the bottom of the funnel to make ROAS look higher at the expense of bringing in new leads and prospects six months from now. Or they'll offer a discount that lowers CAC but also lowers lifetime value at the same time, because now you're acquiring customers that only want a discount and will continue to only buy when one is offered. That's a direct hit to your margin.

Most teams that want to improve their ROAS immediately reach for attribution software. Platforms like Northbeam and Triple Whale, or a full customer data platform, will measure precisely whatever you point them at. That means they rely on your system of metrics. And if your metrics aren't tied to revenue, the risk becomes buying more measurement precision before you know that precision is accurate.

Accuracy requires organized data and a way to read the qualitative signals that explain why a number moves. The biggest pitfall in media management and attribution is precise measurement of the wrong metrics instead of focus on the right messages. Pulling your sales leads and traffic metrics into a single report you actually read would do more than another layer of tracking.

Consider a 5:1 ROAS. That means you generated $5 in revenue for every $1 spent on ads. Sounds great. But that number doesn't account for your cost of goods, your margin, your return rate, your cancellation rate, or whether those customers ever come back. At one company with a 20% product margin, a 5:1 ROAS meant they were giving the product away for free after manufacturing, shipping, and platform fees. The agency touts success, but finance knows the truth.

Here's where this gets real. At Old School Labs, where I was the CMO, we had two products that told opposite stories depending on which number you looked at.

Product B had the healthy ROAS. It recovered the cost of acquiring a customer in four months.

Product A cost twice as much to acquire a customer, and on the first purchase it came in below 1:1, meaning we spent more to make the sale than the sale brought in. Any team optimizing for ROAS would have killed it immediately. It looked like a loser.

Product A recovered its acquisition cost in two months. Half the time of the product that looked twice as healthy on the dashboard, because those customers came back on a much shorter cycle. They also stayed around three times longer.

Over the life of a customer, product B returned a modest amount of margin. Product A returned orders of magnitude more. At the cohort⁵ level it was the most profitable segment in the entire business, and cutting it would have taken a large share of the company's profit with it.

But even tracking these paired metrics wouldn't have been enough. We needed to understand who was actually buying product A. They were older men who had started noticing their energy and their strength drop off. The obvious play for a supplement company is to go after younger guys who want to be more fit, or the weight loss market, which were two of our other biggest cohorts. This entire segment would have been invisible if we had just cut the campaign with the lowest ROAS instead of investigating who was actually buying and why.

That's why qualitative data matters. Attributing performance to just channels or individual tactics would have missed this completely. Sometimes the message matters more than the metrics.

Let me be clear: none of this justifies a low ROAS. The point is to stop looking at any leading metric in isolation and start looking at everything as a system. This is what will enable your marketing teams to hold agencies accountable. It's also what will prevent your IT teams from getting bottlenecked by long, expensive implementations that only add to your operating expense at the cost of your margin.

Your acquisition cost, your lifetime value, your payback period, your cancellation rate, your margin by customer group. When those numbers are paired, the real story shows up. When they're not, you're making million-dollar decisions on a single data point. And when you add qualitative data to the picture, you find the segments that no dashboard can surface on its own.

Without all of that, most CFOs would rather write a check to one of Zuckerberg's charities than keep spending on the platform.

Everything we've talked about so far has been in the context of customer acquisition. But the same pattern shows up in operations, and the consequences compound just as fast.

Let's look at an example from team resources and workforce management. At 1-800 Contacts, the chat support team was handling two simultaneous conversations plus email. Volumes were high, queues were long, and the team wasn't meeting their SLAs. The obvious metric said the problem was capacity: not enough people to handle the volume. So the company offered overtime. The team was working 50 to 60 hours a week. They were burning out, and they still weren't meeting targets. Workforce management tried to incentivize with even higher bonuses per hour worked and the opportunity to earn more paid time off they could use in the future. The team was at risk of churning some of its top talent, and it would take a month before they could train anybody new on the floor.

One of my counterparts and I looked at the situation differently. The constraint we saw was cognitive load. An agent had to hold a happy customer in one window, a disappointed customer in another, and a logistics issue in a third, switching between all three continuously.

There is real research on what that costs. Sophie Leroy's work on attention residue found that when you move from one task to another, part of your attention stays stuck on the task you left, and it's worst when you were forced to leave that task unfinished.⁶ Earlier work by Rubinstein, Meyer, and Evans found that switching costs climb as the rules governing each task get more complex.⁷ Three different customers with three different problems are about as rule-complex as a shift gets.

Handle times ran longer because an agent was never fully present with the conversation in front of them. That’s the part a single metric would never show you.

The volume itself was also overstated. A meaningful share of it was the same customers arriving twice, through a second channel, because the first one hadn't answered them yet. We all know how that feels. Your package is missing, or you received the wrong product, or you simply ran out of something you need. It's urgent. You want it solved now. And the fear of never hearing back makes people tenacious. None of that shows up in a capacity metric. It shows up as more work, which reinforces exactly the wrong diagnosis.

The fix was specialization. Chat handle time dropped from 7 minutes to 3.5. Email handle time dropped from 10 minutes to 1. Overtime was eliminated. SLAs were met, and the team was happy. Volume actually decreased, because the duplicate contacts stopped.

That meant over $250,000 in annual costs removed without layoffs. The demands of the customers were met without any new hires or any new tools, and employee retention improved because people stopped burning out.

Capacity was the metric everyone was watching. It was the wrong one. Today, a large share of those chats and emails could be handled automatically with the right knowledge base behind them. But back then, one number tracked in isolation led to the wrong prescription.

Everyone can relate to that feeling of holding an Allen wrench in your hand, staring at the last step of the instructions written in bad English, then looking back at the packaging you excitedly ripped open only an hour ago, wondering how you're going to get all of this garbage back into the box to return it to Amazon.

Just as you were feeling relief about almost being done, having managed to get through every cryptic step, at times being tested on only your most elegant French, you realize the one piece that holds everything together wasn't in the box. That's exactly what cost one of our clients over a year of IT resources and hundreds of thousands of dollars before bringing us in to start everything over from scratch.

At an organization with an operating budget north of $100 million, the marketing team needed to track website analytics through Google's latest platform, GA4. Reasonable ask. But GA4 requires a connection to BigQuery, Google's cloud data warehouse, to access the full dataset. The team set that up. It worked, except for one critical problem: one of the most important data points, the session ID, wasn't available through the API. Google gates it behind their own cloud environment.

None of this surfaced before the commitment. The team was simultaneously transitioning from one data warehouse to another because they wanted something that would automatically scale with their data volume without having to manage infrastructure manually. Also reasonable. But their implementation vendor didn’t map the capabilities against the limitations before they started building.

The cascade: six months building out the platform. Three months designing the dashboards on top of it. Another three months before the dashboards showed them anything was wrong. It wasn’t until the project hit quality assurance that the original stakeholder asked why they couldn’t track data by session anymore. A year of work. Multiple teams involved. And the root cause was an assumption about an API that went unvalidated in week one.

Experienced product managers and engineering teams know you need to test small, early and often, before getting to an MVP and scaling to a solution of this size if you want to prevent these types of issues. But even when you're not building custom solutions, there are other Trojan horses that can sneak into your company and derail your strategy. Most of them start with a free trial.

There are a lot of middleware solutions and SaaS products that work as a band-aid for gaps like this. But none of them should ever be a substitute for doing the homework on how the underlying systems actually work, what’s possible, and what’s not. Most APIs expose all of their endpoints. They don’t force you to buy additional products just to access your own data. But some do. And if those dependencies go unmapped before the first line of code gets written, you end up with a year-long detour that was avoidable with a week of research. And you will be locked into a vendor because their terms and platforms don’t allow you to export your own data. They make you rent it instead.

While this goes a little deeper than paired metrics, it still illustrates the same point when applied to technology decisions. You're not just tracking the cost of a platform. You're tracking cost paired against capability, paired against dependencies, paired against time to value. The same principle that found the hidden customer segment at a supplement company and the division of labor fix at a contact lens company applies here: one number, tracked in isolation, led to the wrong prescription.

Before you start labeling me a Luddite or a deliberate contrarian out to rage bait for attention, hear me out: AI is extremely capable. It's true. Our entire company couldn't run at the scale it does without it. It's unequivocally been the most important invention since the internet... when used correctly.

But many of you are already seeing that AI doesn't fix broken data on its own. There are tons of SaaS companies and data vendors beating their chests on the idea that AI is only as good as the data you feed it. But much like treating diabetes, nutrition is only part of the equation.

Adding fruits and vegetables while continuing to eat candy and drink beer isn't going to fix your blood sugar, no matter how much you start running at the gym. Just the same, adding new clean data into your data warehouse for AI to look at isn't going to do much for your messy Salesforce objects that are disconnected, or your existing tables crowded with proxy metrics that don't track revenue and profitability. Because of context windows, throwing unstructured data at AI without a system to clean it or identify the features relevant to your business doesn't just overwhelm the model. It's one of the fastest ways to get your team reacting to advice that treats symptoms, not the system.

I don't know about you, but when it comes to my health, I'd much rather look to the doctor treating my whole system than the random YouTuber telling me all I need is a shot of pickle juice to enjoy as much cake as I want.

You're probably thinking: isn't that what Scale AI or similar platforms are supposed to fix? They're no different than any other SaaS product built to appeal to the lowest common denominator. They build out features they think everybody should be tracking.

But when it comes to commercial software, we always have to question the incentives. Consider how much of the research you encounter was paid for by a company with a product to sell. The same goes for the safety committees inside large organizations, and for the industry bodies that publish the best practices everyone else adopts. The people funding the study, staffing the committee, and writing the standard all have something riding on the conclusion. When a company's own product is the thing causing the problem, the incentive to go looking for that is zero, because finding the root cause would mean destroying the business model.

So whenever someone hands you a conclusion, the question worth asking is who wins if you believe it. It's like the old Roman poet my dad is named after says: "Who will guard the guards?"

I've seen this firsthand from working with both Equinix and Digital Realty, two companies that on paper do the same thing. Their business models and go-to-market strategies are completely different. Some of the audiences they target overlap, others don't. They use separate tech stacks. The conversations they have with their customers about specific needs use completely separate language, each addressing unique pain points and opportunities.

You need to identify your own features in the data and build out measurement strategies that work for your specific business model. It makes no sense to continue renting your insights and your own data back from vendors that are trying to embed themselves in your ecosystem permanently.

The same single-metric problem that led one company to build a strategy on a 30% repeat rate, that caused the team to spend a year building on an API limitation their implementation vendor never mapped, that led an operations team to throw overtime at a specialization problem, is now being accelerated by tools that make it faster to act on bad assumptions.

AI amplifies whatever you feed it. If you feed it a single KPI tracked in isolation, it amplifies the mistake faster, at greater scale, and with more confidence than any team member would have had making the same error manually. That's a whole Texas-sized sheet cake drowning in a family-sized jar of pickle juice.

That’s why the measurement system has to be right before you point AI at it.

You may recognize some of these problems in your own organization, or some that are similar. As we cover in our article on the S-curve of growth, scale naturally causes organizational drift and creates silos where teams start measuring different KPIs and answering to different incentives. The customer experience fractures first, and the employee experience follows it.

If silos are the root cause of these issues, the solution isn't to get rid of your team and try to scale back to being a founder-led organization. That's not what we're saying. While solving some of these symptoms with better systems will naturally correct headcount, the goal is to get all teams driving toward the same outcome and make sure every person in the company understands how they contribute to increasing profit.

Another consideration with paired metrics goes beyond tracking numbers over time. Those metrics have to be shared across departments. Any metric that lives inside only one department's dashboard eventually creates conflict, because two teams end up defending two different versions of the same quarter. If you feed AI paired metrics tracked over time, shared across departments, and validated with qualitative data, it amplifies insight. The system gets smarter because the inputs are connected.

We write in detail about the controlled experimentation methods that actually prove what's working and what isn't. But the experimentation only works if the metrics feeding it are paired and shared.

Marketing and sales need to work together on CAC because marketing controls the upstream activity that drives demand and lead generation. They need to make sure they're capturing the right attention and driving it toward the sale. It's sales' job to close. The entire equation of acquisition cost should be shared by both teams.

Sales also has to make sure they're closing the right customers and setting the right expectations to prevent attrition. That makes lifetime value a shared activity, bridging all the way upstream to marketing and the people responsible for building the product, rather than something customer experience and retention carry alone.

In order to make sure every team understands where these activities break down, it's the responsibility of BI and RevOps to give all of these teams a clear picture of where each activity hands off from one team to the next. That way, responsibility is shared. Accountability is visible. And the metrics that connect them are tracked together, not in isolation.

Without this, the entire organization gets stuck in effortful, deliberate thinking on every decision. Daniel Kahneman, author of Thinking, Fast and Slow, calls this System Two thinking: the slow, analytical, energy-intensive mode your brain uses when it doesn't have enough information to act intuitively.

When your teams don't have shared metrics giving them a clear picture, every decision requires grinding through incomplete data, second-guessing other departments, and competing for resources with a scarcity mindset. It causes stress, leads to overanalysis, and it's exactly where collaboration breaks down, because the numbers haven't earned enough trust for anyone to act without protecting themselves first.

This isn't theory. This is Thinking, Fast and Slow illustrated in your organization in real time. There's a reason he won a Nobel Prize in Economics as a psychologist. When shared metrics give every team the same clear picture, more decisions can move into System One: fast, intuitive, collaborative. That's because instead of trying to understand what moved behind just a single metric, teams can be armed with real insights that span how their activities function within the entire system. The team can direct their energy toward the customer instead of toward defending their dashboards.

Every person in your organization needs a balance of both: the intuition to act quickly and the rigor to validate what's working. But you never want those two modes to turn into business units competing with each other instead of functioning as a single team. That transitions the organization from a fractured group of silos into a single unit competing with actual competitors instead of cannibalizing each other from the inside.

That shift is the difference between treating blood pressure alone and diagnosing diabetes.

No one wants to hear that their department or their organization might be in disarray. It's a fair question to ask "how bad can it really be?" when no one on the team is raising these problems. But nobody's report covers the gaps between functions, so no one's job is to raise them. We've written about why people don't feel safe raising these issues and what it costs when the feedback loop is broken.

My dad didn't want to hear he had type 2 diabetes either. But the cluster of symptoms told the truth before any single number could have. Fortunately, catching it in his late 40s likely prevented many more serious consequences of leaving the disease untreated.

Most companies call us when revenue is declining, which is the last sign that the system has been suffering for a long time. The early warning signs are almost always visible months or years before the revenue catches up. Those early warning signs are your leading indicators. Revenue is a lagging indicator, the last thing to move. It's important not to ignore them. If your current solution is to treat rising CAC with more spend, to cap it at a specific target MER, or to treat declining retention with a loyalty program without ever connecting the symptoms to ask whether they share the same root cause, your business may not just have high blood sugar. It might already be suffering from clogged arteries.

Fourth Order Intel pairs your metrics, maps them to revenue impact, and identifies the 15 to 20 percent of operating expenses that single-point measurement misses. The average engagement produces findings within 30 days.

References:

¹ Beer, M. and Nohria, N., "Cracking the Code of Change," Harvard Business Review, May-June 2000.

² McKinsey & Company, "The keys to a successful digital transformation," and McKinsey Global Surveys on transformations, 2012, 2014, 2016, and 2018.

³ MIT Initiative on the Digital Economy (NANDA), 2025. Analysis of enterprise generative AI deployments.

⁴ CMO Council and Deloitte Digital, "Humanizing + Analyzing Relationships To Drive Revenue, Retention And Returns," 2021. Survey of 150 global marketing leaders.

⁵ A cohort is a group of customers who share a common characteristic or acquisition period, allowing you to compare performance across segments rather than in aggregate.

⁶ Leroy, S., "Why is it so hard to do my work? The challenge of attention residue when switching between work tasks," Organizational Behavior and Human Decision Processes, 109(2), 2009, pp. 168-181.

⁷ Rubinstein, J.S., Meyer, D.E., and Evans, J.E., "Executive Control of Cognitive Processes in Task Switching," Journal of Experimental Psychology: Human Perception and Performance, 27(4), 2001, pp. 763-797.

My grandpa is a collector of vintage cameras. Like any old collector, he's quick to tell you about how much better the engineering of older equipment is.

"Things today are built to fall apart… It's planned obsolescence."

Unfortunately, he's right. And it's not just cameras. In many business models, the failure point is the business model.

From the auto manufacturers whose belts wear out at predictable cadences, to the batteries inside phones that will need to be replaced after so many charge cycles. These models are designed to extract revenue over long periods from the same customers.

But those failure points aren't advertised, and in the case of your vendors they most certainly don't come up on discovery calls. They're what you find out later. After the contracts are signed and the check clears on the other side of the sale.

The costs of "learning" on ad platforms and keeping up with the algorithm, or maintaining your cloud architecture, quickly balloon from an experiment to a daily operational expense. And often, once you get going, you can't stop.

That “can’t stop” fear is just one myth we're going to talk about today. By the end of this article, you will see how incentive structures create industry “best practices” designed to create hidden failure points to maximize revenue for vendors, while costing you your profits.

It probably will take you only 30 seconds of doom scrolling before you see an ad to "get the most out of your data" or a promise about how a new tool will "help you scale your business." If you have the discipline not to doom scroll, your LinkedIn inbox is flooded with the same types of messages.

We've seen that message on everything from analytics connectors to Salesforce. It's ambiguous. It doesn't offer a specific outcome. It feels vague, almost as if they hope you’ll imagine what that means on your own.

Unfortunately, the only thing most tools predictably scale is your costs, not your revenue. Not quite the promise we expected.

The sales team pitching you knows what their tool does. They don't know how your business makes money. And that means the tools they are pitching are designed as one-size-fits-all. You can imagine how that works out.

Most marketing agencies charge a percentage of your ad spend or a retainer that scales with the scope of work. The more you spend, the more they earn. That's the model. It means the agency is structurally incentivized to recommend increasing your budget, expanding into new channels, and launching more campaigns, regardless of whether the current ones drive profit. That is a perverse incentive: they make more money even when you lose money.

At a large insurance brand, a new agency reported decreasing Customer Acquisition Cost¹ (CAC). We were brought in as a second set of eyes on the initiative because some of the internal stakeholders were skeptical. In the interest of transparency, this agency sold and managed programmatic² ads, a service our firm did not offer and had no stake in, except to protect the interests of our client.

Their first test did show lower acquisition costs within the first three months, according to their reports. But, looking outside of the organization, we found a seasonal demand wave created by external triggering events, which neither the agency nor the business had been monitoring.

Because their test was global to all states and regions where this company could legally advertise, their data was noisy. They didn't know that. Or if they did, it wasn't in their best interest to admit it.

A seven-figure renewal was on the line for the agency.

For the insurance brand, on the other hand, that represented a 20% increase in ad spend.

The agency pressed for one more test, and this time we had full control over experiment design.

We deployed the attribution and measurement systems we've used to grow businesses from $3M to $100M in just a few years, and on turnarounds for brands like Potbelly.

Part of that system involves controlled experiments focusing on contribution margin.

The agency insisted they wanted recall to be the primary metric. We put our foot down on looking at net new acquisitions.

Acquisition costs went up in every treatment where the ads ran. When CAC did go down, it went down globally, across both the control and treatment areas, aligning with larger news events and seasonal waves that impacted demand. As we suspected, the agency had been riding a seasonal demand curve.

To be fair to the agency, we did note opportunities in their creative messaging and the structures they used to design their cohorts so they’d be armed with what they needed to succeed in the future. But the findings were clear: the ads hadn’t reduced CAC as the agency originally reported. So, the seven-figure renewal didn't happen.

When the firm doing the analysis also sells the media or the creative, there's no incentive to tell you what's not working, unless they are invested in your success. When they aren’t, their revenue depends on you continuing to spend and scale even when it doesn't work.

As we mentioned in our article on what marketing audits and consultants miss, we know an agency owner who told us directly that their strategy was to "extract revenue from enterprises." Even when they knew internal teams were perfectly capable of doing what they were doing and could have pointed it out, they didn't mention it. The cruel irony is that this was a vendor hired specifically to identify growth opportunities within the business and to enable teams through skilling them up. Which implicitly means they should have had the goal of finding profit.

It got worse when they found out another vendor was doing something similar. They spent half their time and resources trying to figure out how to undermine that other vendor so they could win that contract too. Instead of finding profit, they were protecting their territory, afraid of being found out, all the while hurting the morale of the internal team. The right thing to do would have been to enable the internal team as much as possible, even if that means working yourself out of a job.

It's not just ad agencies of course. It's any vendor that sells you services where there's an exchange of time for money. The incentive to expand scope, extend timelines, and add deliverables is baked into many vendor models. First year is capital expenditure. Every year after that, it comes out of operating expense. And it never gets questioned because it's "the agency we've always used."

It's the reason we scope every engagement up front. Our fees are flat and not tied to your spend, and we don't bill by the hour, so you know exactly what you're getting. ROI is planned from the start.

Marketing agencies have their place and often can do things no internal team can. But why do some agencies insist on staying long past their welcome?

Let's zoom out for a minute. All of us have heard that "breakfast is the most important meal of the day." If you've ever visited a high-end home from the early 20th century, you've probably walked through a music room. Neither of these ideas showed up on their own. Both of them trace back to Edward Bernays, the father of propaganda.

It turns out bacon and eggs were mostly eaten at dinner. But the Beech-Nut Packing Company wanted to sell more bacon, so Bernays got 4,500 doctors to confirm that a hearty breakfast was healthier than a light one, then published the results in newspapers across the country.³ The American breakfast was born. Just the same, Bernays outlined in his 1928 book Propaganda⁴ how to sell pianos without ever asking someone to buy a piano. Instead, you convince architects to include music rooms in their home designs.

You stage exhibitions of music rooms by well-known decorators. You invite famous musicians and society leaders. Before long, having a music room is a status symbol, and no one really understands why.

Thought leadership from industries walks this same fine line between being genuinely helpful and creating unnecessary problems. The agencies we work with follow the same trend reports and industry narratives that all of us read on LinkedIn or Forbes. Most of it is genuinely helpful.

But it's hard to know when there's another motive at play. They've got to sell you on ideas that make the industry's offerings seem both irresistible and essential for the next stage of your business.

Let's look at how their incentives don't always align with yours.

B.F. Skinner proved you can make a pigeon superstitious.⁵ Put it in a cage and arrange for food to appear at regular intervals. Whatever the pigeon happens to be doing when the food arrives, spinning around, bobbing its head, it will keep doing over and over again, convinced that the dance caused the food to appear. Seth Godin wrote about this in Fast Company, and the parallel to business is uncomfortable: there's plenty we do, plenty we've always done, that has nothing to do with what actually works. But once we've made up our minds, we're like pigeons. We don't want to change our behavior, regardless of how much data we see to support a better alternative.

There are a few key superstitions that advertising platforms have created that impact what your agency does downstream.

The myth of awareness. The concept of awareness has been bastardized. It's been twisted inward to have businesses focus on the idea that we need to help customers become aware of who we are. "Here I am!" Look at us. Notice us. Remember our name.

The original intent of the customer awareness model comes from Eugene Schwartz, whose 1966 book Breakthrough Advertising is still considered one of the most influential works on advertising ever written. By his own account, 85% of his ads paid out, a hit ratio he described at a Rodale Press seminar and one of the highest anyone in direct response has claimed.⁶ He would know. He literally wrote the book on this. And his model was for businesses to be deeply aware of what customers cared about, in a specific sequence:

First, helping them organize their symptoms into a simple, easy-to-state diagnosis, moving them from unaware to problem-aware.

Second, helping them understand what solutions were available and comparing and contrasting yours.

Finally, helping them by educating them on what your product can do for them, promising them a transformation.

That last step is where most marketing funnels begin. Even sophisticated models like the bowtie funnel introduced in Revenue Architecture⁷ by Jacco van der Kooij make this same mistake: assuming that awareness starts with telling people "here I am" and trying to capture a lead where you then have the opportunity to educate them. Why would anyone want to give you their email address before they even understand they have a problem?

It all goes back to how advertisers have made money since the newspaper days. They make money when you buy an ad. The more ads you buy, the more budget you put behind them, the more money they make. But what if you don't get sales right away? They want you to think that direct response is something only tacky marketers do, and that you're doing something more sophisticated by "raising awareness through branding."

Yes, there are brands that win with a brand play. But those brands have massive budgets and years of market presence.

If you're BMW, you can afford high acquisition costs and have a huge moat as a luxury vehicle.

If you're Popcorners betting on a Super Bowl commercial, you don't do that before you're in every grocery store to capture the moment someone wanders down the chip aisle looking for snacks.

Distribution models rely on a sequence of logistics. Successful advertising relies on a sequence of needs and wants, not just the ability to state you exist.

For businesses where there's no organic rediscovery, or where the problem is far disconnected in the customer's mind from the available solution, awareness doesn't translate into sales. It wastes money.

Think about Spirit Airlines. You knew they were around when they existed. But only as the butt of a joke. Being aware of them doesn't mean you're going to book them for your 10th anniversary trip. Awareness does not equal action.

Keep everything running no matter what. You've probably heard this from your reps. It doesn't matter whether you're advertising on Meta, Google, or TikTok. Every one of them will tell you "don't turn this off, it will reboot the learning phase" or that one ad is helping these other ads get credit. They want you afraid to turn off what's not working, thinking it will hurt everything else that is working.

One of my friends owns several successful chiropractic clinics and was spending nearly 15% of his revenue each month on advertising heading into the pandemic. When he thought he wouldn't be able to continue operating, he shut off all his ads. Revenue stayed the same. It turned out everyone was finding him organically through Google My Business and local SEO. When he was ready to open a few new locations and actually grow, we were able to get him better results for a tenth of what he had been spending before, helping him open two more clinics.

It turns out a century later, this quote hits just as hard as it used to: "Half the money I spend on advertising is wasted. The trouble is, I don't know which half." In a world of bots, ad blockers, and third-party data becoming less and less available as privacy regulations expand, even the attribution tools you're paying top dollar for can't do much for that problem. It really doesn't help when every platform wants its credit.

So how do you know what's actually working? It's common practice to "raise awareness" on Meta, driving to a vague home page, leaving prospects to Google to figure out what it is you actually do. They end up reading about you on an affiliate site or a reviews blog, only to come back in organically, if they come back at all.

You've forced them to jump from channel to channel. In this case, all channels were necessary, but that's not always the case. It happens in organizations of all sizes, from the Fortune 15 down to brand-new businesses. And the more fragmented channel management gets, the more they start cannibalizing each other between competition for budget and credit.

These strategies can look like a win because revenue goes up. But 15-30% of your margin gets lost in the inefficiency of having to leave everything on to protect a fragile system.

Between gaps in the customer journey and fragments of information spread across multiple channels, most customer experiences are an obstacle course when they should be an amusement park.

Don't let your agencies substitute research with "industry best practices" designed to benefit the platforms that created them. It's the reason a certain platform keeps telling you that shorter is better when it comes to ads. That "attention spans are only 8 seconds."

How many times have you heard people share that attention spans are only 8 seconds? It's still shared on LinkedIn by executives. To their credit, the reports shared by their teams likely corroborate this, but not for reasons they'd expect.

The latest data shows that the most popular creators on YouTube are long-form. MrBeast averages about 30 minutes per video. Kill Tony, one of the biggest comedy podcasts, runs for three hours. Alex Hormozi goes for an hour. Cody Sanchez averages about 20 minutes. Even Ninja Kids runs 15 to 20 minutes for content directed at children.

But creators like Jack Neel, Joe Rogan, and the Diary of a CEO average 3 hours for long-form conversations.

The most watched creators? The wealthiest creators? The most trusted? Same creators. All long-form content.

It’s a hell of a lot longer than 8 seconds. This is yet another "breakfast is the most important meal of the day," undermining your ability to connect meaningfully with your customers.

Maybe you're thinking, "But Omar, we're not content creators. I don't want to sit in front of a camera all day." All content has to be designed to persuade, and importantly, hold attention.

Someone in your network has shared an infographic or a quote from the great Robert Cialdini's book, Influence.⁸ This principle is proven time and again: what we pay attention to, we deem more important. If you can hold attention long enough, you will be considered more important in the customer's mind.

One of our longest-running ads was over 3 minutes with a watch retention of 80-90% for almost 3 years. We only had to relaunch it after an algorithm update had us remove the word "ass." And it kept running for years after that. 3 minutes.

In fact, most of our ads have been long-form for the past decade, holding attention to broad audiences for 90 seconds or longer. And the most important thing: they drove tens of millions of sales. That ad, three years untouched and a few more after that single edit, took a business from $300k per year to $20M per year and a healthy exit.

If watch retention is 8 seconds, the content is aimed at the wrong person, says the wrong thing to them, or misses on both. That's what makes this next myth so brutal.

This one is ironic. Any ad platform is a content platform first. They are designed to keep users inside their ecosystems by showing engaging content. Whether that's blogs, videos, social media, or music, people came to pay attention.

So the idea that your Return on Ad Spend⁹ (ROAS) is the most important metric when it comes to promotion embeds an assumption: you have to spend money to be seen.

For most brands, this is true, and there's nothing wrong with advertising. When you're unknown, it's the fastest way to get noticed. But advertising doesn't compensate for not having the best product in the market. As Robert Stephens, the founder of Geek Squad, once said, "Advertising is the tax you pay for being unremarkable." The less remarkable you are, the more you pay.

Think about this. Tesla reached $81.4B in revenue and a trillion-dollar valuation with an ad budget of under $200k. About $0.15 per vehicle sold.

Then there's Apple. Every campaign had a goal to reshape a conversation and redefine the culture around its products. And it paid off.

From their 1984 campaign that drove $155M in sales, to the iPod ads with colorful backgrounds, and the iconic "I'm a Mac, and I'm a PC" campaigns, Apple has achieved more profitable ad efficiency than any other company of their scale.

They became the first trillion-dollar publicly traded company. And they did it, spending 1 to 2% of their revenue on ads.

Their secret weapon was never the media buy. It was knowing their customer. The better they knew that customer, the less they had to spend to acquire customers.

Apple advertising investment relative to revenue. Source: Form 10-K filings, 2006–2015.¹⁰

In Steve Jobs' own words: "You've got to start with the customer experience and work backwards to the technology. You can't start with the technology and try to figure out where you're going to try and sell it."

The 1984 campaign held the subtext, "you're worth more than just being a corporate drone," leveraging motifs from one of the most famous works of dystopian literature.

The iPod commercials doubled down on standing out from the crowd with white headphones instead of the black ones everyone else had. And they never had to say you can dance around without your CDs skipping. They showed it.

And the "I'm a Mac" campaign solidified Apple as the go-to brand for creatives and a status symbol for those who could afford them.

Over the decade, Apple's ratio of marketing spend to revenue improved because its target wasn't restricted to ROAS or whether it outspent the competition.

Goodhart's Law puts it plainly: once a metric becomes a target, it stops being a good metric. And teams focusing on short-term ROAS wins to keep the status quo are missing top-line growth while squeezing margin every second they ignore insights about their customers.

As we cover in our article on the S-curve of growth, market fit goes deeper than just the product, and it's not a one-time event. And neither is scaling your ads to a specific monthly budget.

That's what makes this next myth a gut-punch.

As we just looked at, the platforms exist to capture your attention for as long as possible. They do this with algorithms.

If knowing your customer better directly translates into better profits, it's no surprise the platforms with the largest data stores have the best algorithms and the margins. Their decision engines aren't built on just age, income, and demographics.

They're built on what people do, how they react, and what they pay attention to. It's why two people on opposite ends of the world, in different age groups, with different career paths, can both be shown the same shows on YouTube, or even the same ad for a SaaS product. Relevance is about behavior cohorts more than simple personas.

So the good news: algorithms are not designed to steal money from you or to hurt your advertising. They're designed to connect better with customers and provide users with a better experience.

Nobody has more data on what each of us cares about, how we feel about brands and products, and social issues than social media platforms. AI is quickly catching up. As of 2026, millions of people are using AI as a therapist, a confidant, and a sounding board for their deepest concerns. It's no wonder OpenAI just started selling ads.¹¹

These qualitative data stores are massive and growing larger every day.

The bad news: They can't share it with you. Privacy laws don't allow them to. And it's their competitive advantage against the next platform trying to sell you ads.

Instead, they aggregate their insights to the lowest common denominators:

"Others in your niche are seeing the best results from short-form content."

"You've got decent hooks, but you really want to be using UGC or testimonials."

"Have you looked at your landing page? Maybe there's a better place to drive traffic."

Meanwhile, everyone's content breaks when there's an algorithm update.

This is where all of that bad industry advice across domains compounds from seeming genuinely helpful to financially dangerous.

Let's think about how that advice sits against the real world.

So, UGC and testimonials are all that matter? Because they have the highest ROAS?

Discounts often have the highest ROAS, too. Should we be giving away the product to increase ROAS?

It's no surprise that the ads that perform the best using the ROAS metric capture people at the end of their customer journey and are reported on a metric built for and tracked by the pixels placed on your site by the people who make money by having you pay for ads.

Last click attribution is like saying Apple is only successful because Tim Cook was the CEO.

It's a clever frame. They know you won't like the ROAS metrics of the content that does the real persuasion and changes public opinion, because it doesn't sell right away. They know you won't like the truth that not every click or impression can be directly tracked. But that indirect measurement is critical for growth.

And what they most don't want you putting together: dollars out have to multiply dollars in. Not just in revenue, but in profit. And that's the real problem with ROAS. It reports ad costs relative to revenue, not profit.

That's why with Potbelly, we made spend five times more efficient at level budget, and grew average order size, before we scaled.

This approach helped swing Potbelly's net income from a $23.8 million loss to $4.3 million positive.¹² That efficiency supported a 30% lift in revenue in 2021 and 19% the following year, with same-store sales up 18.5% and full-year shop-level margins expanding from 7.4% to 10.5%. Their own earnings release credited "strength in digital marketing and traffic."

And we did that by focusing on the feeling our ads created, built on hundreds of insights we gathered from the market in real time, only treating ROAS as a leading indicator, not the destination.

The incentives at play keep you from the industry insider strategies that dwarf "best practices" because they're designed to gate data and to serve reps who just want to see you spend more on their platforms.

Remember earlier how we talked about how it's not just ad agencies, but all vendors whose incentives are not aligned to your business? Unfortunately, it's not just the ad platforms either.

From SaaS companies that want to sell you more seats to justify their sales teams' quotas, to cloud platforms that make more money the more you store on their systems, all of these businesses survive on the vague promise that you'll get better insights and make your team move faster. It is the same perverse incentive at work: these vendors earn more when your costs go up. None of them are willing to be accountable to your actual top and bottom lines.

They know it. You have to wonder who created the term "middleware" to begin with.

In the age of AI, a lot of these tools are poised for disruption. Our team has been able to cancel several subscriptions by replacing them with on-demand APIs and Claude Code, completing the same work at a fraction of the cost with more flexibility and no recurring contracts. But there's a catch: we are starting to see token costs increase. It's what made us pause to focus on building deterministic software and minimizing context into each LLM call, avoiding the new industry trap: agentification.

The people telling you to use more tokens are the ones selling the tokens. This builds on the trend we've already seen with cloud providers.

They perpetuate myths like "data gravity," the idea that the more data you have, the more tools, data, and talent it attracts. Of course they want you to think you've got to spend more money storing every single data point you've ever tracked, even when you're measuring proxies like ROAS or views or shares. Things that don't track back to revenue.

Most of the data that lives inside these clouds can be archived without any impact to business operations. Once the actual insights of what happened and when are extracted, they go from raw data into real intelligence.

In fact, against most enterprise systems we've encountered that cost millions to build, we've seen teams relying on raw data for decision making in dashboards, while the actionable intelligence was extracted with simple tools like spreadsheets and local statistical analysis. Free tools.

This extends the point we made in our article on what marketing audits miss: half of purchased martech tools go unused by the teams that bought them, which means the cost of implementation, maintenance, and roll-outs was wasted. That waste is only accelerating as quicker, better, and cheaper options become available that your teams may not even be aware of, yet.

With all of these channels running and conflicting narratives bombarding us all day, it's no wonder that when we identified gaps in the go-to-market and data infrastructure plans with the CIO of a half-billion-dollar insurance company, his word to describe our findings was "dizzying." He recognized the truth in them right away.

Others in his position have asked, "With everything I have running, how do I even know what works and what I'm wasting money on?" That question is the origin of our most successful frameworks based on subtraction. We've relied on it for attribution modeling, causality, and tracking incremental revenue lift in every engagement since 2016.

If there's any wisdom to take away from this analysis of external incentives, it's not to trust the advice of those who don't actually do the work or take accountability for the results at face value.

Your ad platform reps, your cloud vendors, your SaaS sellers all have their own agendas because they've got families to feed, bosses to impress, and team members to support. You making more money or hearing the full truth doesn't always align with their incentives.

But if that's what's happening outside your organization, what's happening on the inside?

Fourth Order Intel's diagnostic identifies 15 to 20 percent of operating expenses as recoverable within 30 days. The average engagement pays for itself in the first month.

Notes:

¹ Customer Acquisition Cost (CAC): total marketing and sales spend divided by the number of new customers acquired.

² Programmatic: automated ad placements across websites and video platforms, purchased through bidding algorithms.

³ Tye, L., The Father of Spin: Edward L. Bernays and the Birth of Public Relations, Crown Publishers, 1998. Documents the Beech-Nut Packing Company "hearty breakfast" campaign and the 4,500-doctor survey.

⁴ Edward Bernays, Propaganda, Horace Liveright, 1928.

⁵ B.F. Skinner, "'Superstition' in the Pigeon," Journal of Experimental Psychology, 1948.

⁶ Eugene Schwartz, Breakthrough Advertising, 1966. Hit ratio as described by Schwartz at a Rodale Press seminar and widely reported in direct-response literature. Self-reported, not independently audited.

⁷ Jacco van der Kooij, Revenue Architecture, Winning by Design, 2023.

⁸ Robert Cialdini, Influence: The Psychology of Persuasion, 1984.

⁹ Return on Ad Spend (ROAS): revenue divided by advertising cost.

¹⁰ Apple Inc. Form 10-K filings, fiscal years 2006–2015. Advertising expense disclosed under Summary of Significant Accounting Policies; net sales from segment reporting tables.

¹¹ OpenAI confirmed advertising in ChatGPT January 2026, with rollout to Free and ChatGPT Go tiers in February and a self-serve Ads Manager in July 2026.

¹² Potbelly Corporation, fourth quarter and full year 2022 results. Full-year revenue of $452.0 million, up 19%; same-store sales up 18.5%; shop-level margin 10.5% versus 7.4%; net income of $4.3 million versus a net loss of $23.8 million.

Related: