Many Fortune 500 companies unknowingly undermine their profitability by treating owned acquisition channels, such as homepages, or in-product promotions, as “free” resources. This practice results in inefficient resource allocation and discourages effective targeting, as teams have little incentive to optimize when advertising comes at no cost. By failing to account for opportunity costs and economic margins, businesses risk misallocating valuable assets and missing opportunities for growth.
This post explores how understanding opportunity costs and economic margins can drive better targeting decisions, maximize the yield on owned channels, and unlock long-term profitability. Indeed, innovative businesses are instead pricing it and allocating it via market mechanisms, using Retail Advertising Networks. No matter what solution you adopt, one thing is clear: Treating owned attention as “free” is a costly mistake that could cost your business hundreds of millions of dollars every year.
Motivating Example: The Economics of Opportunity Costs
To fix ideas suppose an online merchant sends out a mailed brochure every November for its Panettone Christmas cake. Because printing and mailing the product brochure is costly, the firm only wants to send it to customers who would make a purchase if they receive the printed brochure but not otherwise.
Assume a price of $3 per cake, costs of goods sold (COGS) of $1, and operating expenses of $1 to design, print, and mail a brochure. The gross margin (i.e. revenue minus the cost of goods sold) on each product sold is $2. To make matters even simpler, assume customers are limited to buying at most one cake, and the firm makes cakes on demand so we can ignore inventories etc. As a result, if a customer gets a brochure and makes a purchase the net margin on that sale, defined as gross margin minus operating expenses, is \(2 - 1 = 1\). If they don’t make a purchase, the firm incurs a loss of \(-1\), the cost of the mailer.
Consider two extreme policies. Let \(\pi^0_t(i) = 0\) denote the policy that does not send a brochure to any customer \(i\) in year \(t\). Let \(\pi^1_t(i) = 1\) denote the policy that sends brochures to all customers. Let \(y_{i,t}\) denote whether customer \(i\) makes a purchase in year \(t\). Specifically, \(y_{i,t} = 1\) if they do, and \(y_{i,t} = 0\) otherwise. Let \(y_{i,t}(1)\) and \(y_{i,t}(0)\) denote the potential outcomes if the customer receives or does not receive a brochure, respectively.
Now, partition the set of customers according to their potential outcomes (the first two columns of the table below) as follows:
| \(y_{i,t}(0)\) | \(y_{i,t}(1)\) | Gross Margin \(\pi^0_t(i)\) | Gross Margin \(\pi^1_t(i)\) | Net Margin \(\pi^0_t(i)\) | Net Margin \(\pi^1_t(i)\) | Opportunity cost \(\pi^1_t(i)\) |
Economic Margin \(\pi^1_t(i)\) | Count | |
|---|---|---|---|---|---|---|---|---|---|
| Adverse Responder | 1 | 0 | 2 | 0 | 2 | -1 | 2 | -3 | 2 |
| Always-taker | 1 | 1 | 2 | 2 | 2 | 1 | 2 | -1 | 10 |
| Never-taker | 0 | 0 | 0 | 0 | 0 | -1 | 0 | -1 | 80 |
| Responder | 0 | 1 | 0 | 2 | 0 | 1 | 0 | 1 | 8 |
Understanding Customer Segments and Their Impact on Margins
The customer segment labelled “Adverse Responder” is defined by customers who would make a purchase if they do not get a brochure (\(y_{i,t}(0)=1\)) but not otherwise (\(y_{i,t}(1)=0\)). Under the policy \(\pi^0_t(i)\) the firm makes $2 net margin per Adverse Responder but loses $1 under policy \(\pi^1_t(i)\) (zero gross margin minus one dollar in operating expenses to send a brochure).
The customer segment labelled “Always-taker” is defined by customers who would make a purchase whether or not they get a brochure. Under the policy \(\pi^0_t(i)\) the firm makes $2 net margin per Always-taker, but only $1 under policy \(\pi^1_t(i)\).
The customer segment labelled “Never-taker” is defined by customers who would never make a purchase whether or not they get a brochure. Under the policy \(\pi^0_t(i)\) the firm makes $0 net margin per Never-taker, but loses $1 under policy \(\pi^1_t(i)\).
The customer segment labelled “Responder” is defined by customers who would make a purchase only if they get a brochure. Under the policy \(\pi^0_t(i)\) the firm makes $0 net margin per Responder, but makes $1 under policy \(\pi^1_t(i)\).
The opportunity cost of policy \(\pi_t^1(i)\) is the net margin forgone by not choosing policy \(\pi_t^0(i)\). In the case of Adverse Responders the opportunity cost of choosing \(\pi_t^1(i)\) over \(\pi_t^0(i)\) is the two dollars in forgone net margin under \(\pi_t^o(i)\), and so on.
Economic margin is computed by subtracting explicit costs, like costs of goods sold and operating expenses, as well as implicit opportunity costs, from revenue. Equivalently, since Net Margin already subtracts explicit costs, we need only subtract opportunity costs from Net Margin to get to economic margin. This is what the table shows for policy \(\pi_t^1(i)\). As can be seen, the economic margin is substantially worse than the net margin for all segments except Responders.
Finally, the last column labelled Count shows the number of customers by type in a hypothetical customer population. In this population of 100 customers there are two Adverse Responders, ten Always-takers, and so on.
The True Cost of Misallocated Resources
The data in the previous table enables us to calculate the following metrics for the lift of \(\pi_t^1(i)\) relative to \(\pi_t^0(i)\), or the average treatment effect (ATE):
| Concept | Lift (ATE) | Stages of Enlightenment |
|---|---|---|
| Conversion | 6 | 😊 |
| Revenue | 18 | 😃 |
| Gross Margin | 12 | 😄 |
| Net Margin | -88 | 😦 |
| Opportunity Cost | 16 | 😖 |
| Economic Margin | -104 | 😱 |
The lift values in the table above are calculated by comparing the outcomes under the two extreme policies: sending brochures to all customers (\(\pi^1_t(i)\)) versus sending to none (\(\pi^0_t(i)\)).
Since “Never-takers” and “Always-takers” do not change their behavior in response to treatment, the net effect of the intervention comes entirely from “Responders” and “Adverse Responders.” Specifically, “Responders” generate a positive lift because they only purchase when treated, while “Adverse Responders” generate a negative lift because they only purchase when not treated. The overall lift is thus the sum of the gains from Responders minus the losses from Adverse Responders, weighted by the count of each segment and the magnitude of their respective effects.
This approach isolates the true incremental impact of the intervention by focusing on those whose behavior is actually changed by treatment. For example, the lift on Conversion of sending a brochure to all customers is 6. This is the difference between 8 Responders that convert because of treatment, minus 2 Adverse Responders that no longer convert because of treatment. The other concepts in the table are computed analogously.
There are two key takeaways here:
-
The technical problem is clear: We would like to send brochures only to the Responders. Now, because we do not fully observe potential outcomes we cannot partition the population exactly as in the table above. Hence we need to infer whom the Responders are by combining good data, with defensible assumptions, and appropriate statistical techniques.
-
Taking into account direct operational costs (to compute Net Margin), and opportunity costs (to compute Economic Margin) makes a huge difference to overall lift, and the advisability of sending brochure to all customers. Thus, whereas sending a brochure to all customers looks great in terms of conversion, revenue, and gross maring, it is a non-starter in terms of Net, and Economic Margin.
Now, before you say “Duh, this is obvious”, consider that a major driver of customer acquisition in Fortune 500 companies is owned channels. As I show below, owned channels are typically priced at $0 to owned businesses, resulting in significant capital misallocation.
Targeting Is Not Just A Technical Problem: The Role Of People And Process
Faced with a technical problem most technically adept leaders will immediately reach for a technical solution, like using Machine Learning or AI to help with the targeting. Here I want to challenge the assumption that the technical problem outlined above is even perceived as a problem in many organizations.
As noted, many Fortune 500 companies have a major owned acquisition channel, like the company home page. They also have numerous new initiatives or business lines, all vying for attention in owned channels. Hence, the demand for promotion in owned channels far exceeds the available promotion inventory.
Typically, this excess demand is reconciled via an ad hoc bargaining process among the executive leadership team. The final allocation is then given away for free to the winning teams. Thus, the chosen teams get to run campaigns in owned channels at zero marketing cost to them.
Moreover, because these teams are unaware of opportunity costs, or have zero incentive to include them in their calculations, their focus is conversion, revenue, and gross or net margin (these are identical with zero operational cost). However, ignoring opportunity cost and economic margin can result in a terrible misallocation of resources, as shown in the table above.
The problem with giving away scarce resources for free is that there is no such thing as a free lunch.
If the promotion space is so valuable, the opportunity cost of handing it to one initiative versus another, or indeed a third party, is almost certainly not zero. Presumably, the firm could have sold the advertising inventory in the open market, much like Amazon has done in Amazon Marketplace (as opposed to promoting its own products), and maybe make more economic margin.
Yet by deciding to give away lunches for free the firm can severely impair its decision making. In effect, it is creating incentives for teams to adopt a “spray and pray” posture of blasting all customers indiscriminately (\(\pi_t^1(i)\) above) over carefully targeted interventions even when the former is clearly inferior in terms of economic margin.
Why Ignoring Opportunity Costs Discourages Targeting
“Intelligent people make decisions based on opportunity costs”
– Charlie Munger
Net margin, or accounting profit (what Finance and Wall Street typically measure) can be very misleading for the purposes of optimizing targeting and maximizing yield on business assets.
First, the winning teams have zero incentive to target. The operational cost to them of promotions in owned channels is zero, yet the opportunity cost (to them) of not sending a brochure is some positive probability of missing a Responder. For these teams targeting is the problem, and blasting everyone with an ad the obvious solution.
Second, why worry about theoretical yield losses when actual (accounting) yields are positive? Remember, these companies only measure the first three concepts in the previous table: Conversion, Revenue, and Gross Margin (which is the same as Net Margin when attention in owned channels is given for free). Bringing up opportunity costs in this scenario will ruin the party for everyone. After all, the numbers that they track look great!
Thus, opportunity costs are doubly problematic: They are often implicit, and many teams have zero incentives to think about them. Yet the difference between accounting’s Net Margin and Economic Margin – between firms that “think inside the spreadsheet” and “outside the spreadsheet”, or that are finance led versus business led – has substantial implications for long-run firm performance.
What is to be done?
As a leader in the executive team, how do you know whether your firm is tracking net margin versus economic margin? Below are some quick diagnostic heuristics:
-
First, ask your “growth teams” if they are personalizing all in-product and owned channel promotions. In the likely event that inventory is given away for free, ask also if there is any effort to measure opportunity cost, even if it is as crude as benchmarking to market prices for attention. If neither is true your organization may not be set up to maximize yield on assets, even if, on paper, it is killing it on accounting profits.
-
Second, are your data science and finance teams familiar with the notion of opportunity cost and economic margin? One way to know is to go to your company’s HR website (you can do this right now) and look up the role descriptions or job adverts for finance and data science roles. Next, search the text for “opportunity cost”. My guess is you’ll find zero results, even for concepts like “economics”, “price elasticities”, or “decisions making under uncertainty” that might suggest awareness of opportunity costs. You will, however, find plenty of results for terms like “machine learning”, “big data”, “Spark”, and so on. If so, you may have a problem with people (skill gaps), process (hiring strategy, hiring criteria, economic objectives), and technology (cost and outcome measurement).
-
Third, ask how your firm allocates its owned acquisition channels. Are they allocated like in the former USSR, via central planning committees and interpersonal bargaining, or like in a market economy, via the price system? My sense is that in the overwhelming majority of Fortune 500 firms it is done via the former. This, too, is a people and process problem. The technical aspects of efficient allocation systems, like the price system or auctioning, have long been known. Some innovative companies, like Marriott, are experimenting with owned retail media networks to better allocate owned attention.
Ironically, if you have any of the above people, process, or technology problems, then you likely don’t have a technical targeting problem. Not because the problem does not exist but because it is not even recognized as such by the organization. In fact, the organization may perceive targeting as the problem, and blasting all customers the solution. Such incentives likely contribute to enshitification, and, quite possibly, the slow but inexorable decline in your business despite rosy topline numbers in the short- to medium-term.
Moreover, because the strategy clearly generates accounting Net Margin, whilst opportunity cost and economic margin remain unmeasured, you may have a hard time getting the lack of targeting to be recognized as a problem. Indeed, your concern for needlessly annoying customers with promotional blasts may get you labelled as soft: “Don’t you know we are a business? Our role is to make money! Get on with it and stop worrying!” might holler the cynics.
“A cynic is a man who knows the price of everything, and the value of nothing.”
– Oscar Wilde
As a data scientist, beware. Before diving into technical solutions, take a moment to check that your executive leadership team perceives targeting as an opportunity, and not as the problem to be avoided.
As an investor, next time you are in a earnings call and some leader reports topline growth excluding acquisition costs, ask:
- What percent of acquisitions come from owned channels?
- How are owned channels priced and allocated internally?
If owned channels drive a lot of acquisitions, and if they are given away for free, then worry that your capital is not being put to its best use, no matter how rosy the reported numbers. Hundreds of millions of dollars, maybe billions, may hinge on the answer to these questions.
Conclusion: Unlocking Long-Term Profitability Through Smarter Decisions
Treating owned channels as “free” is a costly illusion that undermines targeting, misallocates resources, and erodes long-term profitability. The problem is insidious: conversions, revenue, and gross margins may look great on paper, masking the deeper inefficiencies caused by ignoring opportunity costs and economic margins. Over time, this approach leads to stagnation and decline, even as topline metrics continue to paint a rosy picture.
The choice is clear: recognize the hidden costs, embrace smarter allocation strategies, and unlock the full potential of your business assets; or risk falling behind in a competitive landscape that rewards those who think beyond the spreadsheet.