The tragedy of the commons in marketing: shared attention and the cost of one more message

Your customers' attention is a field you share with everyone else

Your customer has one inbox, one feed and one page of search results, and you own none of them. Every shop that wants the same customer sends into the same inbox and bids into the same feed. When all of you push a little harder, each of you gets a little less back.

Economists have a name for this pattern. It is the tragedy of the commons, and it explains why a channel that worked beautifully three years ago can feel worn out today, even though nobody in particular did anything wrong.

In this article I take the idea from the original source, show where it goes wrong, and do the arithmetic with four invented shops sharing one audience. The businesses and numbers in the worked examples are made up so I can show every step. Nobody real, no client.

If payoff matrices and the Nash equilibrium are new to you, my article on game theory in pricing builds them from scratch. I lean on it here rather than repeat it.

Hardin's pasture: one more animal, a fraction of the cost

In 1968 the ecologist Garrett Hardin published a short essay in Science called The Tragedy of the Commons.¹ Picture a pasture open to all. Each herdsman asks a simple question: what is the utility to me of adding one more animal to my herd?

Hardin's answer has two parts. The herdsman gets all the benefit of the extra animal, so the positive utility is nearly +1. The cost of overgrazing is shared by everyone who uses the pasture, so the negative utility for any one herdsman is only a fraction of minus 1.¹ Add one, and the sum is positive. So he adds another, and another. Hardin's conclusion was bleak: "Freedom in a commons brings ruin to all."¹

Hardin was mostly writing about population, not pastures.² But he also turned the idea round. Pollution, he wrote, is the tragedy of the commons in reverse: instead of taking something out, people put something in. One of his own examples was advertising signs placed in the line of sight of travellers.¹ So advertising was a commons problem from the start.

Worked example: one more animal, in euro

Ten farmers share a field. Each extra cow earns its owner €100 a year in milk. But every cow also eats grass the others needed, and the field produces €50 a year less for everyone put together. That €50 loss is spread across all ten farmers.

Step one, the owner's gain: €100.

Step two, the owner's share of the loss: $€50 \div 10 = €5$.

Step three, the owner's net: $€100 - €5 = €95$. Clearly worth it.

Step four, the field's net: $€100 - €50 = €50$. Still positive, so far no harm done.

Now suppose the field is fuller and each extra cow costs the field €150 a year. The owner still gains $€100 - €15 = €85$. The field as a whole comes out at $€100 - €150 = -€50$, a loss of €50. The farmer adds the cow anyway, because from where he stands it pays. That gap between what the decision is worth to one person and what it is worth to everyone is the whole tragedy.

Where Hardin went wrong, and why it matters for you

Hardin's story is memorable, and only half right. Two criticisms change what you should do.

The first is that Hardin mixed up two things. A true commons is usually shared by a defined group under rules. What he described was open access: anyone may come, nobody sets limits. Looking back after 50 years, Frischmann, Marciano and Ramello name this as one of the essay's two conceptual mistakes, the other being to confuse the resource with how it is governed.²

The second criticism came from the field. Elinor Ostrom spent decades studying real fisheries, pastures, forests and irrigation systems. In Governing the Commons she showed that neither state control nor privatisation consistently solved these problems, and that local groups sometimes manage shared resources well without anyone coercing them from outside.³ In 2009 she shared the economics prize in memory of Alfred Nobel "for her analysis of economic governance, especially the commons".⁴

From that field work Ostrom drew eight design principles that long lasting commons tend to share. In short:⁵

1. Clear boundaries, both around who may use the resource and around the resource itself.
2. Rules that fit local conditions, with costs in proportion to benefits.
3. The people affected by the rules can take part in changing them.
4. Monitoring, by monitors who answer to the users.
5. Graduated sanctions, mild for a first slip and heavier for repeat or serious breaches.
6. Quick, low cost ways to settle conflicts.
7. Outside authorities recognise the users' right to make their own rules.
8. For larger systems, governance nested in layers.

A review of 91 studies by Cox, Arnold and Villamayor Tomás found these principles well supported by the evidence.⁵

For a shop, the lesson is that the tragedy is not fate. Overuse happens where nobody sets and enforces limits.

Worked example: open access against a governed commons

Take the field from before, with ten farmers and a cow that earns €100 but costs the field €150.

Under open access, every farmer adds the cow, because each sees a private gain of €85. With all ten doing it, the field loses $10 \times €50 = €500$ a year net.

Now give the field Ostrom's kind of rules: a limit of four cows each, farmers taking turns to count, a €20 fine for a first breach and a lost season for a second. The extra cow still earns €100 and costs its owner €15 in shared loss, but now it also carries a real chance of a fine and a lost season. For most farmers that tips the sum negative, and the field keeps its grass.

Attention is the commons every marketer grazes

The resource is a person's attention. It renews, like grass, but only so fast. Push more messages in than a customer can take and the extra ones get skimmed, ignored, unsubscribed from or reported.

The herdsmen are the senders. Each gets the full benefit of their own extra message. The cost is spread across everyone the customer hears from, because a tired customer opens fewer emails from all senders and scrolls faster past all ads.

Three commons matter most: the inbox, which every shop you ever bought from can fill, the feed, where every advertiser chasing your town or interests bids for the same few slots, and the search results, where every site wants the same query. None of the three is open access any more. Each has a manager with rules, as I show later. But first, the maths of why senders overgraze when nobody stops them.

The maths: each sender gains, everyone pays

A shared audience receives $N$ marketing emails a week in total, from all senders put together. Each email sent to the whole list brings in a number of orders that falls as the inbox gets fuller:

$$q(N) = a - bN$$

Here $q(N)$ is the number of orders one send brings in when the audience receives $N$ sends a week in total, $a$ is the number of orders a send would bring in an empty inbox, and $b$ is how many orders each send loses for every extra email the audience receives that week. A straight line is the simplest shape that captures diminishing returns (Foundations: regression explains straight line models). Real response curves bend, but the logic is the same.

Shop $i$ sends $n_i$ of the $N$ emails. One more send changes its own weekly profit, and the combined profit of all the shops, by

$$\begin{aligned} \Delta \pi_i &= m \left[\, q(N+1) - b\, n_i \,\right] \\ \Delta \Pi &= m \left[\, q(N+1) - b\, N \,\right] \end{aligned}$$

Here $\Delta \pi_i$ is the change in shop $i$'s weekly profit from sending one more email, $\Delta \Pi$ is the change in the combined weekly profit of every shop, $m$ is the margin in euro on one order, $q(N+1)$ is the orders the new send brings in once the inbox holds $N+1$ emails, $b\,n_i$ is the orders the shop loses on its own existing $n_i$ sends because the inbox got fuller, and $b\,N$ is the orders lost across all $N$ existing sends from every shop. These are marginal changes: the effect of the next send, not the average of all of them (Foundations: marginal versus average).

The two lines differ in one place. The shop feels the damage on its own $n_i$ sends, the group on all $N$. The ratio $\frac{n_i}{N}$ is Hardin's "fraction of minus 1", written down. Whenever $q(N+1)$ sits between $b\,n_i$ and $b\,N$, the next send pays the sender and costs the group.

Worked example: four Cork shops and one list

Four independent shops in Cork sell gifts and homeware: Shandon Ceramics, Ballintemple Plant Room, Douglas Paper and Blackrock Candles. Over the years the same 20,000 local customers have signed up to all four newsletters. I set the numbers so the arithmetic stays clean:

• In an empty inbox, one send to the 20,000 brings in 100 orders, so $a = 100$.
• Every extra email the audience receives in a week costs each send five orders, so $b = 5$.
• The margin on an order is €20, so $m = €20$.

So one send brings in $q(N) = 100 - 5N$ orders.

Step one, everyone sends once a week. $N = 4$. Each send brings in $100 - 5 \times 4 = 80$ orders. Each shop earns $80 \times €20 = €1{,}600$ a week. Together: €6,400.

Step two, everyone sends twice. $N = 8$. Each send brings in $100 - 40 = 60$ orders. Each shop gets $2 \times 60 = 120$ orders, so €2,400. Together: €9,600. More email helped everyone.

Step three, the temptation. Shandon Ceramics adds a third weekly send while the others stay at two. Now $N = 9$ and each send brings in $100 - 45 = 55$ orders. Shandon gets $3 \times 55 = 165$ orders, 45 more than before, so €900 more a week. Each of the other three gets $2 \times 55 = 110$, ten fewer, so €200 less each. Check with the formula: $\Delta \pi_i = €20 \times (55 - 5 \times 2) = €20 \times 45 = €900$. For the group: $\Delta \Pi = €20 \times (55 - 5 \times 8) = €20 \times 15 = €300$. Still positive, so this extra send was fine for the field too.

Step four, everyone copies. All four send three a week. $N = 12$, each send brings in $100 - 60 = 40$ orders, each shop gets 120 orders. Exactly what they had at two sends, for 50% more work.

Step five, the tragedy. Shandon tries a fourth send. $N = 13$, each send brings in 35 orders. Shandon gets $4 \times 35 = 140$ orders, 20 more than at step four, so €400 more a week. The other three each get $3 \times 35 = 105$, 15 fewer, so €300 less each. The formulas agree: $\Delta \pi_i = €20 \times (35 - 5 \times 3) = €400$, and $\Delta \Pi = €20 \times (35 - 5 \times 12) = €20 \times (-25) = -€500$. The send pays Shandon and costs the four shops €500 a week together.

Step six, where it stops. If all four send four a week, $N = 16$, each send brings in 20 orders and each shop gets 80, back to €1,600 a week. Would any shop now add a fifth? It would get $5 \times 15 = 75$ orders, fewer than 80, so no. Would any shop drop to three? It would get $3 \times 25 = 75$, also fewer than 80, so no. Nobody gains by moving alone. That is the equilibrium, and it is where they end up.

Step seven, the bill. At two or three sends each, every shop earned €2,400 a week. At four sends each, every shop earns €1,600. Across the four shops that is $4 \times €800 = €3{,}200$ a week lost, about €166,000 a year, for twice the email of the best case. Every single step was rational. The destination was a tired list.

The horizontal axis is how many emails each shop sends a week, all four sending the same number. The vertical axis is orders per week, for one shop and for all four combined. Both lines peak at two to three emails each, and the equilibrium, four a week, sits on the falling side. Left alone, the shops stop where no single shop can gain by moving, not where the list works best.

The inbox has a manager now: Gmail and Yahoo since February 2024

From February 2024, Google and Yahoo both started enforcing rules for anyone sending large volumes of email to their users.⁶⁷

Google applies its strictest rules to senders who send more than 5,000 messages a day to Gmail accounts.⁶ Once a sender reaches that level, Google treats it as a bulk sender permanently.⁸ The rules, in plain words:

• Authenticate your mail with Sender Policy Framework (SPF), DomainKeys Identified Mail (DKIM) and Domain based Message Authentication, Reporting and Conformance (DMARC), so the inbox can tell who really sent it.⁶⁷
• Support one click unsubscribe in marketing messages.⁶ Google asks senders to honour unsubscribe requests within 48 hours, and Yahoo within two days.⁸⁷
• Keep the spam rate low. Google says to keep the rate reported in its Postmaster Tools below 0.10% and never reach 0.30% or more.⁶ Yahoo's limit is 0.3%.⁷
• Yahoo adds that you should only send to people who specifically asked for your mail, at the frequency they signed up for, and never buy lists.⁷

The spam rate in Postmaster Tools is the share of your messages delivered to engaged recipients' inboxes that those recipients then mark as spam.⁹ Bulk senders above 0.3% lose eligibility for Google's help until they stay below it for seven days in a row, and since November 2025 Google has been rejecting non compliant mail, temporarily or permanently.⁸

Read that list again with Ostrom's principles beside it. Authentication draws the boundary: who is a sender. Postmaster Tools is monitoring. A 0.1% guideline, a 0.3% line and then rejection are graduated sanctions. One click unsubscribe gives every reader a fast, low cost way to settle the conflict personally. The inbox found its manager.

The rules turn part of the shared cost into a private one. A complaint now counts against your own domain.

Worked example: is the fifth email worth the risk

Go back to Cork. Shandon Ceramics now sends four emails a week to its customers and wonders about a fifth, a clearance offer to people who have not opened anything in six months. It expects that fifth email to bring in €300 of margin. It also knows that unengaged readers are the ones most likely to press the spam button.

Its owner estimates two outcomes, using expected value (Foundations: expected value):

$$E[X] = \sum_i p_i\, x_i$$

Here $E[X]$ is the expected profit of sending the email, $p_i$ is the probability of outcome $i$, $x_i$ is the profit in euro if outcome $i$ happens, and the sum $\sum_i$ adds the products over all outcomes.

Step one, the good outcome. With probability 0.75 nothing goes wrong and the email earns €300.

Step two, the bad outcome. With probability 0.25 the complaints tip the domain over the line, more of the next fortnight's mail lands in spam, and the shop loses €2,400 of margin it would otherwise have made. The email still earned its €300, so this outcome is $€300 - €2{,}400 = -€2{,}100$.

Step three, the sum. $E[X] = 0.75 \times €300 + 0.25 \times (-€2{,}100) = €225 - €525 = -€300$.

On average the fifth email loses €300. The probabilities are guesses, but the shape of the sum is the point: a small certain gain against a modest chance of a large loss.

The feed wears out too: frequency and ad fatigue

The feed is a commons with a twist. The platform runs an auction for each slot, so the price of attention rises when more advertisers chase the same people (I explain the mechanics in my article on auction theory for paid media). On top of that, each person sees your own ad again and again, and every repeat earns less than the one before.

The platforms say so themselves. Google warns that rising cost per view can be a sign of creative fatigue once a video ad has run for a couple of weeks, and suggests rotating two or three ads.¹⁰ Google Ads also offers frequency capping, which limits how often your ads appear to the same person.¹¹ Neither page gives a magic number of views, and neither do I. The right cap depends on the audience, the offer and the creative.

Worked example: a Lahinch shop and its sixth impression

A Lahinch surfboard shop runs a feed campaign to 50,000 people in Co. Clare. It pays €6 per 1,000 impressions. Google itself warns that creative wears out with repeat views, so I give the click rate an invented falling pattern for this example: per 1,000 impressions, 12 clicks on the first view, then nine, seven, five, four and three by the sixth.

Step one, cost per click on the first view: $€6 \div 12 = €0.50$.

Step two, the same sum for each later view: $€6 \div 9 = €0.67$, then $€6 \div 7 = €0.86$, then $€6 \div 5 = €1.20$, then $€6 \div 4 = €1.50$, then $€6 \div 3 = €2.00$.

Step three, the comparison. If a click is worth €1.25 of margin to the shop, the first four views pay and the fifth and sixth do not. A cap of four per person, and the saved money spent on a fresh audience or a new ad, is the obvious move.

The horizontal axis is how many times the same person has seen the ad. The vertical axis is what each click costs at that view. The bars climb steadily: diminishing returns, made visible. Cap where the cost per click crosses what a click is worth to you.

Search results have a manager as well

The search results page is the third commons. If every site in a niche fills the web with near identical pages, the results get worse for everyone. Google manages this commons too. Its spam policies define scaled content abuse as generating many pages for the primary purpose of manipulating search rankings, and say that sites that break the policies may rank lower or not appear in results at all.¹² Again the shared cost becomes a private one for whoever causes it.

Worked example: 300 town pages in Galway

A Galway bathroom fitter is tempted by a quick plan: one page for each of 300 towns and villages across Connacht, each saying the same thing with the town name swapped.

Step one, the private gain the owner imagines: 300 pages that might each catch a few searches.

Step two, the cost to the commons: 299 pages that add nothing the first one did not, crowding out useful pages.

Step three, the manager's response: pages built mainly to manipulate rankings fall under the scaled content abuse policy, and the whole site risks ranking lower.¹²

Step four, the better plan: a handful of pages for the towns the fitter really serves, each with real jobs and prices from that town.

Irish law polices the inbox as well: the ePrivacy Regulations

In Ireland the inbox commons also has a legal fence. The ePrivacy Regulations (S.I. No. 336 of 2011) set the rules for marketing emails. They sit alongside the GDPR and the Data Protection Act 2018, and the Data Protection Commission (DPC) enforces them.¹³¹⁴

Regulation 13 is clear. You must not send a marketing email to an individual unless they have told you they consent.¹³ The one limited exception is for existing customers. If you collected someone's email address when you sold them a product or service, you may email them about your own products of a similar kind, as long as you gave them a clear, easy and free way to object when you collected the address, you give them one in every message, and the sale was no more than 12 months before the email (or you have used the address for marketing within that 12 months).¹³ Every marketing email must also say who sent it and give a valid address to reach the sender.¹³

The DPC's guidance adds two points that matter for a shop. Someone who browsed but never finished the checkout is not a customer, so the exception does not cover them. And where consent is needed, silence or a pre ticked box does not count.¹⁴ Sending marketing email to someone who has objected is a criminal offence.¹⁴ For a shop the core rule is short: no consent, no marketing email, unless the customer bought from you in the last 12 months and could say no.

Worked example: a Kilkenny shop and a list of 5,000

A Kilkenny kitchenware shop is offered a list of 5,000 local email addresses for €400. It also has 3,000 past customers who bought from it online in the last 12 months, were offered an opt out at checkout, and have not used it.

Step one, the bought list. The people on it never gave the shop consent and never bought from it. The existing customer exception only covers addresses a business collected from its own customers in a sale, so it cannot apply. Emailing them breaks regulation 13. The list is worth less than nothing.

Step two, the past customers. They bought within the last 12 months, they were given an opt out when the shop collected their address, and every email will carry an unsubscribe link. The shop may email them about its own similar products. If a customer goes more than 12 months without a purchase and without a marketing email from the shop, they fall outside the exception until they buy again.

Step three, the commons view. The legal list is also the healthier one, because people who chose you complain less, and complaints are what the inbox managers count. My article on consent and email capture shows how to grow it properly.

Your own customers are a commons inside your business

The commons that hurts you most is often inside your own walls. In a growing shop the newsletter, the promotions calendar and a good few automated flows all email the same customers. Each owner sees their own numbers, each extra message looks profitable, and nobody owns the total. That is Hardin's pasture, with your colleagues as the herdsmen.

This commons you can govern, and Ostrom's principles translate almost line by line:

1. Boundaries. Decide which messages count as marketing contact and which are service messages (an order confirmation is not marketing).
2. Rules that fit. A weekly contact budget per customer, higher for people who engage.
3. Participation. Everyone who sends to customers agrees the budget together.
4. Monitoring. One shared view of contacts, unsubscribes and complaints, per send.
5. Graduated sanctions. A send over budget loses its next slot. A flow that keeps drawing complaints is switched off.
6. Conflict resolution. One person settles clashes over the same day.
7. Recognition. The managing director backs the rules.
8. Nesting. Flows sit inside a weekly budget, which sits inside a monthly plan.

Agreeing with rival shops to limit marketing is a different matter, and competition law territory, a line I cover in strategic interaction models. Inside your own business, coordination is simply good management.

Worked example: a contact budget at a Waterford tea shop

A Waterford online tea shop has 12,000 subscribers and three senders.

Step one, count the total. Newsletter one a week, promotions three, automated flows three on average: seven emails a week per subscriber.

Step two, set the budget: four marketing emails a week for subscribers who opened something in the last 90 days, one for everyone else. Order confirmations do not count.

Step three, allocate. Newsletter one, promotions two, flows one, with flows taking priority on any day a customer has an open basket.

Step four, measure for a month: orders per subscriber, unsubscribes and complaints, against the month before.

Step five, decide. If orders hold and unsubscribes fall, the budget stays. If orders fall, test five a week for engaged readers. Either way, one person now owns the total, and that is the change that matters.

How I decide whether one more message is worth sending

This is the order of questions I go through before any new send, ad set or batch of pages.

Read it from the top. The legal question comes first, because no return justifies an unlawful message. Then the load on the audience, because that is the commons. Only then the marginal return, including the damage to my other messages: the $b\,n_i$ term from the formula.

Being the loudest sender is not the same as winning. Hardin was right that open access ends in ruin. Ostrom was right that it does not have to stay open access. The inbox, the feed and the search page all found managers. The one left for you to manage is your own.

Footnotes

1. Garrett Hardin, The Tragedy of the Commons, Science, volume 162, issue 3859, pages 1243 to 1248, 13 December 1968, full text reproduced by The Garrett Hardin Society, https://www.garretthardinsociety.org/articles/art_tragedy_of_the_commons.html

2. Brett M. Frischmann, Alain Marciano and Giovanni Battista Ramello, Retrospectives: Tragedy of the Commons after 50 Years, Journal of Economic Perspectives, volume 33, issue 4, pages 211 to 228, 2019, https://www.aeaweb.org/articles?id=10.1257/jep.33.4.211

3. Elinor Ostrom, Governing the Commons: The Evolution of Institutions for Collective Action, Cambridge University Press, 1990, Canto Classics edition 2015, https://www.cambridge.org/core/books/governing-the-commons/A8BB63BC4A1433A50A3FB92EDBBB97D5

4. The Royal Swedish Academy of Sciences, The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2009, press release reproduced by idw Informationsdienst Wissenschaft, 13 October 2009, https://idw-online.de/en/news338479

5. Michael Cox, Gwen Arnold and Sergio Villamayor Tomás, A Review of Design Principles for Community based Natural Resource Management, Ecology and Society, volume 15, issue 4, article 38, 2010, https://www.ecologyandsociety.org/vol15/iss4/art38/main.html

6. Google, Email sender guidelines, Gmail Help, accessed November 2026, https://support.google.com/a/answer/81126

7. Yahoo, Sender Best Practices, Yahoo Sender Hub, accessed November 2026, https://senders.yahooinc.com/best-practices/

8. Google, Email sender guidelines FAQ, Gmail Help, accessed November 2026, https://support.google.com/a/answer/14229414

9. Google, Postmaster Tools dashboards, Gmail Help, accessed November 2026, https://support.google.com/mail/answer/14668346

10. Google, tips for optimising Video campaigns, Google Ads Help, accessed November 2026, https://support.google.com/google-ads/answer/3013684

11. Google, Use frequency capping, Google Ads Help, accessed November 2026, https://support.google.com/google-ads/answer/6034106

12. Google, Spam policies for Google web search, Google Search Central, accessed November 2026, https://developers.google.com/search/docs/essentials/spam-policies

13. Government of Ireland, European Communities (Electronic Communications Networks and Services) (Privacy and Electronic Communications) Regulations 2011, S.I. No. 336 of 2011, regulation 13, Irish Statute Book, accessed November 2026, https://www.irishstatutebook.ie/eli/2011/si/336/made/en/print

14. Data Protection Commission, FAQ on Consent for Electronic Direct Marketing, April 2020, accessed November 2026, https://www.dataprotection.ie/en/dpc-guidance/faq-consent-electronic-direct-marketing