Sixteen companies. What happened when the source of audience growth they relied on stopped working.
01 / What history says
Finding another source of traffic usually bought time. It did not rebuild the business.
Across the completed cases, companies that lost the channel that built their audience rarely solved the problem by finding another platform or traffic source.
The companies that emerged stronger generally did one of three things.
Each required changing the business around the audience, not just finding another place to reach it. These transitions took years, not quarters.
The same four steps show up again and again, and they arrive in this order.
Traffic or reach starts becoming less reliable. Facebook changed its feed. Google started answering questions itself.
The company monetizes the remaining traffic harder, buys traffic, or diversifies where it comes from. BuzzFeed held revenue for years. NerdWallet grew it 22% in its worst search year.
It buys traffic, pushes harder on conversion, or finds other ways to reach customers. The business can still grow, but more of that growth now depends on distribution it has to pay for or rebuild. People Inc. watched Google fall from 60% of its traffic to 24%.
Once acquisition economics deteriorate far enough, the weakness finally shows up in the financials. BuzzFeed’s advertising, content and commerce lines all fell in the same quarter. Vice went bankrupt. LittleThings shut.
This is why traffic alone is a late and often misleading signal.
Three lessons carry most of the weight in this history.
New York Times · Consumer Reports · ESPN
The strongest survivors did not just find more people. They created something people deliberately returned to.
For the Times, that meant products such as Games, Wordle, Cooking and sports, which get used on a schedule rather than read once. It bought Wordle for a low seven-figure sum and it brought in what the company called tens of millions of new users, many of whom stayed to play other games. It bought The Athletic for $550m the same month.
Consumer Reports had something even more valuable: people already had a paid direct relationship with the company. Because that relationship was already in place, weakening discovery mattered less to the underlying revenue model. Membership is about 70% of a $266m business, with no advertising at all.
ESPN is the open version. Its app is a place people already open on game days, and it launched direct to consumers in August 2025. Whether that replaces what cord-cutting took is not yet known.
The Free Press · Consumer Reports · The Athletic · Barstool as the contrast
Millions of followers can disappear behind a single platform decision. An email subscriber, registered user, paying member or app user is different. That is someone the company can reach again without starting from zero.
The Free Press started from nothing in 2021 and by October 2025 had 1.5 million subscribers, more than 170,000 of them paying. Paramount bought it for roughly $150m and made its founder editor-in-chief of CBS News. A broadcaster with a hundred-year-old news operation bought a four-year-old list rather than building one.
The Athletic’s Pulse newsletter grew from 2.4 million to more than 3.5 million subscribers in a year. Consumer Reports has more than five million paying members.
The contrast is Barstool. Penn Entertainment acquired one of the largest young male audiences in America and reached blended US sports betting share below 5% before writing off up to $850m. The attention was real. The ability to move those people to a specific decision was not.
Sports Illustrated · Britannica · People Inc. · NerdWallet · WW as the late version
Several companies survived by changing what they sold.
Some moved the brand to where the money was. Sports Illustrated’s owner licensed the name into ticketing, events and naming rights. Ticketing alone is expected to generate around $500m in 2026, and less than half of the brand’s revenue now comes from media at all. Nobody rebuilt the audience.
Some sold something other than media. People Inc. built an advertising product that runs on other publishers’ sites, which means part of its answer was to be less of an audience business. NerdWallet is moving from sending clicks to brokering the transaction itself.
Britannica ultimately stopped trying to recover the household customer. It rebuilt around schools, libraries, software and licensing instead.
WW tried to move closer to the transaction by buying a telehealth business, but the shift came after the core model had already deteriorated and much of the medication economics still sat elsewhere.
Often preserved volume temporarily but left the company dependent on someone else’s distribution.
LittleThings is the clearest example. It adapted repeatedly, and became more dependent each time. The January 2018 feed change cut its traffic 75% and the company shut a month later.
Consumer Reports tried this with Consumerist for nine years and eventually shut it. Those readers did not convert upward into the paid product.
These cases do not provide enough evidence to support that conclusion either way. See the note in Limits.
When the traffic source that built an audience weakens, finding another traffic source can keep revenue going for a few more years. But unless the company creates a direct reason for people to return, builds a relationship it can reach again, or changes how it makes money, the audience problem eventually becomes a revenue problem.
The completed failures suggest this takes roughly three to six years. Upworthy ran about three years from the algorithm change to its sale. LittleThings about four from peak to shutdown. BuzzFeed about five. Vice five to six. That range is directional rather than a forecast, and the dates for Vice and Upworthy are not crisply established in public sources.
There is also a slower path that the sentence above does not cover. A company can give up on the original customer and sell to a different one, which is what Britannica did. That took about twenty years.
NerdWallet is the case to watch. If its registered users and direct relationships allow customer acquisition costs to fall over time, this study’s central conclusion should be revised.
06 / The number to watch
What does it cost to acquire the next customer?
A company can lose free distribution and still report healthy revenue, because it buys traffic, monetizes harder, or shifts toward higher-value customers. That can hide the underlying problem for years.
The useful internal view is customer acquisition cost by source, compared with the long-term value of customers from that same source.
Conceptual pattern · not measured data
Acquisition cost rising while revenue stays flat or grows.
That is the warning sign the historical cases suggest appears before the financial damage becomes obvious. In every completed failure here, the company could still produce the same revenue at the point where the decision should have been made.
For the operating team: track CAC and three-year value by acquisition source and customer cohort. This is a direction the cases support, not a measured threshold. Nothing here says how fast the cost has to rise, or how far.
Stop treating traffic as the main health metric
Track what it costs to acquire a customer from each source, and what that customer becomes worth.
Build one reason for people to return every week
Treat it as a product, not another stream of content.
Turn anonymous visitors into people you can reach again
Registration matters only if those users return and become economically more valuable.
Use partners that already see the customer at the right moment
A strong distribution partner should have a reason to integrate even without a referral payment.
Start before the revenue breaks
The historical transitions took years. Waiting until revenue declines leaves much less time to build the replacement.
These are the sixteen businesses behind the findings above.
| Company | Channel it relied on | What changed | What it did | Outcome | |
|---|---|---|---|---|---|
| 01 | Encyclopaedia Britannica | Door-to-door sales force | CD-ROM, then web | Late CD-ROM, then schools and libraries | Changed the customer |
| 02 | BuzzFeed | Facebook referral | Algorithm shifts 2016–18, platform retreat 2023 | Commerce, creator network, acquisitions, AI | Collapse |
| 03 | LittleThings | Facebook News Feed | Jan 2018 News Feed change | Repeated tactical adaptation | Shut down |
| 04 | Upworthy | 2014 clickbait crackdown | Original video, then sold | Diminished | |
| 05 | Vice Media | Social and YouTube | Social referral decline | TV, agency, studios | Bankruptcy |
| 06 | People Inc. (Dotdash Meredith) | Google organic | AI Overviews | Ad tech, licensing, brand-led rename | Revenue held, traffic did not |
| 07 | NerdWallet | Google organic | AI Overviews and LLMs | Paid acquisition, registered users, vertical integration | Open |
| 08 | Chegg | Question-shaped search | AI answers | Own AI product, then a different customer | Decline |
| 09 | New York Times | Print decline, then referral decline | Bundle of habit products | Success | |
| 10 | The Athletic | Direct subscription | Never cleared alone | Acquired into a bundle | Success, inside someone else’s |
| 11 | ESPN | Cable bundle | Cord-cutting | Direct-to-consumer app | Open |
| 12 | Sports Illustrated | Mass print circulation | Print collapse | Brand licensing into transactions | Brand survived, audience business did not |
| 13 | Morning Brew / Money with Katie | Email newsletter | Diversification, not crisis | Acquired a personality-led brand | Reverted after four years |
| 14 | The Free Press | None (built new) | — | Subscription-first, personality-led | Acquired at $150m |
| 15 | Consumer Reports | Paid membership | Free ratings, then AI answers | Separate youth brand, then brand campaign | Membership held, youth brand failed |
| 16 | WW (Weight Watchers) | In-person meetings | GLP-1 drugs | Bought a telehealth clinic | Chapter 11 |
At its 1990 peak Britannica sold more than 100,000 sets and generated roughly $650 million in revenue through a door-to-door sales force of around 2,500 people. Reported Print set sales fell to about 51,000 in 1994 and roughly 10,000 by 1997. Reported The company priced its own CD-ROM at $995 against Encarta’s $50–70, and bundled it free with the print set to avoid a sales force revolt. Reported That protected the channel and did not stop the decline. Print ended in 2012, at which point the print set was about 1% of the business. Fact (company president, HBR, 2013)
The company that survived sells education software to schools and libraries, owns Merriam-Webster, runs an AI unit, and licenses its reference database. Revenue was around $100 million in 2022 with reported pro forma margins near 45%, and the company filed confidentially for an IPO in January 2024 seeking about $1 billion. Reported The consumer household never came back. Britannica did not rebuild distribution to the old customer. It found a new one.
BuzzFeed was taking more than half its traffic from platforms in 2017. Reported (Nieman Lab) Facebook engagements on BuzzFeed stories fell from 329 million in 2016 to under half that by 2018. Reported (NewsWhip via The Verge) By March 2023, BuzzFeed’s social referrals were down 81% from April 2020 and its Facebook referrals down 92%. Estimate (Similarweb)
The response ran through commerce, a creator network, acquisitions, and AI-generated quizzes. BuzzFeed News shut in April 2023; Peretti’s memo said he had been slow to accept that the platforms would not fund journalism built for them. Fact Valuation went from $1.7 billion in 2016 to $237 million, and market capitalisation was under $35 million by November 2023. Reported In Q3 2025 all three revenue lines fell at once: advertising $22.2m (-11%), content $7.2m (-33%), commerce $17.0m (-15%), with US time spent at 68.5m hours against 80.3m a year earlier. Fact (8-K)
Peaked at 58 million uniques in under four years, entirely on Facebook. Reported The company was unusually good at adapting: it rewrote headlines through the 2016 clickbait purge, A/B tested relentlessly, used dark posts, and pivoted to live video when Facebook wanted live video, to the point where 75% of show views came from live. Reported (Digiday) The January 2018 News Feed change cut organic and influencer traffic by 75%. Fact (CEO memo) Acquirers walked mid-process. The company shut in February 2018, putting 100 people out of work. Fact
This is the most instructive case here, because repeated successful adaptation to one platform is what killed it. Every adaptation deepened the dependency.
Fastest-growing media site of its moment, built on Facebook, hit by the 2014 clickbait crackdown. Reported Response was original video, then sale into GOOD Worldwide roughly three years later. The brand persists at far smaller scale. Evidence quality here is weaker than the rest of the set and it is included for the trigger and the timing, not for a precisely measured outcome. Estimate
Valued at $5.7 billion in 2017, filed for bankruptcy in May 2023. Reported Across the top 100 news and media domains, social referrals fell 55% between April 2020 and March 2023. Estimate (Similarweb) Vice’s response was expansion into television, an agency, and studios, none of which restored the audience.
The best-executed traffic diversification here. When IAC merged Dotdash and Meredith in 2021, about 60% of combined traffic came from Google. Fact (CEO, IAC earnings call) That fell to just over a third by Q1 2025 and to 24% by Q3 2025. Fact Core sessions fell 2.7% in Q1 2025 and AI Overviews appeared on roughly a third of search results for their content. Fact
They signed an OpenAI licensing deal in May 2024, built D/Cipher, a first-party contextual ad product later extended to third-party sites, and renamed the company after its biggest consumer brand. Q1 2025 performance marketing was up 11% with affiliate commerce up 26%. Fact Digital ad revenue still fell 3% in Q3 2025. Fact
The important detail is what they actually replaced. They did not replace the traffic. They replaced the dependency by building an ad network that monetises other publishers’ inventory, which means part of the answer was to be less of an audience business.
The case that most complicates the finding. Organic search visibility had a “pretty brutal” quarter in Q3 2024 and credit cards revenue fell 16% to $45.3m. Fact (CEO, earnings call) Full-year 2025 revenue still grew 22%, with Q4 up 23%, as performance marketing, direct, and non-search referral more than offset organic declines. Fact The CEO said LLM referral traffic appears incremental and converts at much higher rates. Fact
Through 2026 the company has been explicit about the strategy: an AI transition changing how people get money answers, and a response built on owned audiences, user registration, and vertical integration rather than a click-only marketplace. Fact (Q1 and Q2 2026 calls) It acquired College Finance in February 2026 for $17m cash, is building a framework to route insurance calls to agents, and is now investing marketing against internal IRR targets for sticky audiences rather than in-quarter profitability, at a cost of $15–20m to 2026 non-GAAP operating income. Fact Q2 2026 revenue was $197m, up 6%, with credit cards still falling on organic search pressure. Fact Trailing twelve-month adjusted free cash flow doubled to a record $141m. Fact
Unresolved Revenue has held. The acquisition engine has not been replaced; it has been rented.
Covered in Study 002. Non-subscriber traffic fell roughly 49% year over year as AI Overviews expanded; revenue and subscribers fell roughly 30%; the company responded by building its own AI product and then pursuing a different customer. Reported Included here as a bridge, not re-litigated.
The clearest success and the one most often misread. The replacement for print was not better journalism reaching more people. It was a bundle of things people use on a schedule. Wordle cost a low seven-figure sum in January 2022 and brought what the company called an unprecedented tens of millions of new users, many of whom stayed for other games. Fact The Athletic cost $550m the same month. Fact The Times disbanded its own sports desk in July 2023 and handed coverage to The Athletic. Fact
By Q2 2025 there were 11.3m digital-only subscribers, about 6.02m of them on bundle or multiproduct plans, digital-only ARPU $9.64, digital subscription revenue up 15.1%, print subscription revenue down 2.8%. Fact The deceleration is worth naming: net adds were 450,000 in Q4 2025, but bundle and multiproduct starts fell from 320,000 in Q4 2024 to 210,000 in Q4 2025, and Q2 2026 missed on subscribers. Fact / Reported The engine works and it is slowing.
Founded 2016, raised about $140m, reached roughly 1.2m subscribers at about $72 a year and around $65m revenue by the end of 2021, having spent close to $100m against $73m of revenue across 2019–20. Reported It never made money on its own. It was acquired for $550m, lost $6.8m in its first two months inside the Times, $11.3m in Q1 2023, $8.7m in Q1 2024, and turned profitable in 2024 after being integrated into the bundle. Fact / Reported
By May 2025 it had passed 6 million subscribers, mostly through the bundle, and the Times had stopped disclosing standalone numbers. Its Pulse newsletter went from 2.4m to more than 3.5m subscribers in a year. Reported (Axios) This is the best-run direct-destination subscription build here and it did not clear its own economics.
Lost close to half its pay-TV subscribers over fifteen years, from nearly 100 million households paying about $4 a month in 2010 to roughly 60 million paying more than triple that. Reported The direct-to-consumer app launched 21 August 2025 with around 25 million subscribers migrated from ESPN+, many of whom already had cable. Fact / Reported Ratings for the first half of 2026 were the best since 2012. Fact (Pitaro)
Disney’s sports segment operating income ran $2.7bn (2022), $2.5bn (2023), $2.4bn (2024), $2.9bn (2025), then fell 23% year over year in Q1 2026, partly on a YouTube TV distribution dispute costing about $5m a day. Reported Asked directly whether app subscriptions were outpacing cord-cutting, Pitaro did not answer. Reported That non-answer is the most informative data point in the case. Unresolved
Licensed to Arena Group in 2019. Arena missed a quarterly licensing payment of about $3.75m, published AI-generated product reviews, and had the licence terminated by Authentic Brands on 18 January 2024. Fact (SEC filings) The media business was discontinued on 18 March 2024. Fact (10-Q)
Print circulation is now reported at around 400,000 and Authentic says less than half of Sports Illustrated’s revenue comes from media. Reported (LA Times, 2026) SI-branded ticketing is expected to generate around $500m in 2026, alongside stadium naming rights, events, and a FAST channel up 60% since launch. Reported Nobody rebuilt the audience. The name was moved into transactions and the audience business became the smaller half.
Money with Katie was doing roughly $250,000 in year two when Morning Brew acquired it at the end of 2021 in an acquihire that bought the social accounts, podcast, email list, website, branding, and IP, with no equity for the founder. Fact (founder’s own account)
Morning Brew built a creator network around this model: the Morning Brew Daily podcast passed 50 million downloads and about $5m of annual revenue in under two years, up over 700% in 2024, and Good Work reached 1.16m YouTube subscribers. Reported More than 60% of Morning Brew’s audience engagement now happens off email. Reported (Axios)
On 1 January 2026 the founder bought Money with Katie back and took it independent to Substack, monetising through direct sponsorships negotiated off-platform. Fact Four-year hold, full reversion, and the list went with her.
Founded 2021 with nothing. By October 2025 it had 1.5 million subscribers, more than 170,000 of them paid, revenue up 82% over the trailing twelve months and subscribers up 86%. Fact (Paramount press release) Paramount Skydance acquired it for roughly $150m in cash and stock and made Bari Weiss editor-in-chief of CBS News, reporting directly to the CEO. Fact
A broadcaster with a hundred-year-old news operation bought an audience relationship four years old. It did not build one.
The cleanest test of the separate-youth-brand strategy here. Consumer Reports bought Consumerist.com from Gawker at the end of 2008 specifically to attract younger readers who might convert to paid subscriptions, and shut it in October 2017. Reported (CJR) Over that period revenue from subscriptions, newsstand and other sales fell from $234.2m (FY2013) to $205.5m (FY2017), down more than 13%. Reported
What held was the paid relationship. The organisation reported $266m of revenue and about $1.7m of net income in its March 2025 filing, with membership around 70% of total revenue, more than five million members, and no advertising. Reported In March 2026 it committed about $3m to its largest brand campaign in five years, with stated KPIs of lower customer acquisition cost, newsletter signups, and membership conversion. Reported The membership model shields revenue as traffic declines. The separate youth brand did not.
Distribution was in-person meetings for sixty years. GLP-1 drugs removed the reason to attend. Revenue fell from about $1.04bn in 2022 to $711m in 2025, with a 2024 net loss of $345.7m and subscribers down 14% year over year. Reported / Fact
The response was to buy Sequence in 2023 and become the place the prescription happens. Clinical revenue grew 55% in one quarter. Fact (8-K) The medication revenue flows around WW to pharmacies. Reported Chapter 11 was filed 6 May 2025, ran 42 days, eliminated $1.15bn of debt, and the company emerged private on 24 June 2025. Fact Q4 2025 total revenue was still down 12%. Fact
Moving toward the transaction after the channel broke was the right instinct and it arrived late and captured the wrong part of the money.
Five cases push against the conclusion above. They are not filed as exceptions.
NerdWallet may prove that paid acquisition plus registered users can become a durable replacement. It is too early to know. Revenue grew 22% during a steep search decline, which should not happen if buying traffic never works. The reconciliation is that it moved money from free traffic to paid traffic and moved closer to the transaction at the same time. But bought traffic is rented, and management has effectively said so by shifting its marketing targets toward long-term returns rather than in-quarter profit. If acquisition costs for its direct user base fall over time, the central finding should be revised.
Losing search does not have to be fatal to a search-dependent publisher. Digital revenue grew while Google’s share of its traffic collapsed. But the mechanism was building an advertising business that runs on other people’s sites, which is a decision to be less of an audience business. That is a real answer, and it is not the answer most publishers think they are choosing.
This is the strongest argument against building your own destination. It raised $140m, hired the best writers in the market, executed well, and never made money independently. It became profitable only inside somebody else’s bundle. If a direct destination is the plan, this case says it needs more than one reason to exist.
You may not have to reach the next generation at all. Britannica never got its household customer back and reached roughly 45% margins anyway by selling to schools and libraries. Sometimes the right answer to “how do we reach the next generation of this customer” is that you should not try.
The audience may not be the asset. The brand is now reportedly more than half a non-media business. The audience business shrank and the name became worth more.
Limits
These sixteen skew toward companies whose outcomes were public and toward failures loud enough to be reported. Businesses that quietly rebuilt distribution are underrepresented, and that bias works against this study’s conclusion rather than for it.
Four outcomes are still open: NerdWallet, ESPN, People Inc., and WW after bankruptcy. NerdWallet is the one most capable of overturning the central finding.
On creators and short-form video: no case here shows a legacy publisher building durable economics on short-form video after losing a channel, but that is an absence of evidence within sixteen publicly documented cases rather than a demonstration that it cannot work. The bias runs in an identifiable direction, since companies that quietly built working video operations have no reason to disclose the economics.
None of these companies consistently disclose audience age, so any claim about reaching younger audiences rests on what was disclosed rather than on measurement. Several traffic figures come from third-party measurement services that the affected companies have publicly disputed. Upworthy carries the weakest evidence base and is included for its trigger and timing rather than for a measured outcome.
Fact filing, press release, or on-record statement. Reported credible third-party reporting. Estimate third-party measurement or modelled figure. Unresolved outcome still open.
The condition for inclusion was structural, not thematic: an established audience or brand, a distribution channel that materially weakened, a documented attempt to build a replacement, and enough subsequent evidence to say what happened. Companies were not included because they involved creators, short-form video, or Gen Z.
BuzzFeed Inc. Form 8-K, Q3 2025 earnings exhibit (SEC). Arena Group Holdings Form 10-Q Q1 and Q2 2024, Form DEF 14A 2024, Form 8-K Q4 2025 and Q2 2026 (SEC). WW International Form 10-K FY2025, Form 8-K FY2025 and FY2026 (SEC). Paramount press release, 6 October 2025. ESPN Press Room, 21 August 2025. Encyclopaedia Britannica / Businesswire, 14 March 2012.
NerdWallet Q3 2024, Q4 2025, Q1 2026, Q2 2026. IAC / Dotdash Meredith / People Inc. Q1 2024, Q1 2025, Q2 2025, Q3 2025 (Neil Vogel on Google traffic share). New York Times Q4 2023, Q3 2024, Q2 2025, Q4 2025. Disney Q1 2026 sports segment. Jimmy Pitaro, ESPN, 2026.
Nieman Journalism Lab on BuzzFeed News, April 2023. Digiday on LittleThings, February 2018 and retrospective. Adweek and TechCrunch on LittleThings shutdown. CNN Business on BuzzFeed News. CNBC on Meta’s retreat from news, January 2024. Press Gazette on Facebook referral decline and on The Free Press. Columbia Journalism Review on Consumer Reports and Consumerist. Adweek on the Consumer Reports 2026 brand campaign. Axios on Wordle, on The Athletic’s profitability, and on the Paramount deal. Harvard Business Review, Britannica president, March 2013. Bloomberg and Crain’s Chicago Business on the Britannica IPO. Los Angeles Times via Awful Announcing on Sports Illustrated under Authentic. A Media Operator on NYT bundle economics and on The Athletic. AdExchanger on Dotdash Meredith and People Inc. Sportico on ESPN Unlimited.
Similarweb social referral analysis, May 2023. Chartbeat and Parse.ly Facebook referral data. Digital Content Next member survey on AI Overviews, 2025. Related studies: 002 (replacing a broken acquisition channel), 005 (what a trusted audience is worth), 022 (the reason to return).