What Happens to a Placement's Traffic After the First Week?

What does the research actually show?

That attention to an online article is front-loaded to a degree most buyers would find uncomfortable, and that the decline begins almost immediately rather than after some period of steady performance.

The most useful study we located is Characterizing the Life Cycle of Online News Stories Using Social Media Reactions, by Carlos Castillo of the Qatar Computing Research Institute, Mohammed El-Haddad of Al Jazeera, Matt Stempeck of the MIT Center for Civic Media and Jürgen Pfeffer of Carnegie Mellon, published in 2013. It examined 606 articles from Al Jazeera English, tracking website visits alongside reactions on Twitter and Facebook.

Its central descriptive finding is blunt. The traffic to about 80% of articles decreased monotonically during the first twelve hours — not merely declined overall, but fell continuously from the start, without a plateau.

The paper describes the early hours as a burst: an initial wave of visits and social activity that “decays rapidly after a short time”. It also notes that after a few hours a large share of remaining visits is internal traffic — readers arriving from elsewhere on the same site rather than from the outside world.

How long is an article's half-life?

Shorter than a week, on the evidence available, and unevenly distributed.

The Castillo paper cites earlier work by Dezso and colleagues analysing visits to a large Hungarian news portal, which found article half-life distributed according to a power law across a broad range, with a mean of 36 hours. A power-law distribution matters as much as the average does: most articles are short-lived, and a small number persist far longer, which means the mean describes almost no individual article well.

The paper also cites research from the URL shortening service Bit.ly finding that half-life varies by subject matter. Business articles have on average a longer half-life; politics, celebrity and entertainment pieces are intermediate; sports articles are comparatively shorter-lived.

The same Bit.ly research found half-life affected by where a link was first posted, with links shared on Facebook longer-lived than those on Twitter.

Attribution matters here. The 80:10:10 finding belongs to the Castillo paper's own sample; the 36-hour mean and the topic and platform variations are findings by other researchers that the paper cites. They are consistent with one another, and none is a measurement of placement traffic specifically.

What makes the exceptions rebound?

Links, mostly, which is the practically useful part of the finding.

The paper identifies a class of roughly one article in ten whose traffic declines and then recovers, and attributes the recovery to internal or external links — another page pointing at the article and sending a fresh wave of readers.

This matters because it identifies the only mechanism in the research by which an article's life extends beyond its own news cycle. An article nobody links to follows the decay curve. An article that gets picked up, referenced, added to a resource page or cited by a later piece re-enters circulation.

It also connects to a finding reported elsewhere in this series. Pew's link-rot study found 5% of outbound links on news pages already broken and 23% of news pages carrying at least one broken link. The mechanism that extends an article's life is the same mechanism that decays over the following years, which means a placement's long tail depends on infrastructure that is itself eroding.

Timing matters as much as existence. A link arriving while an article still has momentum compounds with it; one arriving months later restarts from nothing. That is part of why coordinated coverage outperforms the same pieces published in isolation.

Why is this evidence thinner than it looks?

Because it is old, narrow, and not about placements.

The Castillo study was published in 2013, and the web it describes predates the collapse of referral traffic from social platforms, the rise of AI answer surfaces, and substantial changes in how people encounter news. The direction of its finding is unlikely to have reversed — if anything, attention has become more front-loaded — but the specific percentages should be treated as indicative rather than current.

The sample is 606 articles from a single publisher, Al Jazeera English, whose audience and subject mix are not representative of publishing generally. A study of a trade publication or a local newspaper might produce a different distribution.

Most importantly, none of this research measures sponsored placements. It measures news articles, which arrive through news cycles, social sharing and homepage promotion. A placement in a trade publication may behave differently, and nobody has published the comparison.

The absence is telling in itself. Publishers hold this data for every article they run, and buyers are rarely shown it. A market selling attention has produced almost no public evidence about how quickly that attention disappears.

Treat the shape as the finding and the percentages as illustration. That attention front-loads and decays quickly is robust across every study cited here; that it is precisely eighty per cent in twelve hours is a fact about 606 articles at one publisher in 2013.

What should a buyer actually expect?

That the great majority of whatever traffic a placement produces will arrive within the first day or two, and that the first week is essentially the whole of it for direct readership.

This reframes what a placement is for. If the value proposition is people reading the article, then the value is realised almost immediately and is finished quickly, and the price should be judged against a day or two of attention rather than against a permanent asset.

The durable value, where it exists, lies elsewhere: the page existing as a citable record, its presence in search results for specific queries, the credibility of a masthead in a media strip, and whatever links it attracts. Those persist on different timescales, and the last of them is what the rebound finding is about.

Judged that way, the common disappointment makes sense. A buyer expecting sustained referral traffic from a placement is expecting the behaviour of the 10% rather than the 80%, and the difference between those outcomes is largely whether anyone else decided to link to it.

It also explains why placement reporting so often arrives quickly and then stops. A report compiled a fortnight after publication is describing a finished event, and there is rarely more to add because the curve has already flattened.

What this means for anyone buying placement

Ask the publisher for the traffic curve on comparable articles. They have it, it costs them nothing to share, and a publisher unwilling to describe how their pages perform after week one is telling you something.

Do not price a placement as an annuity. If the case for buying rests on continuing readership, the research says that case is weak, and the honest justifications are the durable ones — record, searchability, audience reached in the first days.

Plan for the rebound rather than hoping for it. The one documented route to extended life is being linked to, which means the work after publication — getting the piece referenced, cited, included in resource pages — does more for its longevity than anything in the commissioning.

Media Morpheus sells placement. What a placement reliably delivers is a page, an audience reached largely in its first days, and a citable record that persists as long as the publisher keeps it up. Sustained traffic is not part of the product, for us or for anyone, and a seller promising it is promising the exception as though it were the rule.

Methodology and limitations

The primary source for this report is Castillo, El-Haddad, Stempeck and Pfeffer, Characterizing the Life Cycle of Online News Stories Using Social Media Reactions, arXiv:1304.3010v3, October 2013, based on 606 articles from Al Jazeera English with visit data and Twitter and Facebook reactions. Figures on article half-life are attributed within that paper to other researchers: the power-law distribution with a 36-hour mean to Dezso and colleagues' analysis of a Hungarian news portal, and the variation by topic and posting platform to research by Bit.ly. Six limits apply. First, the research dates from 2013 and predates major changes in how readers reach articles; percentages should be read as indicative rather than current. Second, the primary sample is a single publisher whose audience and subject mix are not representative. Third, none of this research measures sponsored or paid placements, only news articles, and no published comparison was located. Fourth, the half-life figures are cited secondhand within the paper and were not independently verified against their original publications. Fifth, a power-law distribution means the reported mean describes few individual articles. Sixth, no original measurement was conducted for this report.

Conclusion

Attention to an article collapses fast. In the best study available, traffic to roughly eight articles in ten fell continuously through the first twelve hours, with only about one in ten recovering afterwards — and that recovery came from other pages linking in, not from anything intrinsic to the article. Separate work put the median article's half-life at well under two days. The evidence is thirteen years old, drawn from one publisher, and measures news rather than placements, which is itself worth noticing: an industry that sells attention has published almost nothing about how quickly it disappears. For a buyer the practical implication is clear enough. A placement's readership is a two-day event, and everything durable about it — the record, the search presence, the links — is a different asset with a different justification.

Frequently Asked Questions

Sources

1. peer-reviewed Castillo, C., El-Haddad, M., Stempeck, M. and Pfeffer, J., 2013. Characterizing the Life Cycle of Online News Stories Using Social Media Reactions, arXiv:1304.3010v3, based on 606 Al Jazeera English articles.
2. peer-reviewed Dezso et al., analysis of visits to a large Hungarian news portal reporting power-law distributed article half-life with a mean of 36 hours, as cited in Castillo et al.
3. vendor research Bit.ly research on article half-life by topic and posting platform, as cited in Castillo et al.