Why the next phase of reader engagement is about relevance, timing, and trust
Attention math has changed
Let's start with a simple equation: more content divided by the same amount of reading time. A 2024 study in Quantitative Science Studies found that the number of articles indexed in Scopus and Web of Science in 2022 was about 47% higher than in 2016, far outpacing growth in the number of working scientists. Meanwhile, long-running surveys by Carol Tenopir and Donald King suggest that researchers' reading has reached its natural limit. In 2012, US scientists reported reading about 22 articles a month, down from 27 in 2005, and the time they devoted to each article had leveled off at just over half an hour.
Taken together, these findings point to a sobering reality for publishers and societies. The volume of research competing for a reader's attention keeps climbing, while the hours a reader can devote to it have stopped growing, meaning every email you send draws on a reading budget that is already spent. Email personalization in scholarly publishing, when done thoughtfully, respects that constraint and helps each reader spend their limited time on the research most relevant to their work.
The three levels of email personalization
The word personalization is used to describe everything from a first name in a subject line to machine-learning recommendations, which makes it hard to discuss productively. It is more useful to think of it as a progression through three distinct levels, each drawing on richer data and delivering proportionally more value to the reader.
Level one is cosmetic personalization, which inserts details such as the reader's name, the journal title, or their membership tier into an otherwise identical message. It is easy to implement, but readers have largely learned to look past it, because a personalized greeting placed on top of an irrelevant table of contents remains an irrelevant table of contents.
Level two is segmentation, which divides your audience into groups based on information readers have provided about themselves, typically through registration forms, membership applications, or subscription preferences. Common segments include specialty, professional role, career stage, institution type, and the subject areas a reader selected when signing up. A neurology society that sends stroke research to stroke specialists and epilepsy research to epilepsy specialists practices level two personalization, as does a publisher that sends early-career researchers content on grant writing while sending department chairs content on research strategy. Segmentation aligns content with what readers have told you about their professional identity, but it depends on two things many organizations struggle with: complete, current profile data, and the capacity to build and maintain several versions of each email rather than one. (A simple place to start: stop treating members and non-members the same way in email.)
Level three is behavioral personalization, which relies on what readers actually do rather than what they once said about themselves, including the articles they click, whether they go on to read the full text, what they browse afterward, and the times of day they tend to engage. A reader who registered as a general cardiologist but has read eight papers on cardiac imaging over the past two months is revealing an interest their profile will never capture. This is where the most meaningful gains are found, and where reader trust determines whether personalization succeeds or backfires.
Trust is the design constraint
Scholarly readers differ from a typical marketing audience in an important way. Elsevier's Insights 2024 survey of roughly 3,000 researchers and clinicians found that 94% of researchers believe AI will accelerate knowledge discovery, yet 95% also believe it will be used to spread misinformation. In other words, this is an audience that is genuinely receptive to smarter tools while remaining deeply skeptical of any system that feels opaque, manipulative, or careless with the integrity of the research record.
Preserving that trust comes down to four principles. First, personalization should never alter the research itself; it should influence which content a reader sees and when, never what that content says. The second is that recommendations should explain themselves, since a simple note such as "Recommended because you read several recent papers on cardiac imaging" transforms what might feel like an algorithmic black box into a service the reader can understand and evaluate. The third is that inferences should stay within reasonable bounds, so reading history suggests related articles but never implies anything about a clinician's patients or a researcher's unpublished work. The fourth is that readers should always have control, because a clear, easy way to adjust topics or frequency signals that the relationship runs both ways.
Let editors lead and let data assist
A common concern about personalization is that an algorithm will bury important work simply because a reader has not shown prior interest in the topic. The most effective response is to design emails with two lanes, each governed by a different kind of judgment.
The first lane is editorial, reserved for content that must reach everyone it applies to regardless of their click history, such as a major clinical guideline update, a retraction notice, or a landmark trial. Editors decide what belongs there, and no behavioral data overrides them. The second lane is driven by relevance and appears below the editorial selections, where each reader sees three or four articles chosen based on their own reading behavior and stated interests. Editors retain their authority, and readers receive content that feels genuinely selected for them. (Keeping the editorial lane accurate is also where approval workflows for scholarly publishing emails earn their place.)
Timing is personalization too: send time optimization done right
Relevance depends not only on what you send but on when it arrives. Consider a clinician who reliably reads email at 5 a.m. before rounds and again in the mid-afternoon. An email that arrives at 10 a.m. will likely be buried beneath the messages that accumulate before that clinician returns to the inbox.
Individual send time optimization addresses this by analyzing when each reader has historically engaged with past emails and scheduling future delivery shortly before that window. The critical detail is that this analysis should be based on clicks rather than opens. Privacy features in several widely used email apps now preload message content automatically, registering an open whether or not a person ever looks at the email, so open data often reflects when a device downloaded a message rather than when a human being read it, which makes open rates and open times an unreliable guide to actual reader behavior. A click requires a deliberate action from a real person at a specific moment, making it a far more trustworthy signal of attention. For readers who do not yet have enough history to establish a pattern, a sensible default based on similar readers can serve until their own habits become clear.
Why the effort pays off
The business case is well documented. McKinsey estimates that personalization typically lifts revenue by 5% to 15% and makes marketing spend 10% to 30% more efficient. For a publisher or society, that value appears in renewals, submissions, usage figures reported to institutional subscribers, and the quality of audiences offered to sponsors.
The engagement case is just as compelling, and arguably more important for organizations whose mission is disseminating research. McKinsey found that 76% of consumers say personalized communications are a key factor in prompting them to consider a brand, 78% say such communications make them more likely to return, and 78% say they make them more likely to recommend that brand to others. Researchers and clinicians aren't shoppers, but the underlying behavior translates directly to scholarly publishing: a reader who consistently receives relevant content is more likely to open the next issue, return to the journal's site, renew a membership, and forward an article to a colleague. That last behavior matters in a field where peer recommendation remains one of the most trusted ways research travels.
A practical place to start
No organization needs to reach level three overnight, and a phased approach over the first 90–180 days is realistic. Spend the first few months auditing the data you already have and understanding where it lives, including registration fields, email click data, and website reading behavior. Then focus on tagging content by topic and building a two-lane email template that separates the editorial section from the relevance section. In the third month, you can begin testing send timing with your most engaged readers, measuring success by clicks that lead to full article reads rather than by opens.
In a world with more research than anyone can possibly read, the publishers and societies that earn their readers' attention will be the ones that consistently prove useful, and personalization done with care is one of the most reliable ways to get there. If you want to see how a purpose-built email platform for STM and scholarly publishers handles the two-lane template, click-based send timing, and reader controls, talk to us.
Continue the conversation
This article builds on a conversation from our short video series with Samast AI, From Content to Community, in which experts in scholarly publishing, data, and responsible AI explore how publishers and societies can use audience insights, personalization, and emerging technology to build stronger relationships with their readers. Each episode runs about two minutes and takes on a single question, from what meaningful personalization looks like in practice to which reader signals matter most.
Watch the full series on YouTube: From Content to Community — STM Publishing playlist. Next in this series on the blog: Before the Algorithm: Why Connected Reader Data Is the Real Foundation of Personalization.
FAQ: Personalization in Scholarly Publishing
What is email personalization in scholarly publishing?
Using what a publisher knows about a reader — their stated interests, role and specialty, and the articles they actually click and read — to decide which content each reader receives and when, without changing the content itself.
What is send time optimization?
Scheduling each reader's email for shortly before the time they have historically engaged, based on their own past behavior. For scholarly audiences it should be driven by clicks, since automatic image preloading makes open times unreliable.
Does personalization mean letting an algorithm decide what gets published?
No. A two-lane design keeps an editorial lane that editors control for content everyone should see, with a relevance lane beneath it tuned to each reader.
Where should a publisher or society start?
Audit existing data (registration fields, click data, reading behavior), tag content by topic, build the two-lane template, then test send timing with the most engaged readers over the first 90–180 days.
Sources
- Hanson, M.A., Gómez Barreiro, P., Crosetto, P., and Brockington, D. (2024). The strain on scientific publishing. Quantitative Science Studies.
- Scientific American. Scientists Reading Fewer Papers for First Time in 35 Years (reporting on Tenopir and King reading surveys).
- Elsevier (2024). Insights 2024: Attitudes toward AI.
- McKinsey & Company. What is personalization? McKinsey Explainers.
- McKinsey & Company (2021). The value of getting personalization right, or wrong, is multiplying.
