4.1 Why bibliometrics?

Bibliometrics, or the scientific study of scientific publications, is becoming an increasingly important part of scientific research and science policy. In biomedicine, as in most scientific disciplines, an increasing number of researchers are competing for a flat or decreasing amount of research funding (Alberts, Kirschner, Tilghman, & Varmus, 2014), contributing to an overall aging of the biomedical workforce (Daniels, 2015) and leading to an increased emphasis on research evaluation as a tool for deciding who does, and does not, receive funding. This emphasis is further complicated by the increasing pace of scientific publication—PubMed indexed over 1.25 million scientific articles in 2016—and the increasing specialization of scientific research. All three trends place enormous pressure on the peer review process, in which researchers and experts read and qualitatively evaluate each other’s work. Although peer review remains the gold standard for research evaluation, it is not perfect. Aside from the fact that peer review is difficult, expensive, and time-consuming to perform, different peer review panels come to different conclusions when reviewing the same work (Bornmann, 2011) and peer reviewers are subject to conscious and unconscious biases in reviewing the work of others (Lee, Sugimoto, Zhang, & Cronin, 2013).

These issues with peer review have led many scientific evaluators to turn to bibliometrics as an alternative for evaluating scientific research. Bibliometrics are now regularly used in national research evaluation exercises in countries around the world, including the United Kingdom (Higher Education Funding Council for England, 2014), Italy (Ancaiani et al., 2015), Australia and New Zealand (Nature Outlook: Assessing Science, 2014), and China (Shao & Shen, 2011) and are incorporated in many high profile university rankings such as the Times Higher Education ranking and the Shanghai Ranking. Bibliometrics are also routinely used at the individual level to evaluate researchers for promotion and tenure and, in some cases, to make funding decisions. Despite this wide usage, many bibliometrics and nonbibliometrics researchers have criticized how bibliometrics are being used in these evaluations. Debates over the Italian exercises (e.g., Abramo & D’Angelo, 2015,2017; Benedetto, Checchi, Graziosi, & Malgarini, 2017; Franceschini & Maisano, 2017) are indicative of the main criticism: that bibliometrics are being used inappropriately by evaluators who lack expertise in bibliometric theory and practice. Evaluators may be drawing the wrong conclusions from bibliometric indicators because they do not know enough about those indicators to select the appropriate metrics or to interpret them properly. They are also influencing researchers’ behavior as a result of these evaluations, often in ways that they do not intend (de Rijcke, Wouters, Rushforth, Franssen, & Hammarfelt, 2016). Bibliometrics researchers have even published a manifesto setting out 10 recommended principles for the use of bibliometrics in evaluation (Hicks, Wouters, Waltman, de Rijcke, & Rafols, 2015).

All of this presents informationists with a service opportunity. There is a large and growing need for selecting, generating, and interpreting bibliometric indicators at research institutions around the world, but there is also a lack of knowledgeable practitioners available to fill that need. In this chapter, I will argue that informationists, and librarians more generally, are perfectly positioned to respond to that information need. I will first provide a brief introduction to the science of bibliometrics and then make the case that informationists are uniquely suited and qualified to provide bibliometric services to their research institutions. I will then provide a case study of the bibliometric services program at the National Institutes of Health (NIH) Library and finally give some practical advice to informationists on how to build the knowledge and skills necessary to provide bibliometric services to their customers.

4.2 What is bibliometrics?

Bibliometrics is the quantitative study of scientific research using publications as a proxy for research. The idea is that we can use scientific publications as a primary data source to learn about how scientific research works. Bibliometrics extracts various kinds of data from publications and uses quantitative techniques to analyze that data in various ways to try and learn something about scientific research. Although best known for, and in some circles synonymous with, its attempts to measure the impact of individual papers, authors, and journals, bibliometrics is really a science of science, interested in quantifying and identifying trends in science itself. For a more expansive overview of the field, see Mingers and Leydesdorff (2015).

As bibliometrics is such a large and heterogeneous discipline, it can be helpful to think bibliometrics as being concerned with four major themes: productivity, collaboration, research topics, and citation impact. Bibliometrics is not only interested in each one of these themes by themselves, but also in combination with each other, so that there are analyses on the effects of collaboration on citation impact, or on measuring productivity per research topic, and so on. Analyses in and across themes can also be applied to publication sets at various levels of aggregation—whether by a particular author, or produced by a particular university, or on a particular topic—to inform evaluations (Haustein & Larivière, 2015) or set agendas for future research. In the rest of this section, I will discuss each of these themes in more detail and provide examples of the kinds of analyses that can be performed in each theme.

Productivity analyses typically focus on publication counts, or counting how many publications were produced by a particular entity in a particular period of time. Besides being a measure of productivity, publication counts can also be used to describe a particular publication set by indicating the number of publications per document type (original article, review, editorial, etc.), per journal, per database-defined subject category, and so on. Publication counts per year can also be used to measure research intensity over time, either for an entire data set or, when combined with topic analysis, within a data set to indicate changes in research focus over time.

Collaboration analyses attempt to identify and describe the social network of science. Using network science techniques (Borner, Sanyal, & Vespignani, 2007; Newman, 2003), they seek to understand which authors produce scientific research and how they do, or do not, work together to produce that research. Most collaboration analyses in bibliometrics use coauthorship on scientific publications as a proxy for collaboration, creating networks of authors who are more or less strongly connected to each other based on the number of publications they have written together (e.g., Newman, 2001, 2004). When viewed from above, as it were, these networks often reveal patterns in how authors do, and do not, collaborate with one another. When combined with network-based metrics, these analyses can also identify “bridge” or “gatekeeper” authors who tend to work across disciplines or existing author communities and who connect authors in different areas. These collaboration analyses are usually performed at the individual author level, but they can also be performed at the institution or country level, based on the stated affiliations of the authors in a given publication set, to identify patterns in how institutions or countries collaborate to produce scientific research publications.

Topic analyses seek to understand the intellectual structure of scientific research and how that structure changes over time. Given a set of thousands of publications, topic analysis seeks to understand what those publications are about, how those topics interrelate, which topics have the most emphasis through either publication or citation intensity, and which topics might be gaining or losing emphasis over time (Borner, Chen, & Boyack, 2003). Topic analyses in bibliometrics tend to fall into two general categories: citation-based approaches and text mining approaches. Citation approaches use cited references as a mark of similarity between articles, with the idea that articles cite, and are cited by, other articles that are topically similar to themselves (Borner, 2010). Text mining approaches, on the other hand, analyze the words used in the titles, abstracts, or sometimes even full texts, of articles and assign articles to topics based on how frequently certain terms occur in them (van Eck, Waltman, Noyons, & Buter, 2010; Yau, Porter, Newman, & Suominen, 2014). Either approach can be applied to publication sets by specific authors or groups of authors to assess how closely their research aligns with strategic goals, or to sets in particular disciplines or on certain topics to identify research gaps and emerging areas of interest.

Finally, citation and impact analyses seek to understand which articles, or sets of articles, are most influential in advancing scientific research. It does this by counting citations to specific articles. The idea is that if one paper cites another, that citation serves as an indication that the article being cited was in some way influential on, or at least useful to, the article citing it. Zooming out, then, we can get a sense of the research impact of any given article by counting the number of other articles that cite it. Zooming out even further, we can measure the research impact of a set of articles—by a particular author or research institution, published in a particular journal, or authored by a particular country—by measuring the number of times the articles in that set have been cited. More extensive overviews of citation counts and citation indicators can be found in Belter (2015) and Waltman (2016).

Although simple in theory, using citation counts to measure research impact is, in practice, far more complicated than it sounds. Citation counts are influenced by a variety of factors—age, discipline, document type, database, etc.—that are independent of the usefulness or impact of specific articles. This means that an article with 20 citations is not necessarily more highly cited than an article with 10 citations, or an author with an H-Index of 30 is not necessarily more highly cited than an author with an H-Index of 20. We would need more context about these articles and authors to be able to interpret these metrics correctly (Ioannidis, Boyack, & Wouters, 2016). In addition, citation counts, and all indicators based on citation counts, do not take authors’ motivations for citing other articles into account. We know that authors cite other works for a wide variety of reasons (Bornmann & Daniel, 2008), but these reasons cannot currently be taken into account in citation indicators, meaning that we do not know why a particular article or author has been highly cited. Finally, citation counts only measure a certain kind of impact: impact on future scientific research. They do not measure more abstract concepts such as quality, nor do they measure the broader impacts of scientific research, such as changes in policy or improvements in people’s health. They only measure how useful certain articles have been to the authors of other articles for writing their own articles.

More recently, a new category of metrics has emerged, often referred to as alternative metrics or altmetrics. Altmetrics are measures of online interaction with publications, such as downloading them, mentioning them on Twitter, saving them to online reference managers such as Mendeley, and so on. The theory is that as we can count the number of times a video has been viewed on YouTube, or a tweet has been retweeted, in the same way we can measure the same kinds of interactions with scientific publications and use them as alternatives to citation metrics for measuring research impact. The reality of altmetrics is much more complicated. We know, for example, that most altmetric indicators are not correlated with citation indicators (Bornmann, 2015; Erdt, Nagarajan, Sin, & Theng, 2016), suggesting that they measure something different than citations do, but what, if anything, they do measure is still not really known. Altmetrics are an active area of research in bibliometrics at the moment (Haustein, 2016), so our understanding of these metrics and their uses are changing rapidly, and will likely continue to change in the years ahead.

Finally, it is worth emphasizing that although bibliometrics and altmetrics are useful tools for exploring and evaluating scientific research, they are still proxy measures for the thing itself. Bibliometrics only measures research publications, not research activities. In addition to doing research, researchers also perform a range of other important activities—teaching and mentoring, reviewing other researchers’ work, consulting on legal and regulatory affairs, communicating with the public, etc.—that are critical to advancing and applying scientific research, but do not necessarily result in publications. As a result, none of these activities are captured in bibliometric analyses. As a result of these, and other limitations already mentioned, bibliometrics can inform evaluations and decision-making, but cannot replace expert judgement in assessing and evaluating scientific research.

4.3 Why informationists?

So what does all of this have to do with informationists and librarians? Even if bibliometrics is such a big deal, why should informationists be involved? The first and most obvious answer is that our patrons are increasingly asking us to be. They are asking us for help in gathering publications for analysis, and increasingly for the analyses themselves. In doing so, they are displaying an intuitive understanding that because bibliometrics is about publications, and because they come to the library for publications, then maybe they can come to the library for bibliometrics, too. They may well understand that bibliometrics and scientometrics were largely started by librarians—Eugene Garfield, the creator of the Science Citation Index, now the Web of Science, was a librarian by training—and are an outgrowth of library and information science. However, whatever their reasons for asking, they are not only asking us, but more of them are asking us all the time.

Second, informationists already possess many of the skills necessary to do bibliometric analyses. We are experts in the knowledge ecosystem, which is to say that we understand how knowledge is produced, recorded, and disseminated in the scholarly literature. Our understanding of the information needs and habits of researchers in different disciplines helps us understand the differences in scholarly communication practices in those different disciplines, which helps us design analyses that are appropriate in those different disciplines. Through our work performing literature searches, we also have advanced skills in database searching and publication management that are essential to creating accurate publication sets on which an analysis can be based. Moreover, we are service professionals. We not only understand how to provide good customer service, but also understand how to conduct reference interviews to help our patrons clarify and articulate their actual, as opposed to their stated, information needs. These skills are essential in bibliometrics, because customers asking for bibliometric analyses rarely understand bibliometrics well enough to know what to ask for. Informationists already know how to conduct a reference interview to identify what those customers actually need, and are willing and able to adapt their response to address that specific need.

Third, bibliometrics provide a new way of producing the kinds of products that librarians have been producing for generations. One of the more valuable services that early librarians provided to their customers were annotated bibliographies and topic summaries in which they attempted not only to gather information on a particular topic, but also synthesize that information into a form that more directly answered a customer’s information need. Not content with simply providing a list of references, they attempted to answer a particular question with those references. The increasing volume and technicality of the biomedical research produced today make those syntheses extremely difficult for any but the most narrowly defined topic. Bibliometrics offers methods of creating those kinds of syntheses, but using automated means rather than manual ones. They allow us to create topic guides on topics as broad as Alzheimer disease or physical rehabilitation for cancer patients that would be nearly impossible to perform manually. At the end of the day, no patron asks for a literature search because they want a list of publications. They ask because they want to learn something about the research on a particular topic. Bibliometrics allows us to go a step beyond the traditional bibliography toward addressing that underlying information need.

The final, and most important, reason that informationists should be involved in bibliometrics is trust. We know from a long line of research that patrons trust librarians to provide accurate and unbiased information, and trust is essential when it comes to evaluation. Informationists stand outside the traditional evaluation space—we are neither the evaluator nor the evaluated—so we do not have a stake in the outcome of the evaluation. As a neutral third party, we have no reason to be biased or inflect the results of our analyses in any direction, which allows both sides to have more confidence in the results we provide than the results provided by either of them. Our neutrality, combined with the trust that our patrons already have in the information we provide, make us uniquely suited to inform the evaluation process.

Our involvement in that process can take a variety of forms, from curating and ensuring the accuracy of the publications included in the evaluation all the way to performing the analyses ourselves and ensuring our results are interpreted properly. As we have already seen, there is certainly a need for this information, and informationists are perfectly placed to provide it.

4.4 Case study: Bibliometrics at the NIH Library

Like many libraries, the NIH Library had been getting ad hoc requests for individual-level bibliometric analyses for years. These requests would be handled by staff librarians and informationists as part of their daily activities, but none of these staff members had received any formal training in bibliometrics and could only provide limited services in this area. Recognizing the growing demand for bibliometric services, and building on the successful launches of its bioinformatics and data services programs in prior years, the NIH Library have decided in late 2013 to build a bibliometrics service program. It had two generalist informationist positions open, so it decided to repurpose those positions to be bibliometric specialists who would create the new program. These positions, similar to NIH Library positions in bioinformatics and data, would turn the informationist model on its head by offering a specialized service, in this case bibliometrics, to all of NIH rather than offering a range of services tailored to a specific group. My colleague Ya-Ling Lu and I were hired for these bibliometrics positions in May of 2014.

The library’s bibliometric services program was launched the following month. Our formal charge is to be “a key source of bibliometric information at NIH” and to “promote increased understanding of bibliometrics for the NIH and the NIH Library staff, assist NIH staff with bibliometric analysis and the responsible use of bibliometrics, and further the adoption of bibliometric services by the broader library community.” Our goal was not to replace the ad hoc bibliometric services already being provided by library staff, but rather to formalize and enhance those services by bringing additional knowledge and skills that would ensure these services were performed according to the current bibliometric literature and enable the library to perform more advanced analyses that could assist with a broader range of requests than was possible before we arrived. We also wanted to train our fellow librarians and informationists at the NIH Library in bibliometric theory and practice to enable them to perform more advanced analyses if they wished.

At the NIH Library, we provide four broad categories of services to NIH staff: consultations, training, core analyses, and custom analyses. One of the unique aspects of working at the NIH is that we are both a producer of research, through our intramural research program, and a funder of research, through our extramural grant programs. This means our customer base includes both intramural researchers and research directors as well as extramural grant managers and evaluators, since each of the institutes and centers that comprise the NIH has staff dedicated to managing and evaluating their grant funding portfolios. These different customer groups have different motivations and needs, which have informed the categories of services that we provide.

When we came to the NIH, we quickly learned that although our customers generally knew what bibliometrics were, they did not understand it well enough to know what bibliometrics could do, or the range of questions it could be used to address. We therefore developed our consultation and training services to meet these needs. Through our consultation services, we answer questions about bibliometric indicators and the appropriate use of bibliometrics in research evaluation, provide advice and guidance on bibliometric data sources and tools, and perform consultations on selecting appropriate analyses and designing methodologies to answer specific kinds of questions. Through our training program we provide instruction to NIH staff about bibliometric theory and practice, with courses ranging from introductory overviews of what bibliometrics are, and are not, capable of doing more hands-on courses on measuring an individual’s citation impact or performing topic analyses in the R programming language. We offer these courses both through the library’s standard instruction program and on demand from our customers. NIH’s extramural grant managers and program officers are especially interested in these kinds of customized trainings.

Although we provide a steady stream of consultations and trainings, the majority of our time is spent on our core and custom analyses. Our core analyses are primarily oriented toward understanding NIH’s intramural research program, both through a series of profile reports for each of the institutes and centers of NIH and, more recently, though overarching analyses of NIH intramural research across these institutes and centers. One of the first projects we took over soon after starting the program was to produce a series of institutional research profiles of the intramural research being performed at each of the 27 institutes and centers of NIH. These profiles aim to provide a high-level snapshot of the productivity, collaboration, research topics, and citation impact of each institute and center’s research publications.

We decided to produce these profiles on our own, rather than by request from our customers, for a number of reasons. They not only provided the library with valuable information about our primary customer base, but also gave us in the bibliometrics program a series of tangible products that we could use to tell customers about our program and the kinds of things that we could do for them. We sent each profile report to the scientific director of the institute or center analyzed in that report with a short email introducing our new program and offering to provide additional information about the report or our services. Some directors never responded, others welcomed the reports and asked a few questions, and few others invited us to present the reports to their entire research staff. Many also passed our reports on to colleagues in both the intramural and extramural programs, who then requested analyses of their own programs and passed the results on to their colleagues.

Most of our first year at the NIH Library was spent working on these core analyses, but as word of our services began to spread at the NIH, our focus shifted to our custom analyses. Our intent with these custom analyses was that as a customer could come to the NIH Library to request a literature search, in the same way they could now come to the library to request a bibliometric analysis tailored to their specific needs.

When we receive a request for a custom analysis, we always schedule a meeting with the customer before we begin the analysis to clarify the customer’s needs, their goals for the analysis, and the motivation for their request. This is essentially a modified form of the traditional reference interview, in which we attempt to identify the customer’s actual, as opposed to their stated, information needs. We have found that this is especially important in providing bibliometric services because, as stated above, customers who want a bibliometric analysis often do not understand it well enough to ask for what they really want, so they ask for what they do know, or think we can provide. In some cases they ask for analyses that we know are not appropriate for their intended goals, so we explain to them in this meeting why those analyses are not appropriate for their specific needs and offer to perform other analyses that are appropriate for them. We have found that as long as we take the time to explain the problem and offer a better alternative, customers are happy to accept our suggested changes. In all the cases, we have found it valuable to have this initial consultation to more clearly identify the customer’s needs, the scope of the project, the data sources we are going to use, and the timeframe for the project, because we frequently come away from these meetings with a different understanding of the project than we had going in. It is also valuable to our customers to ensure that the analysis that we spend a substantial portion of our time producing is actually useful to them.

The nature of these custom analyses varies widely in scope and methodology, but some analysis themes have begun to emerge. Our intramural researchers tend to be interested in variations of our institute and center profile analyses, but run for their specific laboratory or department rather than their entire institute. In some cases they want us to do some form of benchmarking where we compare their publication output to other laboratories or departments that they identify as their peers. Our extramural grant managers tend to want similar kinds of profile analyses, but for the specific sets of grants that they manage. Many also ask for analyses of their institute- or center-funded publications on a specific topic like obesity or blood diseases. Some also ask for what we refer to as landscape analyses, or analyses of all published research on a particular topic like Alzheimer disease or cancer rehabilitation, to identify trends or gaps in the research literature that could be used to inform future funding priorities.

When we started the program, we knew that all of these services, and especially our custom services, would each take substantial amounts of our time to complete, so we decided it was necessary to establish policies to protect that time. Since there are only two of us to provide bibliometric services to all of the NIH, we needed policies to ensure that our time was equitably distributed across the NIH institutes and centers. One such policy is that we require at least 2 weeks to complete any analysis, unless the customer is willing to provide financial reimbursement for our time. If a customer asks for an analysis to be completed in < 2 weeks, we inform the customer that they can either pay us for the time we devote solely to their project, and not to others, to meet their deadline, or else we will not provide the service in the allotted time. The other policy is that we will spend up to 40 h of our time on any single project as part of the library’s base services, but no more. If after our initial consultation we estimate that a project will take longer than this to complete, we will inform the customer that we could complete the project, but they will have to reimburse the library for the extra time above the base 40 h necessary to complete it. Otherwise we will not complete the project.

Again, the goal of both policies is to ensure that we have enough time available to help everyone at the NIH and that we do not end up spending all of our time working on one analysis or for one institute or center. With bibliometrics, a single project could end up taking months to complete, so it was important for us to have a mechanism to prevent that, if necessary. Customers do still have the option to have us devote that amount of time to a single project, or for us to complete a project in a short period of time, but our use of financial reimbursement ensures that we spend that time for customers who are willing to support that time. In practice, < 1% of our projects are affected by either policy. Nearly all of our customers are fine with a 2-week turnaround time and we deliberately set the 40-h limit to be high enough that we could complete the vast majority of projects in that time. I will admit that we have not been as strict in enforcing that 40-h time limit as we probably should be, but I am glad it is there. Both policies give us mechanisms to decline or deflect certain projects with unreasonable demands or deadlines, while still allowing us to serve the vast majority of our customers.

In the 3 years since we launched the program, we have completed over 450 consultations, custom trainings, and bibliometric analyses for NIH staff. In June 2017, a fairly typical month for us, we taught 3 tutorials and courses, provided 8 bibliometric consultations, and completed 6 bibliometric analyses. However, beyond the raw numbers, we have been involved in a number of high-profile projects at the NIH. Our analyses are frequently used to inform intramural research reviews at the laboratory, department, and institute levels. We have partnered with intramural researchers on original research projects that they would not have taken on without us. Our analyses have informed evaluations and comparative effectiveness reviews of various NIH funding mechanisms, grant portfolios, and NIH-funded publications on various topics. In addition, we have informed the strategic planning processes for NIH institutes and centers to identify priority research areas for future funding. Although the NIH is a unique organization in many ways, the success of our program, and of similar programs at libraries across the country, speaks not only to the demand for bibliometric services at research institutions more generally, but also to the opportunities available for informationists willing to undertake them.

4.5 Advice for building a program

In talking to other librarians and informationists about my work, one of the most common things they tell me is that they would like to provide bibliometric services, but do not know where to begin. With so much to learn and so few introductory resources available, the entire field of bibliometrics can be intimidating. In addition, providing bibliometric services means becoming an analyst: a role we, as a profession, tend to avoid. Library schools tend not to offer or emphasize analytics courses, library students tend not to take them when they are offered, and practicing librarians tend to shy away from the analytical aspects of our profession. Transitioning into bibliometrics therefore requires us to not only learn the science of bibliometrics, but also how to be an analyst. It also requires us to change our perceptions about our own profession and about how we, as librarians and informationists, go about providing services to our customers.

These dual challenges—learning bibliometrics and taking on the role of an analyst—make getting started with bibliometrics a daunting prospect for most informationists. I know, because I was in the same position. I have not received any formal training in bibliometrics and have received very little training on data analysis techniques. Instead, I have made do with what I could learn on my own from what was available and through trial and error. Having gone through that process myself, I would like to offer some advice for others who would like to do the same.

First, and perhaps most important, learning bibliometrics takes time. It takes time to read books and articles on bibliometric theory, it takes time for that learning to become ingrained in your memory so that you can use it freely in discussions and consultations, it takes time to learn the tools and techniques of bibliometrics, and it takes time to perform any analysis, no matter how skilled and knowledgeable you become. All of this time is an important consideration when building bibliometric skills, because you need to ensure you have dedicated time for learning and doing it. Practically speaking, this usually means that you have to stop doing something else in order to have enough time to learn and do bibliometrics. For an individual informationist, this involves reevaluating how you spend your time, but for a library it often means pulling one or two staff members off other projects so that they can focus on bibliometrics. This time commitment is also ongoing: performing bibliometrics responsibly and effectively means keeping up with the bibliometric research literature and regular practice with the tools and techniques, so it requires a dedicated commitment of time to maintain knowledge and skills in this area.

Unfortunately, resources for learning bibliometric theory and practice are few and far between. Full-length books on the topic include those by Moed (2005), De Bellis (2009), Borner (2010, 2015), and those edited by Ding, Rousseau, and Wolfram (2014) and Cronin and Sugimoto (2014). The references cited in this chapter also represent a recommended starting point for exploring the original research literature on the topic. Asynchronous training in some aspects of bibliometrics is available through the annual Information Visualization MOOC offered through the University of Indiana (http://ivmooc.cns.iu.edu/). Hands-on training opportunities in bibliometrics are, at the moment, relatively scarce. The European Summer School for Scientometrics (http://www.scientometrics-school.eu/) offers an annual week-long course on bibliometrics started in 2010 to meet the growing demand for bibliometrics training in Europe. The Centre for Science and Technology Studies at Leiden University also offers theoretical and practical courses in bibliometrics (https://www.cwts.nl/training-education) both in Leiden and at other sites on demand. Ya-Ling and I at the NIH Library have also begun offering half- and full-day workshops on bibliometrics at the NIH campus, at the 2017 MLA and SLA annual meetings, and at other locations on demand.

Second, you do not need to be an expert in all aspects of bibliometrics to begin offering bibliometric services. Instead, we can take our cue from the start-up culture and begin by offering the minimum viable product. The idea is to begin with the most basic form of a service or product—one that meets only the most basic requirements but that also requires the least effort to produce—and then refine and build on that product based on customer feedback. This model works in the technology sector, but it also works in libraries, and especially with bibliometrics. My first bibliometric report consisted of two pages of black and white tables and a few lines of text. It was not much, but it was more than my customer had before, and it was sufficient for their needs. With so many bibliometric analysis options available, it is easy to become overwhelmed and think that you need to offer everything at once. Not so. Offering the minimum viable product first not only allows you to build basic proficiencies in bibliometrics, but the end product also generates feedback from your customers on what else they might want to have included in the future products, which allows you to target your future learning and skill development based on that feedback.

However, it is important to keep in mind that the minimum viable product cannot simply be the minimum product; it must also be a viable product. To produce a bibliometric report that is both informative and methodologically sound, you need to know what analyses to perform, which indicators to select, and how to perform those analyses accurately. That requires a fair amount of knowledge, not only about bibliometrics, but also about the context in which bibliometric analyses are produced and used. My advice would be to begin with one of the major themes of bibliometrics, such as citation impact, learn as much about that theme as you can, and then begin to create the minimum viable product for that particular theme. You do not have to be an expert in social network analysis to produce an accurate and useful citation impact report, so start with one aspect of bibliometrics and worry about the other aspects later.

Third, you do not have to wait for your customers to ask to start performing bibliometric analyses. We began our series of NIH profiles proactively, rather than waiting for customers to ask us. Offering services proactively, rather than in response to their questions, is admittedly a radical departure from the traditional library service model, but one that has substantial advantages. It is far easier to market a service like bibliometrics if you have something tangible to show. Rather than trying to explain what bibliometrics could do if only someone asked, it is far easier to show people an actual analysis performed on publications at or by your institution. Presenting your customers with an analysis also allows your customers to do your marketing for you. When presented with a bibliometric analysis, customers often pass that analysis on to their colleagues, who then pass it on to their colleagues. This kind of word-of-mouth marketing is far more effective than any messaging that you or your library could produce. Finally, it gives you the opportunity to practice, which is by far the best way to learn.

Fourth, start building on that minimum viable product as a way to expand your skills. After I produced that first two-page report, I challenged myself to add something to every subsequent report, whether it was a graph instead of a table, a new kind of analysis, or an improved visualization of a previous analysis. Building on previous analyses in this way allows you to consolidate the skills you have already gained while adding to those skills in a deliberate way. Customer feedback on previous reports is invaluable in this process, because it tells you what to learn next. Bibliometric analyses tend to raise new questions as it answers old ones, so your previous analyses will probably elicit additional questions from your customers. Answering those new questions then becomes the roadmap for your future learning. By constantly adding new pieces to your previous analyses in this way, over time you develop a wide range of analytical skills that are targeted toward answering the specific questions that customers at your institution have.

4.6 Prospects

Although many of the early pioneers of bibliometrics were librarians, the paths of librarianship and bibliometrics quickly diverged and have been relatively separate ever since. Recent changes in the ways that people find and access information and how scientific research is produced and funded are signaling a change to this trajectory. Calls for librarian involvement in bibliometrics have been made for at least a decade (Ball & Tunger, 2006; Joint, 2008; MacColl, 2010). Librarians and informationists, particularly in Europe (Astrom & Hansson, 2013; Gadd, 2011; Gumpenberger, Weiland, & Gorraiz, 2012) and Australia (Corrall, Kennan, & Waseem, 2013; Drummond & Wartho, 2009; Haddow, 2007), are increasingly answering those calls, providing information, guidance, and, increasingly, analyses to support national research evaluation exercises as well as researchers at their institutions. Momentum is growing in the United States as well, with university and government research libraries across the country adding capacity and starting programs to provide bibliometric services to their institutions.

The benefits of bibliometrics to libraries are clear. By providing services directly to administrators and policy makers, bibliometrics programs increase the visibility and prestige of the library to a key stakeholder group that is traditionally underserved by libraries. Adding capacity in bibliometrics also helps the library run more effectively. Bibliometrics can be used to inform collection development, provide insight into the institution’s research priorities, and identify targeted customer groups for library messaging. Most importantly, bibliometrics adds value to the services of the library. It allows the library to adapt to the changing information needs of its customers and provide new kinds of services that contribute directly to furthering the institution’s research mission.

Librarianship has always been a profession built on change. Librarians have always had to adapt to the changing needs of our customers, whether through the development of computerized information storage, the development of online databases, or the creation of the informationist service model. Bibliometrics is an opportunity to continue that change into the future, giving us new tools to provide more numerous, more efficient, and better services to our customers.