Looking back on six years of cultural heritage tweeting

Screen Shot 2020-02-11 at 11.00.36

When I left the Bodleian a few weeks ago, I said goodbye not only to Digital Bodleian (which I still care about more than nearly anything, and which will soon be relaunched very gloriously) and my habit of working from the Clarendon Building porch in between meetings, but to life as a very, very, very low-scale Twitter influencer. I had tweeted for Bodleian Digital Library Systems and Services for nearly six years: three years alone, and three with the help of my colleague Tim. In that time we grew our audience from just over 1000 to 20,000. It started as something I did while I was procrastinating about other things, but it evolved into a significant responsibility for both Tim and me, although possibly he and I were the only ones who saw it that way.

I started tweeting because I enjoyed it. I spent most of my time digitizing things–making images and metadata available online–and the work felt incomplete if I couldn’t share those things. I knew that someday people would come across them within the normal course of academic research–they would be browsing Pinakes or the ISTC and would click on a hyperlinked shelfmark–but I didn’t want to wait for that to happen. I wanted to connect people with the digitized books and manuscripts now, and to do it in vivid and memorable ways. During my time at the Bodleian, this grew from a personal compulsion to an institutional mandate. Over the last few years, libraries and museums have demonstrated that it’s possible to make collections accessible to non-academic users via social media; to make good on the promise of democratization through digitization. If it’s possible, funding bodies want you to do it.

After demonstrating that our work on Twitter accounted for 20% of the Polonsky Foundation Digitization Project site traffic (and later, 50% of the Manuscripts from German-Speaking Lands site traffic), I began to get practical questions about our social media strategy: how we crafted our tweets, how we measured up against peer institutions, what successful Twitter performance looked like. Essentially, I had to make answers up, because there is very little out there about what good cultural heritage Twitter looks like or what it’s meant to achieve. In fleshing out what might charitably be called a social media strategy, I drew heavily on blog posts by Adam Koszary (formerly of the Bodleian and the MERL) and Russell Dornan of the Wellcome, and I looked at tweets (mine and others’) that I thought did a good job and tweets that missed the mark somehow. In case anyone else is out there trying to answer the same questions for their director or collaborator or funding body, here is the essence of what I’ve learned.

The basic rules

These are the guidelines I would give to a brand-new BDLSS tweeter. Some of these are things I’ve learned over time, some of them are obvious to anyone who uses Twitter, and some of them might just be my own personal preference.

  1. Try to include an image in every tweet to catch the eye. (Usually easy, because we were basically always tweeting about digitized collections, and even if we weren’t tweeting about a specific item, we could usually find something tangentially related to the tweet. We never used stock images to liven up non-image-based content; to my mind this gives things a very corporate vibe.)
  2. Use alt text to describe images for people using screen-readers.
  3. Cite image sources. We included shelfmarks and folio labels in alt text and captions superimposed on the images themselves (although I am much lazier about this than Tim is), and always linked to the digitized image. I will acknowledge that including links and shelfmarks in the tweet text makes it look kind of ugly and dilutes the impact of a really funny tweet, and some people get around this by including the citation in a second tweet, but that doesn’t stop people from screencapping or quote-tweeting just the first tweet, and the whole purpose of what we’re doing is to drive people to the digitized collections, right?
  4. Use hashtags sparingly. More than two hashtags at the end of a tweet makes you look like a bot; more than one hashtag mid-sentence makes you look like an out-of-touch corporate account. (I am making these numbers up.) Also, most hashtags are not particularly useful. “We’ve just #digitized five new #medieval #manuscripts! #digitization #libraries” is hard to read and helps no one. However, we did use event hashtags and project hashtags (e.g., #PolonskyProject and #PolonskyGerman) because they provide some context and continuity without having to explain the project or link through to the website every time.
  5. Proofread very carefully. As a mere offshoot of the main Bodleian Libraries account, we got relatively little heat from people criticizing our usage, but it still really tanks things if you’re trying to tweet something a bit playful and you misspell “Bodleian” or use the wrong kind of “it’s”. I suspect that people are more tolerant of experimentation and irreverence from an institutional Twitter account if the tweets contain no errors. This is a bit tricky given that humor on Twitter often relies on unconventional punctuation or grammar, so you have to be careful about what lines you cross. I debated “yesterday they” vs. just “yesterday” in this tweet for a while, but I think it was funnier with the unnecessary “they”, and nobody seemed to mind. (God bless the MERL, which, when called out by pedants, always doubles down.)
  6. Use first-person plural in tweets unless you are answering a specific question, in which case use first-person singular. This makes it clearer that you are providing only a single person’s expertise rather than an institutional ruling.
  7. Never belittle your collections, your staff or your users; always work from a position of tremendous respect for all three. (This is especially important if you are working with sacred objects or objects from other cultures.) Early in my tweeting career, I got in a bit of hot water for the #damagedmanuscriptThursday hashtag. Honestly, I think this hashtag played a significant role in our early follower gains (people really, really love parchment repairs), but I should have framed the tweets better, with more of an emphasis on conservation and damage stabilization. This tweet and the linked blog post (which was reviewed before publication by colleagues in special collections and conservation) revisited the same topic with a more careful touch.
  8. People respond best to content that allows them to connect with the people who made or used the objects in our collections: marginal doodles, signs of use, details of how a manuscript was made or stored or photographed. Find these things and highlight them. If you’ve got a girdle book, for example, you might want to find a picture of someone carrying one to illustrate how it would actually have been used, and study the images for evidence of wear that shows it was used that way.
  9. Threads are time-consuming and lower your average per-tweet engagement rate (most tweets in a thread are unlikely to do as well as any standalone tweet), but people like them and engage with them, and they allow you to tell richer and more complex stories, so make time for them.
  10. You can and should plan some things in advance (for example, substantial threads that require research, or tweets tied to particular dates), but Twitter is an inherently informal and spontaneous medium, and you have to be able to react quickly to trending hashtags or other people’s tweets. (#WorldCatDay seems to happen every few weeks, and we didn’t always pick up on it, but whenever we did, we reaped the rewards.)
  11. If people reply to you, reply back.

What does a successful tweet look like?

I haven’t been able to find any cultural-heritage-specific benchmarks for social media engagement. However, I have read some reports on corporate and institutional social media performance indicating that a satisfactory engagement rate for a higher education institution’s tweets would be about 0.5%. I think a digital library account should expect to do better than that, for two reasons: a) there’s no reason why we can’t lard our posts with attractive images and b) we are a very specific arm of a respected institution with nothing to sell, meaning most people who see our tweets are already on our side. Our analytics at BDLSS certainly bore this out.

For the use of colleagues at other institutions who have been asked difficult questions about social media benchmarks, here are our numbers from October 2019:

  • Total number of tweets: 93
  • Mean number of impressions per tweet (i.e., how many people saw the tweet): 3600
  • Mean engagement rate (i.e., the rate of people who retweeted, liked, commented on, or clicked on images or links in the tweet): 2.4%

Our top tweet was this one, which earned about 20,000 impressions, with an engagement rate of about 4.2%:

Our lowest-performing tweet (excluding replies and threaded tweets) was this one, which earned about 3000 impressions, with an engagement rate of about 3.6%:

This engagement rate is higher than the mean because the mean includes replies as well as standalone tweets.

(By the way, a tweet that earns more impressions will often have a lower engagement rate, because it’s being seen by people who don’t follow your account and who are more likely to simply scroll past rather than interacting with it. That said, engagement is what we’re really looking for: we want people to view the images and click through to the digitized objects, and it’s not much use to us if lots of people view the tweet but no one actually clicks on it. So it’s worth looking at both metrics.)


Some of your tweets shouldn’t be successful

Twitter is important to BDLSS. It drives a significant proportion of our audience to our websites. That said, some of our tweets were duds, and I think that’s a good thing.

What BDLSS is trying to do on Twitter is engage people with digitized objects. Some digitized objects are extremely engaging: your book of hours, your Canterbury Tales, your exquisitely decorated Qu’ran. As you can see, it took no effort at all to make this Hebrew manuscript a Twitter hit:

But visually striking objects are only a tiny sliver of the Bodleian’s collections, and we wanted people to get to know the more boring objects too. Of course, we were very lucky, because nothing we’ve digitized is actually boring. After looking at a digitized object closely, I often ended up with a thread’s worth of fascinating things to tweet about, although the tepid response to some of these tweets served as an occasional reminder that what is interesting to me, a librarian with a quasi-parental bias toward everything in Digital Bodleian, might not be interesting to anyone else.

Some of the things our followers weren’t very interested in: archaeological tracings, maps of Oxford, photographs of the steel industry, Oxfam posters, sheet music.

Now, the MERL team have shown that it’s possible to get people excited about fat sheep, so why can’t we get people excited about Oxfam posters? Well, frankly, the MERL model isn’t for everyone. In order to make your cultural heritage content go viral without significant pushback, you have to really know your stuff; you have to be a small enough or new enough institution that your mentions won’t be full of people correcting your Latin; you have to demonstrate real curatorial knowledge as well as social media savvy; you have to let your audience in on the joke. (A big part of the MERL’s social media presence is a dry acknowledgement of the weirdness of using memes to get people interested in historical farming–a brilliant approach that tempers humor with genuine expertise, although (without sounding too smug) I don’t know that it would work for a collection with broader appeal. One of my favorite MERL tweets asks the reader to make the most of summer weather by “staying indoors, in the dark, in a museum, looking at horse brasses”. It’s a reworking by Adam of one of his earlier tweets for the Bodleian, which also performed extremely well but lacked the absurdist punch of “horse brasses”.) But most importantly, even the skilled GLAM-Twitter humorist needs to mix meat-and-potatoes tweets in with the funny stuff, so that their account doesn’t come across as wholly frivolous.

We certainly played with tone and humor at @BDLSS, especially when we knew the images wouldn’t speak for themselves. Case in point, this tweet by Tim, highlighting one of the 4000 pieces of Victorian sheet music digitized for the What’s the Score at the Bodleian? project:

We also sneaked duller objects into threads that started with more exciting ones:

But we never got all that silly. We never used a reaction GIF, and we very rarely piggybacked on trending memes. I experimented more with silly tweets than Tim did, I think, but neither of us wanted to stray too far from the persona we had created for ourselves: discursive, dry, highly enthusiastic, knowledgeable but not excessively so. (Given that we were tweeting about manuscripts that neither of us could read, “excessively knowledgeable” was not an option.) I suspect that our Twitter persona worked to the extent that it did because we never seemed to be trying too hard to garner likes; we appeared to be–and in fact generally were–simply sharing things we found interesting. It’s easy to say that I’m glad we never ended up in a Buzzfeed roundup of libraries who won Twitter this week, given that this was never within the realm of possibility; but still, I’m glad.

The truth is that a drawing of a proposed sewer system in central Oxford is not going to light many hearts afire. A few people might find it extremely interesting, but most people won’t, and if you tweet about it your engagement rate will suffer, and that’s OK. It’s still important to get the content out there, so that the library collection is fully represented, and so that people who do find that sort of thing interesting will be able to explore it. Like most ways of measuring impact, Twitter analytics runs the risk of distracting us from what really matters. We should be using social media to highlight underappreciated collections, and to experiment with new forms in order to reach underserved audiences; we should not be peddling #libraryporn and perpetuating the already alarmingly wide gap in public attention and institutional resourcing between, say, medieval Western manuscripts and 18th-century administrative archives. Finding new audiences is slow work, but it’s the essence of what we’re trying to do on Twitter. Sometimes tweets that you think will be duds aren’t duds, as they let you connect with an audience you didn’t know you had: for example, tweeters in the Middle East responded enthusiastically to our digitized Arabic and Persian manuscripts, and locals of Haslemere, Surrey enjoyed this map. But even the duds that are actually duds are still worth tweeting. For that reason, I would caution against setting hard performance goals.

Here are my favorite things I’ve done on Twitter