The expected (feared ?) news is out: Amazon has embraced the «all you can eat» subscription model. Kindle Unlimited (KU) is on everybody’s mind, lips and pens. Movies have been Netflixed, music has been Spotified and, now, books are following the same path. In a recent post, 24Symbols co-founder Justo Hidalgo is urging publishers to recognize that subscription services do bring tremendous value to the publishing ecosystem. In the midst of the Amazon/Hachette battle, the KU announcement has at least one merit. It raises lots of comments, arguments and counterarguments: will KU make the publishing industry KO? Is is good news or bad news and for whom? Do «all you can eat» subscription services have a bright future ?
Well, as always, nothing is neither totally white or black nor totally new or old. Nothing is either bad or good. It all depends. KU or not, the devil still lies in the details. Subscription, streaming, advertising as paying third party are indeed hardly new concepts. Think of radio, TV… They have been around for quite a while. The only difference is that Internet gives them a new lease of life that may or may not be disruptive. To avoid hasty conclusions on these potentially disruptive effects , a step back and a wider perspective are needed. The following lines try to achieve this. They are the output of my business musings and ruminations after almost fourteen years of book streaming and subscription. This is one specific (biased?) angle, but hopefully a good one to start with.
How does (should) subscription really work? Data, pricing and money talk
Cyberlibris, the French firm I co-founded 13 years ago, has been a very active player of the e-book and digital libraries arena since then. Over all these years, Cyberlibris along with French and international publishers has carefully crafted an «all you can eat» subscription streaming based business model for e-books. At the time Cyberlibris was founded, the DSL, WiFi acronyms did not even exist! We started our journey with the academic world and expanded it to ‘main street’ along the way. This has been a long road and a great place to observe and act.
Our view, nothing fancy but easy to forget, has always been to put the reader/user at the forefront. Indeed, understanding what a reader frustrations, when and where they materialize helps a lot when designing a proper answer to these frustrations. Take the example of the academic world. It is a strong international community. You don’t have to invent it. It is already there. The challenge, however, is to find ways for its members to become more efficient learners, teachers and researchers. And, in that respect the reasons for learners, teachers and researchers frustrations are numerous. How come the book I want to read for the exam is not available ? How come I can’t give my students several books to read for the course and so on and so forth?
Analyzing these frustrations (which by the way means that publishers too often leave money on the table), we designed several services including ScholarVox, a service dedicated to business schools and ScholarVox Sciences, a service dedicated to engineering schools. These are community based digital libraries marketed on a BtoB basis where books are accessible 24/24, 7/7, streaming-wise with no restrictions as to the number of simultaneous readers. Most of the above quoted frustrations were gone. But, the design would not have been complete without a properly designed pricing scheme. Fail the pricing and you fail everything.
The pricing model has to be simple, easy to understand and, last but not least, fair. It has to properly align partners incentives. After days of reflexion and market research, we concluded that an «all you can eat» subscription/streaming model with revenue sharing was what was needed. Schools, universities subscribe on a yearly basis on behalf of students, professors and librarians. In a sense, they (yearly) rent books which they never own as opposed to the physical library where owned books are stored. So eleven years ago we went live!
You may wonder why we did not opt for an individual subscription model. Well, first it is costly at to chase subscribers one by one. Second, the whole idea is to embed the digital library into the teaching process and this can only be done at the institution level. Third, which is in fact first, to avoid what insurance companies call adverse selection or anti-selection.
At the end of the day, the best metaphor I can offer for what we do is ‘group insurance’. Indeed, in group insurance, corporations buy a global insurance policy on behalf of their employees. This is usually good for the insurance company because with a large group of people at once the law of large numbers has a better chance to work and, more importantly, the issue of adverse selection may turn out to be less severe. In other words, with a large group, if you do the pricing job properly, you will benefit from the law of large numbers (not everybody has a claim at the same time) and you won’t end up carrying high risk people only, namely those who should have paid a lot more in the first place. Why is all this relevant to the «all you can eat» subscription model? Well, schools subscribe a global ‘reading insurance’ policy (they pay the lump sum annual subscription fee) on behalf of students, professors and librarians. What does this policy cover? Reading needs. And, it turns out with a large crew of students and faculty that heavy readers are compensated for by light readers, a kind of law of large reading numbers ! Not all students are work-alcoholic!
For this to be successful, publishers (and authors for that matter) have to be fairly compensated. Here is how it works. Publishers get a share of the «global insurance (subscription) premium» that schools have paid to Cyberlibris. We pay each publisher according to its percentage share in the total number of consultations across all publishers. By consultation, we understand a book page viewed on screen (one consultation) and a page printed (one additional consultation). Each month, each publisher receives a detailed royalty (consultations) statement which yields the overall amount to be invoiced and the split of that amount across books. This means that each time the same page is read again, a new payment is due. This is in sharp contrast with the sale of a book where whether or not the book is read the same single amount is paid, namely the one time price of the book.
This begs the ultimate question: how do you (fairly) price this insurance premium to make everybody happy? Well, somehow, using actuarial principles, just like an insurance actuary would do. In insurance, what they call the fair premium is the expected value of the claim, that is, in a nutshell, probability of having a claim x size of the claim. To get to this number right, actuaries blend ‘art and science’ in their computations. In our case, one has to figure out what the expected level of consultations is going to be given the size of the academic crew, the type of digital library it accesses and, last but not least, the time users spend on the service. One crucial element is indeed time. Indeed, time spent will determine how many pages can be consumed in that time span. This is where gathering data points is crucial and, in the end, gives you an edge. Because we have been in this business for the last eleven years along with more than 300 publishing partners, we have gathered a rich set of data to feed our pricing scheme. Somehow, a successful subscription service is one where smart time and content actuarial practices are in place. In the end, data and money talk.
Publishers have often a hard time understanding that the subscription price is not an increasing function of the size of the catalog. Indeed, the temptation is to conclude that the more books available, the higher the price should be. Publishers are inclined to believe that the more they are, the less they will receive. But this misses an important point: people have only limited time to read. Accessing twice as many books will not make them read twice as much. Moreover, users may be tempted to conclude that they pay too much for books that they will never read. Where does size matter then? In two areas. The first one states the obvious: the catalog has to be appealing. It will never have everything (but who has everything?) but it matches what users needs (again the design issue). In turn, being relevant means that you attract more and more users which defeats the idea of declining revenues. The second one is that with a large catalog, you have the ability to versionize, namely to market specific libraries to specific audiences. Books may indeed have several lives in several services. In the end, what induces readers to spend time on the service is the curation of the catalog. A well-managed curation yields relevance and serendipity. Serendipity because there are far more books that we don’t know than books we know. I am always surprised in my own field (I used to be a finance academic) by the number of finance books (I mean non best-sellers, not to mention non finance books) I should have known about!
The insurance analogy may help understand why many artists, such as David Byrne, complain about music streaming services such as Spotify and Deezer. These are individual subscription services which incur the risk of attracting heavy music listeners (the adverse selection phenomenon) who should have paid a lot more. Everybody pays 9.99/mo. But for some it is too much and others not enough. As Nobel Prize winning economist Georges Ackerlof would put it, you end up with a «market for lemons». To add insult to injury, music is a non rival activity. You can drive and listen to music (something you can’t with a book unless it is an audiobook). To make things worse, it seems that $9.99 is a kind of psychological threshold. People don’t want to pay more. So you end up with heavy listeners who don’t pay the right price.
As a matter of fact, the same holds true for book individual subscription services. The risk is that they attract heavy readers and end up with an unbalanced portfolio of subscribers. My guess is that carefully crafted vertical services have a better chance to succeed than horizontal ones. Indeed, in vertical ones, you attract ‘aficionados’ that are less elastic to price than subscribers to horizontal services. Furthermore, aficionados are more willing to add a new subscription (provided it makes their day) to the whole set of subscriptions we all have to pay (Internet, phone, TV etc…). With aficionados the right price will be paid even if it is North of $9.90.
To be fully honest and to conclude this first section, there is one frustration that quite early on our clients voiced. They told us that even though they truly enjoyed the simplicity of, say, ScholarVox, they felt frustrated not owning the books. But when you think of it, especially in the area of academic publishing (where obsolescence is quite rapid), you never ‘own’ physical books. Yes you do own a book published in a given year. But, you will have to purchase the new edition in 3 year time (otherwise patrons will complain that the library is not up to date). As a result, it is as if you were renting the book except that the cash-flow cycle is not the same as a true rent. You pay say $60 one shot for a print copy and $60 again three years after. Ignoring discounting this is a $20 / year rent! Not to mention the fact that you will have one way or the other to get rid of the old copies (because of lack of physical space). Hence the subscription model is not that remote from the ownership model in that example. Again the devil lies in the details : all books are not born equal!
As years went by we expanded our subscription services from the academic world to the corporate world, to the public libraries sector and to families. We know cover the full reading cycle of any individual: education, business and family and count readers in roughly 40 countries.
The economics of subscription services is simple and subtle. Simple because everybody knows what it stands for. Subtle because (and sadly enough nobody really talks about it) the room for mispricing is significant. Any pricing error can be paid dearly and the unfortunate result is some sort of Gresham Law where bad money drives out good. Pricing is not something that you draw from your hat. Pricing has to be part of the service design. Mispricing is a good warning of a poorly designed service where one does not really know who the readers are going to be. As a result, both readers and publishing houses are dissatisfied and rightly so. It should also be emphasized that most of the recent subscription services do not rely on a genuine shared subscription scheme. As far as I see it, they are more or less kind of Ponzi schemes: they pray that they won’t have to cash out to publishers more than they cashed in. Each time a book is, say, 10% read the service has to pay publishers the wholesale price. In other words, these services «pray» that lots of users won’t read much and that they will continue attracting this type of readers. This is bad because fully read books are treated the same way as partially read books. Rather unfair indeed. How come now that we can distinguish between fully read books and partially read books we don’t take advantage of it ? The unfortunate result is that instead of growing the pie and then sharing it, publishers capture the funding of these new services and, in the end, strangle them. Money changes hands without triggering growth. In a properly priced revenue sharing model, this will never happen: you can’t pay more than what is available to share. Incentives are fairly aligned towards growing the pie. This pricing misdesign is bad both for the service and the publishers not only on straight monetary grounds but also on qualitative grounds. Indeed, if you rely on the fragile assumption that people should not read much, you won’t collect what is the true gem of a subscription service: the reading data. And, if you do, it means that your readers read too much and you end up broke before capturing the gem data.
Is there life beyond (proper) pricing? The hidden treasure of reading data
On top of straight money, publishing houses can get and learn a lot from properly designed subscription services. Similar strategies to the ones that led publishing houses to market hard covers first, paperbacks afterwards, pocket books etc… are available. Take the following example by assuming the latest book of a best-selling author. What about having a pre-release of the book in digital format readable with no additional fee, say for one week, in the subscription based library. After a week of availability, the book is accessible through purchase/download only. Finally, after a few months, the book is again available in the digital library through the subscription service. After all, this is what Le Livre de Poche is all about. In the medium term, such a string of events does help screen the impatient readers from the patient ones. There are many ideas along the same vein that can be explored such as testing manuscripts before fully investing in them. Whatever the idea tested, it will trigger data points. This is where subscription based streaming services are hard to beat: they produce formidable reading data flows and readers leave footprints that are most valuable if properly understood, managed and used.
To truly understand what’s at stake, let’s go for a moment to places where books are traditionally read and sold, namely physical ‘brick and mortar’ libraries and bookstores. These spaces are a tribute to Euclid, the famous Greek mathematician, the Father of Geometry, and its no less famous postulates. Between two given books goes one straight shelf only. Two straight shelves never cross each other. As a result, musing in a physical library or a bookstore is a highly structured, organized experience. Books sit (and wait) on shelves, not any shelves though. These are the shelves that librarians classification have elaborated over years of metadata efforts classifying books. As a result, library, a bookstore are truly great interfaces (not all of them though!) where the eye can collect lots of information in one go. The physical space is structured such that the books affordance is maximized for patrons or clients walking along shelves. This is a great plus. On the minus side, these are spaces where not many reading data are collected.
Musing in a digital library is very different. Even though it is easily accessible once the subscription is paid, it can be a frustrating experience. Frustration again! Indeed, the reader is limited to a small 2D space where not much can be displayed in one go. This a hard constraint. This is a minus. On the plus side, lots of data can be gathered: books that are read or not, books that are assigned to digital bookshelves, books that are commented etc… New metadata are created as we are all metadata of the books we read and love. As a result we have two sets of data, the traditional metadata and the usage metadata. A natural question to ask is the following : How would one (spatially) organize the digital library once one takes into account all the available data.Which book would be close to which ones? Which books would not? The question is more ambitious than Amazon famous recommendation device. It addresses the organization of the whole library. It is not restricted to the neighborhood of a given book. To use a biological metaphor, with traditional metadata one can «compute» the genome of the library. With additional data provided by streaming services, one can «compute» its phenome. This is why we called our internal project the Book Phenome Project (BPP). BPP had a clear objective: to be able to compute for any digital library its social graph to, then, visualize it. By social graph, we mean the graph of the books contained in the library. The structure of the graph (the location of books, the links and distances between books) is influenced by both traditional metadata (genotype) and extra environmental data (reading habits, phenotype). To achieve this ambitious goal requires some modern artillery, namely the use of machine learning techniques and other big data and information visualization related tools.
The end result can be quite stunning. The BPP output is called the DICE : Digital Content Explorer. Hereafter are two screen captures of the DICE, one computed on www.scholarvox.com (digital library for business schools) and one for www.bibliovox.com (digital library for public libraries) :
Each point is a book. Each book emits a color halo which indicates to what topic it is related to. For instance, in the ScholarVox Dice example, red means Finance, blue means Economics and Decision Sciences. There is a striking difference between the two graphs. One mixes colors a lot and the other does not. Well, this tells that academic users don’t read the same way public library users do. The algorithm detects this from the data. As a result, books are not mapped the same way. In an academic setting, the motivation for reading is tightly coupled to the curriculum and to how the curriculum is organized over the academic year. In a public library setting patrons are not guided by any curriculum. They go freely by their tastes. This explains why cooking books are well identified, close to gardening books («art de vivre») while business books are in a different area. The obvious point from these two maps is that there is a lot to learn from «machine learning processed» reading data.
Assume that a given publisher wishes to know what its social map, in other words its clients graph, looks like globally (all countries spanned by the service included), next locally (say France or Senegal) and finally at a given institution in France. This is indeed possible and easily visualized. The next capture retains the following scenario: Service = BiblioVox, Publisher = Eyrolles, Field = Sciences, Location = Levallois-Perret (suburb of Paris).
For the first time, publishers benefit from a highly granular information drawn from the thorough analysis of reading data.
The DICE and its avatars will soon land on publishers desktops and allow them to map their books, to filter these maps (across users, location, topics etc…) and to trigger proper business actions. Publishing has so far evolved without any proper map, under a lack of proper and relevant information. These days are over. Rich maps are now available. Publishers will no longer venture in the dark.
This thin visualization is made possible thanks to the vast flow of reading data that subscription/streaming data are able to capture. Without these data it would have been impossible to design the maps. Note that the maps are valuable not only to publishers but also to all parties involved one way or the other with the service. This is for instance a great curation tool for librarians. One can easily imagine large screens on library walls displaying the DICE or its equivalents.
Last but not least, data trigger a virtuous circle whereby as more data are deciphered, pricing gets closer to fairness and appealing to more people and so on so forth.
Show me the money, show me the map!
With the irruption of Amazon’s KU, publishers face what was obvious fourteen years ago to the few pioneers of subscription/streaming/revenue sharing services such as Cyberlibris or ebrary: the digitization of books means that it will change the way they are consumed (not to mention the way they are produced and the way they are written). This has happened in the music industry, in the movie industry. It has taken longer because a book is not the same thing as a CD. Gutenberg (and its predecessors) have provided us with a genius invention that is both at the same time the hardware and the software. The print book is a marvel of design and autonomy. Content and device both at the same time. Hat tip Mr Gutenberg !
Hence, it would be silly to go head to head with Gutenberg. One has to figure out how to be orthogonal to Gutenberg. A good place to start with is uncovering current reading (and publishing) frustrations. Another good advice is to look at recipes that have proven their value over (a long period of) time. In that respect, the library is a great place to focus on. A print library has a lot to offer. This is why we still use them. But it has also weaknesses which Internet and digitization may help circumventing. As for most things in life, we have to face a recurring trade-off that Kevin Maney very aptly named the fidelity-convenience trade-off. As long as what we loose on the fidelity front is regained on the convenience front, we’re all right and willing to accept the new deal. This is what distinguishes a good design from a poor one. If subscription/streaming services are to succeed they have to offer convenience gains that compensate fidelity losses. Properly designed (and priced) subscription/streaming services can deliver what print books and physical libraries can’t deliver. This does not mean that both print books and physical libraries are doomed. Again, what matters is the trade-off equation. A print book scores high on fidelity, less on convenience. A digital library scores high on convenience, less on fidelity. The good news is that we can access both trade-offs, print and digital and are better off as a result.
What do all these business musings tell us about the KU event? Well, one word comes to my mind: coordination. Markets often fail because lack of coordination between market players, because of coordination costs. This is the famous prisoners dilemma whereby the two convicts unable to coordinate their actions end up with the maximum jail sentence. Had they been able to devise a joint strategy, their jail time would have been significantly reduced. The book market is replete with coordination costs : Think of book inventories for instance. KU again rings the bell of coordination. If there is a firm that has efficiently internalized coordination costs, this is indeed Amazon. KU is a natural output of Amazon’s coordination efforts. That’s why Amazon is so powerful to dictate its rules. If players don’t try to fairly and openly coordinate, Amazon will do the coordination!
This does not however mean that «la messe est dite». Publishers have a lot of dear assets in their hands, firms like Cyberlibris or 24Symbols too. The key to combine synergistically these assets is to talk, think together, spend the time to fully uncover (as I tried to do in the above lines) what a subscription/revenue sharing model truly is, truly does and what dividends one can draw from it. Yes, it takes time, patience and will. It is often said that well-established firms should think and act like start-ups. After fourteen years of spreading the subscription gospel, I am still puzzled by how poorly understood subscription is, how feared it still is. This is odd ! Again one has to put the reader and its reading environment at the forefront because this is where the money is. Given technological developments, how can one enhance the life of readers? Putting oneself in readers shoes is a great and profitable exercise that helps a lot reconsider Terra Cognita and embrace more actively Terra Incognita. Terra Incognita is not that scary anymore once one starts designing proper maps.
Subscription services shall not be designed in a vacuum. Yes, firms like Cyberlibris are there to push boundaries, to give the initial momentum. But, for the momentum to be sustainable and even stronger requires tight collaboration/coordination of all the players. One should no longer be begging for content as if content were in a sacred safe. One should rather open the safe and as Carl Shapiro and Hal Varian once put it in their Information Rules book, be aggressive instead of being greedy! Productivity gains are achievable thanks to digitization and Internet: a smart and profitable move is to redistribute them fairly. And, in that last respect, there is a lot that subscription services can deliver as long as their life is made easier. There is no reason to be KU or KO, unless of course the status quo becomes the business rule.