[Editor’s note: This is the first part of a two-part blog post.]
For my birthday this year my daughter gave me a DNA sample kit from 23andMe, a company that promises to analyze your DNA for you and then post the findings in a private online account. The tests reveal information in three categories: genetic linkages to specific medical conditions (Do you have a genetic makeup that is linked to Alzheimer’s?), certain traits (Are you likely to become bald or accumulate earwax?), and lineage (Where do your ancestors come from?). The company styles itself as a technology company or a life sciences company, but it is in fact part of the growing number of data publishing companies. It presents a model for publishers in the STM area to follow, even as it raises the essential question: Why did not one of the large established companies conceive of and launch a company like 23andMe first?
Before you can have a data-publishing business, you need data. That’s where the consumer-facing side of 23andMe comes in. I will be describing the consumer side of the business in this post and the back-end data publishing model in the second.
For a fee that varies widely depending on how and where you buy the kit (prices range from under $100 to up to $299), 23andMe will send you a saliva testing kit. The kit comes by mail in a fashionably designed box that includes a tube for collecting saliva, instructions for use (poorly done, in my opinion, though I note that grouchiness is not hereditary), links to a Web site, where you set up your personal account, and a box for returning the sample. The service also has a huge amount of information and cautions about what the tests do and do not reveal, evidence that getting this product through regulatory agencies was a considerable challenge. The single most intriguing item to me was the assertion that 23andMe may also be able to determine if you have genes that implied breeding with Neanderthals. I mentioned this to my wife, who said, “Don’t bother with that test. I already know the answer to that one.” After about a month I received an email saying that my report was now ready for viewing online.
It’s an unexpected irony that for the most important information (that is, concerning the possibility of developing medical conditions), the reports are only as interesting as you have potential problems. I nervously looked up the report on Alzheimer’s, which killed my mother and grandmother, but, whew!, I discovered I did not have a genetic propensity for getting the disease myself. (These tests evoke vivid and painful family memories. So, for example, even as I was thinking about Alzheimer’s, I was reminded of my father’s characteristic super-lucidity as he lay dying of liver cancer. For him Alzheimer’s would have been a blessing.) I do wonder, however, what it is like for someone to discover a genetic link to a terrible disease. I am of the “I have to know” school, but people differ on this point. Some people are worried about the medical issues they will discover, some about their lineage. Some Kitchen readers may recall that James Watson chose to suppress information about Alzheimer’s in his personal genome. On the other hand, we have Andy Borowitz’s hilarious take on genetic testing, which I cannot resist linking to.
There is a great amount of information on 23andMe’s Web site, but I would be surprised if any of it is new to readers of the Kitchen. The writing is characterless, the explanations basic. Upon reflection it became clear why this is the case: 23andMe wants to build as big a database as possible and thus must write to the lowest common denominator. After all, it’s your DNA the company wants, not evidence of your culture or education. I read through the entire Web site and all of the countless emails the company sent me, but it was quite an effort to sustain my attention in the absence of new information.
On the other hand the site has a great deal of sophistication in all of its warnings about how to think about the genetic information and its chorus that it will not violate your privacy in any way. You can hear the lawyers behind this. The company clearly spent many months attempting to get regulatory approval to market the service. Notices about protecting your privacy accompany every one of the multitude of emails the company sends out, as though that were the only question anyone would have. So, for example, I had a question about completing a questionnaire, so I filled out a form for the customer service staff. The response came back: we will protect your privacy. But that’s not what I was asking.
The questionnaire is necessary because, beyond your DNA and the information that can be extracted from it, 23andMe seeks your medical profile, which then can be matched to your genetic profile. The questionnaire is endless. Before you fill it out, you are told that most people take 10-20 minutes to complete the form. I was at 40 minutes without finishing. I had to step away for a meeting; when I returned, the form was closed to me. I tried without success to get access to it again, but all I got for my trouble was a massive number of automated customer service emails (“Your privacy is very important to us”). It is surprising how poorly the user experience of this service is, considering the company’s access to venture capital. They could have hired someone from Netflix or Facebook, but apparently the head of end-user experience comes from Verizon or United Airlines.
The reason the user experience is mediocre is because that is not what the company cares about. The company wants your data, not your blessing, except insofar as your blessing will lead to your recommending other people to the service. And there are attempts to get you to pass the word or to buy a kit at a discount for a relative. You are invited as well to share your genetic information with others with a similar profile. The scientific aspect of this is that genetically we are all part of extended families. The marketing aspect is that of social media: have your users bring other users to 23andMe.
How will the company make money on the consumer side? I see two ways. The first is the obvious one of selling the kits. But I would be surprised if the sale of kits is profitable. They are expensive to make and distribute, and the back-end processing of the genetic sample can’t be cheap. That line of business is further complicated by competition, which keeps the prices down. The most notable competitor is Ancestry.com, a huge social media site — and a truly professional Web presence. Ancestry.com and 23andMe play this game from opposite sides and meet in the middle: 23andMe is a scientific service that uses social media, Ancestry.com is a social media site that leverages its assets (the social graph) to deliver a scientific service. There may not be enough room in the marketplace for more than one consumer DNA testing service, so we should not be surprised to see industry consolidation at some point. After all, the bigger the database of genetic material, the better, and one is thus better than two.
The second way 23andMe can make money from its consumer side is by moving the one-time purchase of the kits to an ongoing subscription for new genetic information. I got a (free) email today from 23andMe, promising new information about my genome. Presumably as genetic science advances, the implications of any strand of DNA will grow, so you can look for someone to sign up for regular bulletins. But this seems beyond 23andMe at this time, partly because there isn’t enough new information to support a subscription, partly because the company seems to lack a consumer publishing orientation. There is simply nobody on board who sees the opportunity in packaging more information for the growing user base.
If the consumer side of 23andMe were the entirety of the business, the company financial prospects would be bleak. But the consumer side is merely a way to get data into the database, and the cost of the kits is simply a way to defray that expense. The real game is in the back end, where the company package the aggregate for other markets. And that will be the subject of the next post.
2 Thoughts on "Publishing the 23andMe Way, Part One: Building the Database"
Hm… I may have some clues on why 23andMe are not going the subscription route. I just had the opportunity to look at the science behind it, and found that they are using gene (or more accurately, SNP) chips for GWAS instead of doing whole-genome sequencing. Now, gene chips are very useful, but the problem is that they do not sample a person’s whole genome, only a fraction of it. The design of the chip involves choosing which part of the genome you want to monitor, so a priori knowledge is an absolute requirement. If it happens that somebody discovers a part of the genome that is associated with an important disease, but which is not represented on the original chip, then a new chip would need to be designed and manufactured (and another set of samples obtained). Since the “new” association would most likely not be covered by the original IRB, then another IRB application would need to be submitted for approval. Let’s just say that the costs in time and money would be substantial.
There have also been some questions about reproducibility and batch effects regarding gene chips which only crop up when you have very large sample sizes. These are rather recent (and unexpected) developments, and may not be well known outside the Big Sample (Biobanks, essentially) community. So having multiple vendors providing the same type of service may not be a bad thing – and may even become a necessity if platform providers are unable to overcome it.