In a recent post I defined the economic value of data as
"the maximum amount of money someone would pay to buy a data set"
I went on to explain that the data value research being conducted by Dr. Jim Short extends far beyond a buyer paying a seller for a data set. Dr. Short has uncovered additional areas where valuation is occurring:
- During an acquisition
- During a bankruptcy
- Monetization without sale
- Data Insurance
In this post I'd like to explore valuation in the context of an acquisition. In order to have a meaningful dialogue about M&A and valuation it would be helpful to find a recent acquisition where data assets were the centerpiece of the transaction.
Fortunately the recent 1.5 billion dollar acquisition of lynda.com by LinkedIn is a great example. Indeed, the CEO of LinkedIn trumpeted the importance of lynda.com's data assets in the following quote:
Lynda.com’s extensive library of premium video content helps empower people to develop the skills needed to accelerate their careers.
— Jeff Weiner, CEO of LinkedIn
So how much of the 1.5 billion dollars went towards the purchase of the data assets (versus other assets such as facilities, people, infrastructure, etc)?
The actual answer to that question would likely only be found via a direct interview of the parties involved (an interview that I am sure Dr. Short would be willing to conduct)!
In the meantime, we can only speculate, and Dr. Short has done just that by assuming that 90% of the "value" realized by LinkedIn is coming from the data assets themselves. Using this model, Dr. Short came up with the following calculations.
This approach by Dr. Short places an economic value of $112.88 per uncompressed gigabyte on lynda.com's data. If he had more insight into the acquisition this value could be higher or lower. Some might argue that Dr. Short should in theory factor the hosting infrastructure into the equation (and potentially reduce the cost of the actual data itself).
I personally think Dr. Short's approach is debatable. What I mean in saying this is that he has essentially drawn a line in the sand by openly discussing one method of calculating data value. His method is then open to debate so that the industry can converge towards agreed-upon standards.
If agreed-upon methods are in place, the M&A discussion becomes more bounded. If LinkedIn is essentially offering $112.88 per gigabyte, lynda.com can point at other valuations in the industry as a way of arguing that the price is too low. This is essentially the equivalant of a sports agent arguing the amount of a player's future salary by comparing athletic statistics against others in the league.
Data valuation during a merger can also be compared against data valuation that occurs in the other business scenarios mentioned above.
I plan on diving into each of these scenarios in future posts. I'd welcome comments from industry readers who have some thoughts or experience in this field.
Steve
EMC Fellow
>If agreed-upon methods are in place, the M&A discussion becomes more bounded. If LinkedIn is essentially offering $112.88 per gigabyte, lynda.com can point at other valuations in the industry as a way of arguing that the price is too low. - See more at: http://stevetodd.typepad.com/my_weblog/2015/06/valuation-during-acquisition.html#sthash.DbRUP9t0.dpuf
I think its also important to note the effective information density in the data. 10GB of MP4 compressed video will likely have a different density (referring here to both Shannon entropy-style information as well as a more semantic view) than 10GB of clickstream data.
So not only do we end up with a dimension that you suggest to start with around raw quantity of data (and the price paid), but also the nature of that data.
Then, how do we differentiate between storage and encoding mechamisms? Do we always, as alluded to, default to effective uncompressed values? And what about the meta data - I'm sure the data that LinkedIn has around person-to-person connections (the graph data) are just as important as the raw (I worked at XYZ for N years).....how do you value connections among the data, given they are tiny compared to the raw data?
Cool questions.
Posted by: Mcowger | June 30, 2015 at 01:21 PM