The value of data

How do you put a value on data? Clearly this is difficult to answer. Apart from anything else it depends on context. For example, knowing that you have a stock level of x is useless unless you at least know what typical turnover for that stock is.

The question, let alone the answer, gets more complicated if you start to involve data over time. That is, how much more valuable is a year’s worth of data compared to six months? Or three years’ worth? Clearly, in many situations, more data is more valuable. For example, suppose that you wish to forecast sales of one or more items over the Christmas period. The only logical way to do that would be to be able to compare previous Christmases, not just with each other, but with trading patterns in the periods leading up to these Christmases. According to SAS you need a minimum of three years worth data to be able to make meaningful predictions about Christmas trading.

Another problem with estimating the value of data is that the value can be derived from third parties. For example, if you are a Telco and you store call data records (as you have to) you can store these off-line, in near-line storage or on-line. But only on-line storage will enable the police and security services to do real-time searches against telephone records. Now suppose that such a search enables the prevention of a bomb plot. How much is that worth? Actually, that’s not quite a rhetorical question: it used to be estimated that a foiled IRA bomb plot was worth £1m. But that was some time ago and inflation needs to be applied to that figure. Moreover, the IRA used to give warnings about their bombs while Al-Queda does not, so the potential loss of life from a successful plot could be much more and the value of thwarting such an attempt all the greater. However, while we can, of course, all see the benefits of this, there is no direct business value to the Telco who is funding the system per se.

So putting a value on data is very hard. Nevertheless, it should be clear that whenever analysis of data is required, then the more the better. However, this flies in the face of what most companies have been trying to do for a number of years. As data volumes have grown the emphasis in many organisations has been on information lifecycle management and archival as a means of reducing the costs of keeping data on-line. But it is arguable that this is self-defeating. Certainly you save costs by archiving older data. On the other hand, the insight that you (or third parties in the case of call data records) can gain will be enhanced from being able to analyse more information gathered over a longer time. So the question becomes: how much more valuable is this greater insight and how does that compare with the additional costs of storage?

Moreover, the equation is changing, as companies such as Dataupia enter the market and allow you to store more data on-line at lower cost (and with a lower administrative overhead). This then changes the tipping point between the value of holding more data on-line for deeper analytical insight, as opposed to the cost savings of archival (which become smaller). The truth is that I don’t know where that tipping point is, and in any case it will vary not just by industry but by company, but it is certainly something that enterprises should be aware of and considering.