As anyone who has passed the third grade knows, all you need to form a line is two points and a ruler. Add another point and you can extend the line or draw a curve, and with one more point you can describe a third-order or declare a trend!
All of that is good in the abstract, but when it comes to predictive modeling you need more data … and some common sense. So I chuckled when I saw this article at TheTechnium predicting free Kindles this year.
The chuckle wasn’t that the idea was preposterous – in fact, here is what they think is happening, based on a quote from a TechCrunch article:
In January Amazon offered select customers a free Kindle of sorts – they had to pay for it, but if they didn’t like it they could get a full refund and keep the device. It turns out that was just a test run for a much more ambitious program. A reliable source tells us Amazon wants to give a free Kindle to every Amazon Prime subscriber.
As for that as an idea, I think it is fabulous … I have been a Prime member for less than a couple of years, but based on the recent unlimited movie streaming, which provides plenty of movies for less than an annual Netflix streaming membership, they have me for another year!
Adding a free Kindle would be a great way to simultaneously reward loyal Prime members, increase their ebook leadership hold, and potentially implement a price increase for Prime membership.
So why did I chuckle? Because there are two things going on here – first off you have the normal business cycle of being first to market with a new device class incorporating technology (E-Ink), followed by new products, competition, and natural price erosion to gain mass market share.
Then you have a targeted business opportunity based on a market growth strategy paid for by a tie-in incentivizing another paid service.
And finally you have a four point line showing price reduction, which is then extrapolated to a zero-crossing to predict a ‘free Kindle date’. All of which is validated by a nebulous non-answer from Amazon that the author took as a validation of his deep analysis!
Again, what is my problem? Bad math and mixed business models.
Anyone who has done basic algebraic linear modeling knows that the error bars increase exponentially outside of the zone of data – in other words, once you stop having data your confidence that the line will continue drops significantly. Considering that the time scale after the last data point is about the same as the entire data scale period … it isn’t even safe to assume linearity in late 2011, let alone predictive ability.
But that isn’t the worst of it – based on the graph and associated assumptions, the ‘free Kindle’ should be based on the same business model as all other prices: anyone gets it for the same price. By changing the assumptions to a specific group of already paying members, they invalidate the initial analysis and are guilty of the ‘bad statistics’ which are self-serving and give analytical reports a bad name!
So will Amazon offer Kindles to Prime members for free? I have no idea – I certainly hope so personally, and think it will be a smart business move … but if they do it will not be based on some hokey non-analysis!