(Hommage au big Lebowtsky).
I was enjoying leisure and was pretty happy with myself, but annoyed.
Annoyed by the non sense around big data. Big data does not measure in how big your data are but in how an heavily significantly simple data comes out.
Big data should be the ultimate data.
Now, let's try to do what all good programmers before coding.
Relax, close your eyes and imagine the better world for your ultimate customer.
Relax even more. Would I switch place with him?
Holy cow, yes!
But, I can't take his place. It sucks to be me. I want to be that guy. So I imagine I become the boss as a fraud. How would I keep my place?
I need something. Something like BIG data so that I can be easily, lazily make bucks. Big buck$. I need the most simple designed evilish software to ease me the trouble of working.
That's how you think. And you are right.
What is it?
A something that makes the business alive. That I can read easily pretending I am a wizard and no one will understand.
Imagine a simple digital clock like device with one info on the wall. What would you put?
The actual flux of income/outcome without any artificial filters.
Imagine how trepiding it must be to have a direct "load" for your business.
Is there a bottleneck in your production? Your load will stagnate.Is it day or night for your customers? You will see if it matters.
Is it Christmas holidays? You will see if it matters on your load.
Is there a bug in a software that results in gains? You can give the coder a bonus to keep up doing the real bucks.
Is there a feature that results in a direct gross loss? Well. Even if it is correct, these people are gonna kill your business.
It would be damn fun and exciting. Just wonder when a good news happens, you have to understand why.
You need to have flux of raw informations maybe coming from other channels (news papers for instance) and a timelime of your value to make the correlation correctly.
You can actually read the results in real time. You could even have deconvolution filters to erase trends, saisonnal activities to improve the results.
And like a load on a computer every scale time make a difference. But actually a skilled eye see the patterns fast so deconvolutions maybe overkill.
And, now, You too can be the big dude with big bucks, with only one skill: looking a figure. Being a boss is just a guy that pays attention to always have a positive cash flow on at least one time scale he controls.
You have to respect lazyness. So, you relax more, 'cause you are not Rockfeller and you have high standards. When you will be the boss, you want to go to Tahiti by cargo boat.
So you think of delegating. You are lazy, but smart.
So you have to be able to delegate the coding without being coned. Because, if you hire another yourself, you know it is easier being well paid selling your boss IT craps than actually trying to provide real services.
Does it requires bigger data?
Nop it requires the ultimate big data: the direct results on
the cash flow of every actions in real time expressed as a figure. Or anything that correlates it. One figure to rule them all: cash. We human are good at out of the box thinking. If I were having a hairdresser's franchise, I think I could correlate the water's consumption in all franchises with my activity with a certain confidence. That might me my ultimate data in this situation. Simple data correlated with my business. It won't be precise; I won't see the financial part.
But what about the finance if I lose my customers faster than my hairs?
It just require smarter ways of having datas. Growing the size of data in an uncontrolled way may not be the solution.
Smarter data may involve a little bit of an archeological domain called "empirical science".
It involves studying our math to sample a smaller set of data and be able to give results with their errors.
Adding errors to the ultimate data is also ultimate.
It enables a controlled tradeoff on cost over exactitude. Precision is about the quantity of information, exactitude is about the quality.
big data is precision.
But ultimate data is cost effective exact data handling with their level of confidence.
Not nice figures unchecked but labeled «exact». Less information that are more significant with their level of confidence that are reliable: data you can trust that now leveled up to the rank of information! You want a small information in real time: are my decisions good or bad. In fact, the trouble is this figure is a glasshouse. Everybody can also see the impact of your wrong decisions. A good tool should be dangerous, else there is no fun.
As
soon as your IT teams will improve, then your error margins will
improve (at the condition you cut some heads every time a value happened
outside of the confidence values previously made and other
precautions).
If you are really paranoid add the IT related cost per customer on a side channel.
Nowadays, the data are growing exponentially with the size of your graph. And, also because of added dynamics, it grows more than linearly over the lifespan. And the more people uses messaging based system for the more operations, the more growth you add.
If we follow the reasoning then big data is just overwhelming bigger data: companies that buys are doomed. Data are consuming electricity to point the less. The OPEX are more than linearly growing by customers&providers. The bigger you become, the more vulnerable until you reach a lock in situation.
Your interest is to control your cost per customers. You want to diminish your cost per costumer while your base of customers grows, not the opposite. If I were your competitor I would not contradict but rather encourage you.
Anyway, the ultimate big data: I will build it with my friends; for us to become the big dudes, relaxed, and having time to spare. It can be done. You just have to relax and focus on what is the essence of the data, not on its accidental nature. The essence lies in simplicity, clarity causality in accord to your goal.
Ultimate data is the simplest tool for measuring in real time your success and failures according to what matters the most. No more, no less.
