- 4/13/2015 6:53:41 PM
If we're talking about the biggest problem in your company, there's not much that the analyst will be able to do in the short term. The pressure will be to use whatever data is available to reach whatever conclusions you can. Any good analyst will make the risks abundantly clear. The audience will ignore all the small print and we'll just hope for the best.
In the meantime, every effort should be made to improve the data collection. After all this is mission critical, right? Hopefully, this allows us to put some integrity into the data before we suffer too much.
- by SaneIT, Data Doctor
- 4/13/2015 7:55:54 AM
@PredictableChaos, that's a fair statement and I agree that you can end up with a black eye but I'm curious how you would handle having partial data, everyone knows you have some data and they want to see something but you refuse to present anything. I know there is a point where you have to say "we need to collect data to be able to report on it" but many companies out there have little bits and pieces of what they really need and they want to get started on reporting now.
- 4/10/2015 6:15:31 PM
SaneIT - yes, you can use incomplete or flawed data, if you are "honest with yourself about how accurate it really is."
If anyone is considering that approach, I would caution against it.
In my experience, no matter how careful you may be to present the analysis together with the caveats or limitations or margins of error - everyone seems to forget all that. They only remember the summary. The conclusion.
And if that turns out to be wrong, you might be the one with the black eye.
It's not fair, but that's often the way it happens.
- 4/10/2015 10:49:02 AM
@SaneIT. I agree, the incomplete data set can be useful as long as you keep it in perspective, and know that you don't have data from certain customers. You also can look at that partial data set and figure out if the customers that it covers are representative of you customer base. That could increase its value.
- 4/10/2015 10:45:37 AM
@Tomsg. Right, the amount of work that goes into the "might be useful someday" data probably isn't worth it. I know that some people believe that every shred of data that you possibly can collect should be saved and made available to analytics apps. I'd love to hear from some of the people who campaigned a couple years ago for saving everything regardless of whether it had identifiable value. Do they still feel that way?
- by SaneIT, Data Doctor
- 4/10/2015 8:23:07 AM
I see the issue of constancy often which I guess falls into the suspect data category. I'm sure that a lot of companies out there could collect the data they just don't or they only capture part of what they really need and only capture that part of the time. If you start doing analysis on a database full of customer data but only half of your customers have a full set of data it skews your analysis quite a bit. Then you get into issues like systems adding defaults if a field is left blank. I'm of the mind that you can use this data you just have to be honest with yourself about how accurate it really is.
- by tomsg, Data Doctor
- 4/9/2015 8:24:45 PM
Some feel that even bad data can be useful. Sometime data can be salvaged and made sueful, but more often than not, the work required isn't worth the results you can get. better to start a data initiative.
- 4/9/2015 10:43:16 AM
As Fabian has pointed out in a number of articles here at A2, too often existing data is suspect, or poorly understood, or defective in important ways.
This may make your problem worse - you don't have the data to do your job. Yet to an outside observer it seems like there is ample data.
Water, Water everywhere; but nary a drop to drink.