Here are some highlights from a few thought leaders and practitioners in the big data space that I found valuable enough to share – to me, they're hitting on similar themes so worth summarizing.
First, let's cut to the chase. Their key points are:
Vincent Granville paints such a great picture of the problem: he lists situations where crude analysis of fast moving, high volume data lead to bad and costly decisions. The best part of the post is a deep discussion of better methods to produce better metrics, including: effective sampling, weighting and modeling (with confidence intervals) and a commitment to identifying the lowest and highest quality data via rigorous data mining or integration techniques for identification, elimination (where appropriate), and analysis.
Ultimately better metrics is "better use of data science".
Avinash Kaushik (who is hilarious and always has awesome posts) is saying much the same: replace real time, flawed information (and focus) with right time data and focus… or better metrics.
He argues, the user is center stage and the information collected is only as valuable as the metrics developed and focused on. He goes further to say in digital marketing user intent is king - focusing on that (the better metric) translates to measureable impact.
Lastly I suggest checking out Kristian Hammond's post from the Harvard Business Review because it's along a similar vein and summarizes the "narrative from data". This flow is exactly what I'm accustomed to in the market research world. The key difference being the automation of data and analysis to stories (very nice) with again the key point being you need solid metrics to get it right - minus that and stories will be pointless.
With so much news and hype in the big data domain I'm always looking for more in-depth information that cuts through the noise and these seemed quite relevant.
For those of you who want more, this video is Tess Nesbitt's (from Upstream Software) talk about their big data work flow, I really enjoyed watching it – great real world use case of metric impact in a well-designed and sophisticated system.