I had the pleasure today to join the #AnalyticsChat "How the Next-Gen CMO Harnesses Analytics" hosted by Gavin Heaton (@ServantOfChaos) from Constellation Research Group. It was a blast of tweets, a lot of fun - but also made me aware of some fundamental misunderstandings around the what 'analytics' is about.
Unfortunately, the marketing professionals are doing the analytics industry a bad service by chasing the latest buzzword and relating analytics to anything Business Intelligence, Analysis and even Reporting. Wrong, so beware the false analytics.
Not enough skills is the problem - this is why Analytics is hard and 'stuck' in Phase 1 / 2 of the Persona Driven Innovation Model (PDIM). A marketing professional can not create analytics - the tools for that are not yet made user friendly enough. And - to find the right predictive model to drive to the action is equally not solved.
Then Start asking questions. Many of them. More. Remember there are no stupid questions - only stupid answers. Because the analytic software vendors have not solved the 'holy grail' of analytics yet - serving you the answer through the automatic selection of the right models - you need to run them yourself. Find a way to automate running them - as today's lame questions maybe hot on tomorrows (enriched) data.
And here is the storification by storify of all the tweets I was involved:
(1) Analytics! - Beware the 'false' analytics:
As mentioned earlier, analytics is the combination of analysis and derived actions (or at least guidance).
 |
| From Wikipedia. |
(2) Too much data
The data problem of analytics has been solved by BigData - namely Hadoop and MapReduce. No need to worry anymore if you can afford to store the data and how to store it.Not enough skills is the problem - this is why Analytics is hard and 'stuck' in Phase 1 / 2 of the Persona Driven Innovation Model (PDIM). A marketing professional can not create analytics - the tools for that are not yet made user friendly enough. And - to find the right predictive model to drive to the action is equally not solved.
(3) Which data to store - which not
Of course you need to store all data you can get your hands on. What may not look like important today - maybe relevant tomorrow. The business impact of not storing the data, will be very likely to outweigh the cost of storing it in the meantime. So don't be afraid to store all that you can get your hands on. Don't fall in trap from the time before the BigData era of thinking you can save money by storing less data. You won't. And you will regret it.(4) Too many silos of data
The history of enterprise automation is full of data silos. But since ERP - Enterprise Resource Planning - the silos are something not caused by the vendors (yes there are exceptions) - but by the users preferring to control their own silos. Except for the CEO, CFO and CIO - all other executives like their silos - including often the CMO. And again - with the answer of (2) storing the data is no longer a problem.(5) Start with the questions - No, start with the data
Usually a smart approach for all kinds of engineering - but not for analytics in the current state. With the data storage problem solved - but not the predictive models for the action element. So we don't know which models to run and we usually also don't know which questions to ask. So don't worry about the questions. This maybe even detrimental to your analytical success as you may end up ruling out data categories, that you may need later. Store all data you can get hold off. No discrimination. It won't cost you much. Keep adding data as you can get hold of it. Never get in a waterfall project mentality with data - you need a continuous data collection and enrichment strategy.Then Start asking questions. Many of them. More. Remember there are no stupid questions - only stupid answers. Because the analytic software vendors have not solved the 'holy grail' of analytics yet - serving you the answer through the automatic selection of the right models - you need to run them yourself. Find a way to automate running them - as today's lame questions maybe hot on tomorrows (enriched) data.
(6) Content, Context and the never ending debate - store all data
Don't worry about them too much, store what you can get. Then ask questions. What is one questions context maybe the content of the next question. The recursive nature of data can make nice discussions - but should not hinder you from your BigData and Analytics success.And here is the storification by storify of all the tweets I was involved: