Text mining for marketing – the state of the art at IBM Business Analytics Live?

Yesterday, Verbal Identity were at IBM’s annual “Business Analytics Live” event. IBM were showcasing their analytical software, along with industry consultants showing what they’ve achieved with it.

We are doing pioneering work with our clients on text mining for marketing and it’s good to see what pioneers in other areas are doing. The high point of an interesting day was a talk by a policeman.

Is it time for a pre-crime unit in the British Police?
Keith Bentley, Chief Superintendent of the Greater Manchester Police (retired) showed how his force has been using data to track offenders, and predict likely areas of high criminal activity.  It echoed some of the points in an article in yesterday’s Guardian on the role of algorithms in our daily life.

Also of great interest was the work of Gary Seaman for the non-profit Medway Youth Trust. This group’s work with unstructured data shows that you don’t need a huge budget to conduct a worthwhile text analytics project.

It is time for us all to hang up our tap shoes.
IBM’s VP of Marketing, Caroline Taylor, stressed the importance of analytics for marketing. She talked about how CMOs would once “tapdance round the boardroom” rather than back up their ideas with sound data. Now, analytics can change this. We 100% agree.

However, ideas still need to come from somewhere, and the message that was lacking at the conference is that people are still the most valuable part of analysis.

Microsoft Word never wrote a novel.
Whether it’s people like us, bringing the linguistic and creative dimension to complete the value loop started by the text analytics, or Keith Bentley bringing the opinions and experience of his seasoned officers to bear in the creation of predictive algorithms for policework – analytics needs people.

Verbal Identity visits IBM's Business Analytics Live

A small selection of the novels Microsoft Word didn’t write.

 

Top Ten Tips if you’re about to start a project text mining for marketing.
Also of note was a very practical talk from Anthony O’Niell, Director of Planning for Irish telecoms giant Eircom, on how to lay the groundwork for predictive analytics best practices in your company. Here are his top tips. (And in brackets, our thoughts):

  1. Get early executive support for analytics. (We’ve found that a low-cost, ‘learn together’ proof of concept is enough to start producing insights and persuading the wider group.)
  2. Encourage enterprise-wide alignment to the use of data in decision-making. (Ambitious, but we’ve seen take up for text mining now from heads of Marketing, Customer Experience, Ops and Human Resources.)
  3. Hire “different types of brains” to ensure data is approached from all angles. (Data feeds the left-brain and the right-brain.)
  4. Define failure and success standards, and build a roadmap based around them. (And revise the standards regularly. We expect greater accuracy in Month 4 of a project than Month 1 – however good we were in Month 1.)
  5. Use dashboards for high volume requests. (Text mining is throwing up such clear insights, we say that our clients can walk into the Boardroom with the results straight after walking out of their meeting with us.)
  6. Democratise data: Make information available to everyone so that they can act on it (e.g. make NPS verbatims available online to facilitate text analytics approaches).
  7. Let your organisation’s goals drive your analysis priorities. (Since text mining allows you to see everything that is being talked about, avoid the temptation to data binge.)
  8. Make sure insights are actionable (It’s a lot quicker to change the brand language than the ops, the buildings or the recruitment policy. We have creative writers working closely with our analysts and linguists. If it’s not actionable, it’s not an insight.)
  9. Multiple projects will be needed across the enterprise, and must be co-ordinated. (Interestingly, we expect to find that the data we get from different departments to enhance each other.)

And, most importantly in our opinion:

10. Realise that all information sources are valuable, but the most valuable information is often the least obvious. (We believe that consumers now expect their favourite brands to have moved from broadcast to dialogue. They’re in a conversation with the brands, and with text analytics, brands can start listening properly.)

If you missed out on IBM’s Business Analytics Live and want to find out more, or if you’re interested in how our work with text mining and analytics is creating value for marketing and customer experience, get in contact by email or just prove you’re not a robot by answering the question below and leaving a comment in the box below.