A few years ago, I was standing in the CEO’s office of an international bank as he waved his iPad around.
“I want this connected to the bank network!”
“But what do you want to do with it?”
“I have no idea, but everyone else has one …”
I look at analytics, AI and machine learning, and hear the same rumblings from senior management teams. They want analytics and AI: after all, it’ll improve the business, reduce costs, enhance customer experiences and probably walk the dog while making dinner for 6. Spoiler alert: it makes terrible lasagne.
More often that not, I walk into a business and see Excel everywhere. Often these spreadsheets require more support and maintenance than any other system in the enterprise. I’ll be kind in my assessment – to me, spreadsheets provide a stale view of largely unstructured or cobbled-together data with incorrect formulae and baseless assumptions littered throughout. Employees at every level spend hours formatting and reformatting cells, simply to repeat the exercise next month. Because it takes so long to validate and format information, cut and paste ‘rules the roost’, further compounding inaccuracies.
Proper analytics tools will take data from its source and present it as useful information. You should have the ability to drill down into the data, should the need arise: “why did confectionary make less profit this quarter?”, drill down, “ahh – the sugar price rose by 10%”. Modern analytics tools should also allow us to predict properly – to use the current example – sugar will continue to rise as the effects of recent catastrophic storms hit us, so we’ll need to increase consumer prices within an acceptable margin so as to maintain a profit but not so much that it would prohibit sales.
In order to do this level of analysis, we need to understand how much the likely base cost will rise, its impact on production, and how much more before the customer, and would either considering moving to the competition or just considering the competition. If we’re on the ball, we can use modern surveying tools to determine customer sentiment and combine this with historical data (tracking previous sugar price rises due to storms and other factors) to feed into predictive analysis systems – and then present the result in an understandable way to those that make pricing decisions. Not only that, we can see how one ingredient price can impact an entire sector – and feed this through to anyone that can help mitigate this impact (sales, marketing, preparation, production …). If it’s a short-term fluctuation, we might bear the cost – otherwise, we’ll have to increase the sale price.
In simplistic terms, working out the correct response actually boils down to answering just one question: “what do you need to know?”. In this case, we need to know whether there’s still a profit in selling sugar-based products. Every day, we design and create analytics platforms and the trick has always been to understand what you need to know to solve your problem. Once we know that, we can design a system, define or import data – and challenge any assumptions before making visualisations or predictive models.
Unlike my banking CEO, I’d really suggest that you think before deploying analytics or visualisation tools: work out what you need to know. Then you know which questions to ask, and from that, you can work out how to collect, transform and process data. You can waste a huge amount of time making data look pretty, but if it doesn’t support your basic needs, then it’s only as good as an outdated spreadsheet.