25 Jul 17

    IT reports are showing your age

    unlokq

    Reports are dead! Dashboards are dead! Perhaps a little alarmist, BUT, have we forgotten the main reason why the ‘data’ is important?… To allow us to take informed actions.

    Earlier this year I went to the Retail Technology Business Expo in London (as I do in most years) and sat in on some talks, some interesting, some less interesting, and some downright boring. But one of the statements that hit me was one that said; ‘we should take actions, not view reports’. BOOM! That hit me right between the eyes, because of its obviousness.

    Why do we create reports? To inform a decision, right? And those decisions lead to an action! In the speed of business in which we now operate, the only element of this chain from which you can derive any value is the ‘action’.

    As part of some other research I happened across a reference to a book, the premise of which was ‘what if our customers are not human?’ BOOM! BOOM! In the context of reports what is it that humans offer? Context? Interpretation? Well, the former can be learned, and at best the latter is founded on past experience also. At worst, it is prone to the subjectivity of the individual and the cognitive biases that are part of human psyche.

    In this world of Artificial Intelligence and automation why would we inflict the delay that reporting introduces between data and action, and why would we subject our business to the biases of individuals?

    History! The capability of IT has improved. Once there was no technology capable of performing the function. Humans had to act as an intermediary consuming output data from a function, either taking action directly, or re-working the data it into a format that can be used by another human as an input to the next process. In fact the volumes of data available can invalidate this approach, since humans would necessarily have to ignore most of the data.

    Bias

    It’s a job protection racket. The people producing reports will justify their existence by saying that they add context to the provided data, enhancing it before it progresses downstream. But what they are really doing is holding onto the data that provides that context. If this user released this ‘enhancing’ data so that it could be incorporated into the original dataset, removing bias. Furthermore, the actions taken, and the results of those actions, can then be used as feedback to further enhance the dataset. This would have previously been ignored, since humans are not great at learning (especially when in large organisations), see @R_Thaler Misbehaving GM car sales example.

    Worse still, humans may manipulate the data based on opinion in the absence of reason, simply because of a lack of trust in the data accuracy or completeness (potentially many times before actions are taken) skewing the data. In recent statistics it has been said that the majority of work in analytics is in cleansing the data. Now, I am not a data scientist or have any experience in business intelligence, and I bow to their superior knowledge, but I raised my eyebrows at this. Specifically at the removal of outliers. Was this not re-introducing human bias into the data? How to judiciously remove of outliers is exemplified in the discovery of the ozone hole

    Sidenote: In predictive environments that is harder. We learn from the past that this is likely to happen again, so we avoid it happening in the future, so then the system no longer learns about that scenario. Thereby creating a smaller sample size which are inherently less trustworthy

    Lean

    New businesses have fewer employees because they leverage technology to do the grunt work. They are more streamlined not by ‘removing’ middle layers of management, and associated reporting functions, but by not needing to employ them in the first place. This by the following rules;

    • Identify all the data that needs to be collected to inform the actions that need to be taken, this needs to be done by, or in conjunction with, people that have a deep understanding of your business.
    • Trusting the data that they have collected, and the algorithms in place to take actions
    • Including feedback loops from the actions taken to enhance the decision environment.

    The evolution of your business may have necessitated these positions, but due to the coupling forces of vastly increasing volumes of data and the increasing computational power to analyse it, this is no longer true. So, why exacerbate this model when the technology IS there? If you had the technology available from the start, would you have made the same decisions? Don’t fall into the ‘This is how we’ve always done it’ trap.

    So, with this done we can get rid of the enterprise users that create reports and dashboards, right? Wrong! Their role needs to be re-imagined, they have a lot of knowledge about your business, likely locked in their heads, that can be used to identify correlations in the gamut of data now available, in other words they can ask the right questions of the data, and identifying other data that is required to further improve the accuracy and trustworthiness of the data, adding the context.

    Can you change?

    Assert within your organisation that your (internal) customer is not a human. Its analysis relies on being fed with as much data as possible. The inference being that DATA is still important, the insights, and more so the actions that you take on the data are important, but the reports, and dashboards themselves are NOT.

    If your report creators are spreadsheet creators rather than question askers, and are not willing or able to cross that bridge then you need to start asking if they are the right people for your future business. Not because of being replaced by #AI, but not adapting as individuals as your business must in this time of Darwinian evolution where only the most adaptable business survive.