The Analytics Industry often promises that we can predict What Will Happen, or discover Hidden Correlations between seemingly unrelated variables in datasets. My experience is that we still spend too much time on Operational Reports that tell users What Has Happened in a single Data Set or Data Source. This post discusses why this is the case, and how to transition our thinking.
To avoid too much repetition in this post I will use the following terms:
- Operation Reporting to mean:
- “What Has Happened?”
- Predictive Analytics to mean questions like:
- “What Will Happen?”
- “What Correlations Are There?”
Why Is Operational Reporting Prevalent?
So why is there so much Operational Reporting, often from a single source? It is because:
- It has been the norm, we are comfortable with it. You buy an ERP, this is what you will get out of the box in many cases.
- The requirements received often request it.
- It is relatively easy to deliver. A bit of SQL knowledge is typically all you need plus a presentation tool.
- It is trusted, as there is a certainty since it has already happened and the report can be reconciled back to a data source screen in many cases.
- It is easy to understand. The end-user does not need any education in the accuracy of an underlying model.
- It is “set and forget”. The model to create the operational report tends not to evolve with changes in the underlying data.
- Self-service is much easier when delivering Operational Reporting. The skills are much easier to find in an organisation.
- A stack of transactional reports gives a sense of security, even if they are not necessarily used.
So What’s The Problem?
Given the many reasons for operational reporting, why is it not ideal?
- As comforting as the view of the sunset is in this photo, Operational reporting tends not to deliver insights. It just helps you keep a track of where the business came from.
- It helps to reinforce a think “inside the box” culture
- This type of reporting is easy to do, so it tends to proliferate. WIth self-service, it quickly leads to a large catalogue of reports that need to be maintained but are of questionable value
Given that you have reached this far I assume that you are interested in moving from backwards-looking to insightful predictions via Predictive Analytics. To achieve this goal you need to Educate, Challenge, Communicate and Analyze with varying amounts of focus:
First and foremost, take your organisation/client on the journey. At a minimum educate:
- Business Analysts, Analysts\Report Writers, Operational Staff, etc. in how to:
- Formulate their ideas into outcomes that have tangible value to the business (Revenue increase, cost reduction, reduced injuries, lower emissions, etc.)
- Challenge their ideas, and the ideas of their colleagues, so they are validated and meaningful to the business
- Present data effectively. This can be Verbally, Visually or in Writing
- Throw away KPIs and Metrics and have lost their effectiveness in driving change
- Analysts on the basic concepts that specifically do not need a “deep” statistical knowledge. These include Clustering, Correlation, Sample Size, Outliers, etc.
- Analysts on the use of existing tools (ETL based Data Preparation, Excel Analysis ToolPak, etc.) or new tools (e.g. NumPy, AI/ML, brute force discovery) available to them. If they are not aware they will never do it.
- Execs in the consumption of predictive content. It will be difficult, but not impossible to achieve without this.
Be prepared for some resistance. This wont be possible without challenging the status quo:
- If existing KPIs and Metrics are not driving change then replace them.
- Move to exception reporting. If the monthly reporting pack is too large to understand at a glance, there is an issue.
- Realise that this is a cultural change that needs the skills that come with a structured Change Management approach that takes into account the nuances of data.
- Actively manage Operational Reports:
- If there are similar reports that deliver the same content then consolidate
- If there reports that are not used then decommission them
- Question the report author why they need to create or publish a report. What does it actually tell them?
Communication skills are needed to get this across the line. It is not a “build it and they will come” situation:
- Identify the Execs who are prepared to communicate on the topic and get them on board. Be prepared that not everyone will feel comfortable with this shift though.
- Find “what will happen” thinkers in the business who have a view that is strategic. They can come from any area of the business (operational, exec, analyst, etc.) but they need to think outside the box, collaborate and advocate for their ideas.
- Create a forum where the thinkers in the business can collaborate and challenge each other.
Ask a technical resource which is the most important activity and this is where they will gravitate to. The actual analysis should be a lower priority than addressing the culture and thinking parts of the problem. If you don’t address these first you will end up with a room of “propeller heads” that are observed with curiosity by the rest of the business but not valued.
In carrying out analysis:
- Formulate a business question and ask your SMEs if it is meaningful before you start. There is nothing worse than discovering an “insight” that is known by others or considered trivial.
- Understand what the “business levers” are and focus on them. Pretty obvious really but we often assume that everyone understands them when often they don’t.
- Understand the longstanding thorny issues that impact the business, or the vertical. See what approaches can be leveraged from other verticals
- Make sure the technology, and approach being used, allows you to Fail fast. There is limited “good-will” for blue-sky thinking if it doesn’t pay off
- Just because new tools can be needed, it needn’t be scary.
- Don’t let this pursuit of insights become an academic exercise. It is very easy to get caught up in sexy new tools and techniques, at the expense of being productive in delivering value.
- Not everyone will able to change their thinking, and that is ok. Be prepared for this and use resources for tasks aligned with their strengths lie.
- Don’t think that throwing the latest tool at the problem with guarantee an outcome. If it doesn’t complement the skillset and inherent thinking, it will only confuse matters.
- Shake things up if you are not getting insights, however, do not let the pendulum too far. There will always be a place for operational reporting. I am not saying it should be avoided at all costs, just challenged.
This is achievable but needs to be approached from several angles.