This article is written by Matt Dent, BI Consultant at Data3
Dashboarding is not Business Intelligence (BI). Nor is reporting. BI is not creating a propensity model, nor is it creating a customer segmentation. BI is all about value, delivering actionable insight; so that decision makers can make informed decisions. Business Intelligence brings robust, reliable solutions to decision makers; Business Intelligence is a series of events to bring out the most useful and informative data.
Before I get into the detail of Business Intelligence and Analytics, it is worth a foreword on ETL. Extract, transform and load. For any successful BI project or journey, you need good ETL. This step is crucial to having clean data that is in the right shape for the job. Without ETL your data won’t be in the right shape to create valuable insight.
I’ve created these 8 windows along the analytics corridor to help display an 8-step BI journey. Let’s go along and open them up to see what is lurking behind.
Let’s jump in with number 1, our good friend, Visualisation
Although Dashboarding is not BI on its own. It is often the first step of the journey. First, we need to display the data so we can inform people what they don’t already know. This is usually done by utilising an array of visualisation methods, usually in the form of charts. The number one mistake in dashboarding is thinking that more is more; remember, BI is about delivering value, not delivering an indigestible chart on everything your data could possibly tell you. Less is more. For users to derive value, good visualisations or dashboards need to be immersive and interactive. The ability to drill down and drill through and explore in more detail brings power to the dashboard and will leave users wanting to use their solution regularly.
Digestibility is key but alongside this, cost and performance are crucial. No one is going to use a beautiful, concise dashboard if it takes minutes to open, and 10s of seconds on anything they want to explore. Speed is another key to the visualisation puzzle. Alongside this, cost is imperative. Visualisation needs to boost ROI by allowing the user to find ways of making and saving money. The visualisation method itself cannot be a financial burden, and therefore a sensibly priced solution for the job needs to be factored in.
So, step 1 is all about creating visuals that will deliver value to the decision makers to help them understand more about their business to make the right decisions and the right time. The keys to successful visualisation are for it to be concise, fast, and affordable.
What about step 2? PING
Step 2 is all about alerting. You may have made a nice dashboard in step 1, but dashboards are dead… apparently, according to ThoughtSpot. Let’s think about this. You’re taking data, creating ETL, and then visualising it in step 1. If you’re really good at step 1, taking value should be easy, but as a stakeholder, you’ve still got to take the time to find the value. What if you want to be told automatically when something important happens? Well, this is BI alerting. We can setup alerts to notify our stakeholders when something important happens, for example: I want to know when my expenses this month go above £100K. Or more intelligently: I want to know when my expenses surpass my average monthly expense.
Decision makers can still have their dashboards. But there is so much value to be gained from alerting. Alerting drives action by the end user. In tandem with this is data actionability within the dashboard. Being able to push actions or data out of the BI tool into other parts of your data infrastructure is a powerful and efficient way to make data driven decisions.
Step 3, lets undertake some projects.
Analytics. To give you more than just reporting. These are the bigger, deeper, insight projects. These will range from: understanding lifetime value of different cohorts; campaign analysis across different channels; or customer demographic analysis. These are the sorts of projects that shine a light on insight that you wouldn’t usually be able to get out of a standard dashboard. However, once you’ve manipulated your data to this extent, you can automate that into regular reporting, but we’ll get to that later in step 7. Alongside these defined projects there is also exploratory analysis; the Adhoc slice and dice of data to further understanding and fulfil curiosity.
Step 4, cut and cluster.
Segmentation is all about grouping up your customers into similar pots. This is most useful for marketing campaigns and can be executed through many different techniques. Here are a quick four:
- Behavioural – How do customer’s behaviours interact with your business, for example; are there customers that buy the same products, from the same channel, along the same frequency? Learning from your historic data who to target for behavioural types
- Psychographic – Generally not one for UK and EU laws, but if you can do psychographic analysis, this will be interpreting customers or potential customers psychology and how they marry up with your products. For example, do they have an affinity to green companies? Are they introverts? Impulsive? Risk Taking? This data is often teamed up with demographic data. Clustering this data can be incredibly powerful, but only if it is legal
- Demographic – Sometimes a divider on the legal or legitimate interest front. This usually uses attributed data rather than actual personal, individual data. Demographic data is often mapped against census data at a postcode, more generic level, and includes data like age and wealth. Again, this can be grouped together and assigned to customers to understand more about the type of person they are and how they might map out against your offerings
- Geographic – This one is quite self-explanatory. Grouping people up that live in similar areas. You know that people in South West London have more disposable income, so you put your more expensive products in that store, and you cheaper ones in your South East London store. Or you know that your audience in Bristol is younger than your audience in Poole, then maybe you can promote your digital channels in Bristol rather than in Poole.
The thing these all have in common, is that they drive value, inform decisions, and even make some decisions for you. These segmentations don’t need to be one off projects. These need to be updated and refreshed over time; this allows them to always be usable, but to also be able to track movement between segments over time.
Number 5, let’s apply some Maths:
Want to forecast out your income data against factors that you can change? Want to group up your customers into pots based on their propensity to respond to your emails? Or maybe something even more, such as econometric modelling, where you want to try and attribute factors you can’t join up: for example – You want to attribute all your marketing efforts to your total sales. All these modelling methods are additional analytics projects above and beyond traditional reporting and dashboarding. They can all derive huge value.
Okay, number 6, time for science
We’ve had some data Maths, let’s go and see some data science. Nowadays, data science is quite a loose term, some claim to be data scientists when they can pivot some data in Excel. Others however, can draw huge value from data with advanced levels of data manipulation, using techniques such as Machine Learning or Artificial Intelligence. Allowing computers to process and model your data can draw deeper conclusions from your data than a human may be able to. Again, this is a way we can derive value from our data to help inform our business decision, I’ll keep this under the banner of Business Intelligence.
Our penultimate step, automation:
This step is all about making your analytics, maths, and or science projects reusable and repeatable. So, it makes steps 3, 4, 5, & 6 even more valuable. If you put all the time and effort into creating your projects once, make them repeatable, using up to date data. You’ll get repeatable value from your hard work, with up-to-date data, and more business value reusing your development work that you put in to build these projects in the first place. If you do something more than once, then there is a business case for building an automation for that process.
Step 8, we’re at the end of the corridor, or are we?
This is less of a step, more of a way of looping back around to step 1.
You’ve put all this work in to create good visuals, you can action them, and alert them. You’ve gone a step further applying maths and science; and you’re automating these projects. Now you need to go back and visualise all this fantastic data you’ve worked so hard to create. You need to display the outputs of your projects and the value they drive. You’ll be able to see where they can improve and alert on performance.
Is this really 8 steps, or is this a continuous BI cycle to continued data knowledge, value and empowerment? Step 8 also goes beyond BI. Use this step to inform other business processes and continue to improve and develop your systems. BI can be the team and processes that join the dots across your business functions.