It’s true: Technology is capable of automating a lot of what we used to do when it comes to analysing data. It can even take this a step further and simulate some of our thought processes. That said, technology has one shortfall: It is not human, and generating insights is an inherently human process that needs human traits to interpret what is happening.

Faced with a deluge of data, finding a way to combine these human qualities with the tools on offer will provide organisations with more opportunities to make high-quality decisions grounded in great insights.

I propose a ten-step approach to accelerate the process of generating and delivering insights, which forms the basis of the Define-Determine-Deliver model. The model draws on a number of sources. First and foremost, it is based on my experiences of working with some of the largest insight-driven companies in the UK and US. (Deloitte defines an insight-driven organisation as “one which has succeeded in embedding analysis, data, and reasoning into its decision-making processes”.) I was able to observe best practice in the way these companies collected and organised huge amounts of diverse data, and I gained a profound understanding of performance and how they were able to engage their people to take the right next steps, which led to stronger performance.

Second, the model takes up the themes being debated by practitioners, experts, and authors, in terms of how to organise and interpret the huge, diverse data sets organisations are now collecting. And the more diverse and complex the data, the greater the challenge of communicating insights.

The model consists of three stages. The define stage will help you clarify what you need to do and why. The determine stage offers a set of principles to help you generate insights, and the final stage looks at how to deliver your message to achieve the level of impact and influence your insights deserve.

DEFINE: PLANNING YOUR ANALYSIS

1. Be clear on the value of your insights. The beginning of the insight process involves being clear about what you are being asked to analyse. Over the years of working for a number of insight-led companies I quickly came to appreciate that the significant first question was not “what?”, but “so what?” Understanding the value (the “so what”) that your insights will add helps you engage with what the person requesting the information is trying to do. When you are informed and engaged, you build a more relevant and more focused analysis plan.

Tip: If the person making the request hasn’t already outlined the “so what”, asking them “How will the analysis help?” is a good way to understand what they are hoping to gain from the insight.

2. Partner with an expert. In my experience, those who seek help from someone who knows the particular area of operations well deliver the best insights. They could be a call-centre agent or warehouse manager, for example. Share what you are trying to do with them and ask their opinion. Their support can come in many forms. They may share their experiences of the topic being analysed, may highlight obvious pitfalls, or simply confirm that what you are doing is on the right track.

Tip: Ask the person making the request to recommend the right contact. Once you have a partner, be curious, ask good questions, and listen well to what they have to say.

3. Create a hypothesis. It is important that when you are doing your analysis, you don’t try to analyse all the data available because this could take too long. The process of forming a hypothesis will help you think about the relationships between your data, which should end with your forming an opinion (your hypothesis) on the answer you might find once you have done your analysis. A clear hypothesis, therefore, provides you with an indicator of what to look out for when doing your analysis, helping you to stay focused, whilst reducing any wasted effort.

Always create a hypothesis statement that captures this belief before you start analysing your data (eg, “product availability has decreased because supplier “˜out of stocks’ have grown as the cost of raw materials has increased”).

Tip: Take time to run through your hypothesis with your expert (from tip 2) or any other relevant people. This will help ensure you have a reasonable and balanced hypothesis, and help to avoid confirmation bias.

4. Visualise your analysis. It is all too easy to just dive in and start analysing data. Before you begin, be specific about what you need to analyse. This involves visualising what your analysis will look like once it is finished.

Tip: Get a sheet of paper and sketch out what your data will look like once you have collected it all, listing the rows of data down the left-hand side and the column headings across the top. Then sketch out the analysis you will carry out or the techniques you will apply. For example, do you plan to create a column of data that looks at the difference between two data points or a graph of certain variables? Be as specific as you can, as this will really pressure test what you are planning to do and whether it will add value.

DETERMINE: DOING YOUR ANALYSIS

5. Collect, clean, stay connected. Developing a plan of how and when you will collect your data is important, as this will help to ensure you have everything you need when you are ready to start analysing. Before you start the analysis, you will need to clean your data to ensure it is accurate, complete, and in the right format. There is nothing worse than unclean data undermining the credibility of your insights. Finally, staying in touch with your expert partner from the previous stage will ensure you get the most out of your analysis.

Tip: It is helpful to have a few (but not too many) expert partners. Picking partners with different types of experience is a great way to get a variety of viewpoints, leading to a fuller piece of analysis.

6. Analyse well. In practice, every piece of analysis is different. Therefore, adapt your approach using these key principles:

• Let the data lead you to the insight. Don’t assume you know the answer before you have done your analysis; this could really bias your analysis. Be open-minded and let the data lead you to the answer.
• If there is an elephant in the room, say so. Sometimes, when it comes to analysis, we don’t want to accept the most obvious insight; we yearn for something more detailed and more profound. But sometimes the most obvious answer is the right one, and it’s OK to accept it.
• Correlation doesn’t equal causality. Take care when verifying whether two variables are linked.
• Focus on what the business needs. If the person asking you for insights needs them in two days to assess an opportunity, then focus on what can be done in that time frame, rather than on the ideal piece of analysis you would produce given more time.

Tip: When analysing data, it is often more useful to focus on trends rather than on single data points. Trends often give you a more reliable view of what is happening. For example, if you are trying to determine which stores are driving low product availability over the year, then focus on the stores that are experiencing consistent decline over the time period (those trending downwards) rather than focusing on one store that had a low score for a small amount of time. (It would be interesting to know why, but don’t miss the big trends contributing to your low product availability.)

7. Bring it all together with a conclusion and indicated actions. Once you have developed some good insights, the next step is explaining what is happening and how the business should respond. This can be a daunting task for finance teams, as the fear of suggesting the wrong thing can create a lot of pressure. Grounding your “indicated actions” in insights will give you confidence in your proposal.

Tip: Seek to ensure your conclusion-indicated actions are correct by writing them out using the following structure: dilemma, insight conclusion, indicated actions:

“I conclude that the reason for ‘the shortfall in sales’ (the dilemma) is because store staff are struggling to get the stock out onto the shelves as the increase in customer numbers means they do not have enough time to restock (the insight conclusion). I propose a pilot project to increase staff in the stores with the biggest declines in sales. If this is successful, I propose a wider review of resourcing in our stores (the indicated actions).”

DELIVER: COMMUNICATING YOUR INSIGHTS

8. Prepare a clear insight message for your audience. The previous step, in which you generate conclusion-indicated actions, is based on what is happening and what you need to do next. The critical difference in this step is that you need to build an insight message to convey to your audience. The insight message is often the only part of your process that the audience sees, and if you want to achieve the right impact and influence, the message needs to be clear and engaging.

Tip: Do the “elevator test” to see if you are ready to deliver your insight message. If you were in the elevator with your manager, could you convey your message (the dilemma, the insights, your recommendation) clearly and succinctly in the time it takes to reach the right floor, all in a way that will resonate and inspire the audience to act on your findings?

9. Craft an engaging message. If you want to deliver an engaging message, then logic alone will not be enough. Engagement requires you to connect to people’s emotions. Your message may well have a good structure, clear visuals, clear arguments, and recommendations grounded in your insight findings. But you also need to build an emotional connection by finding the right tone, forming a connection based on shared aspirations, or focusing on how the proposal will directly benefit the insight requestor and their teams.

Tip: Stories are a good way of helping to deliver a more engaging and memorable message. Stories grab people’s attention, bring messages to life, and help link insights to the big picture. For example, if you are trying to put new customer service metrics into context, you could use statistics. “Customer service scores are at 60%. This is a reduction of 10% versus last year, and we need to do better.” Alternatively, you could tell a story that brings your numbers to life. “Last year we were not at our best for 40,000 customers. That is two out of every five customers that came to us. Here are some of the things our customers said and how we impacted their lives by not being at our best …”

10. Build an insight-led culture. Having a framework is a good way to accelerate the insight process. In the insight-led companies that I have worked for, this framework was embedded into the beliefs of their people, which was demonstrated every day in their behaviours. This level of engagement with the principles of the framework allowed these companies to accelerate insight generation, as well as to adapt those principles to address a particular problem when required.

Tip: Always be a role model for insights, giving your teams or colleagues the confidence and the right to be curious and to always seek out the underlying truth as to what is driving performance.

Source : FM