Hit enter to search or ESC to close
As the year draws to a close, the media is full of predictions for 2020 and reviews of the accuracy (or otherwise!) of last year’s predictions for 2019. Much of this is based on opinion; experience combined with an ability, real or imagined, to read the runes of what has gone before and to apply it to future scenarios.
In the world of data science, predictive analytics is a huge growth area intersecting as it does with our insatiable desire to understand the future before it happens and our love of technological innovation. Market data shows that the global predictive analytics market was valued $4.5bn in 2017 and is expected to reach $16bn by 2026. But what exactly is being predicted and how can these techniques possible be applied to the highly complex, and dare I say, nebulous world of reputation?
Irrespective of the technology being used, the vast majority of predictive modelling is based on the same basic technique; model the future based on a deep statistical understanding of the past. My credit score is effectively a prediction of my likelihood of meeting future credit repayments on time, based on my credit history, payment of loans and so forth.
The key element here is that the score is a rating of a specific scenario – the likelihood of me making future credit payments on time. By narrowing the data set to a specific scenario, we can much more accurately rate and rank the likelihood of something happening or not happening within a certain error margin.
Applied to reputation, the same principles should be applied. The question or scenario cannot be as broad as “what will my reputation be in 12 months’ time”, nor “what reputational risks will I face next year” (still the preserve of the crystal ball and the tea leaves) but rather “if I do xyz, what is the likely impact on my reputation?”
To put that in more practical terms, if an organisation is considering committing to removing non-recyclable plastics from its products, we can analyse the data from all other comparable companies (similar sectors, stakeholders, markets) who have undertaken similar initiatives to calculate the average ‘reputational return’ they have received from doing so with each of their stakeholders.
This will effectively provide a range of historic impact scores for each stakeholder group, which can then be correlated with business KPIs (sales figures, share price, applications per role etc) to calculate a likely real-world impact of undertaking the initiative. In a nutshell, if we do x, we are likely to generate y.
The cornerstone of predictive analytics is still very much regression analysis, which is increasingly being enhanced by machine learning techniques such as neural networks.
In our plastic-free scenario, while it is useful for us to understand the likely range of reputational return we could accrue from the initiative, if we are to proceed, we would want to further understand how to ensure the highest possible returns.
Across the many different examples of companies making announcements to remove non-recyclable plastics, there will be a wide array of different variables present in how they undertaken this. Some companies may have issued specific targets, others may have partnered with NGOs, some will have moved earlier than others, different channels will have been selected etc.
Machine learning models can test the relative impact of each of these defined variables via a process known as Supervised Learning or, can surface these variables independently, which is known as Unsupervised Learning (specifically cluster analysis). This effectively allows companies to build a playbook of the most effective ways in which to execute their announcement for the greatest reputational returns.
The combination of greater sophistication in machine learning techniques and human ingenuity in posing different questions in different ways means that we can more accurately predict the impact of certain events and scenarios than ever before. This represents a fantastic advancement for decision planning, budget allocation and risk management.
Just don’t ask us for our predictions for 2020!
Related posts
Be part of the Stakeholder Intelligence community
