Advanced analytics, leveraging techniques such as machine learning, is a core part of the work that ECI’s Commercial Team do to support management teams across our portfolio in achieving their growth aspirations.
Examples of where we have deployed advanced analytics include building churn prediction engines to identify at-risk customers for more targeted support, or lead scoring incoming sales opportunities based on their predicted likelihood of converting.
We have transitioned away from Excel analysis, where 88% of data workers or approximately 47 million people worldwide perform their work, to newer, more sophisticated tools, such as Alteryx which allow us to blend datasets and perform advanced analytics. These new tools are helping us take advantage of techniques such as machine learning – becoming so called “citizen” data scientists and enabling the companies we are invested in to reap the benefits.
In October, Alteryx hosted their Inspire Europe event in London, where several members of the ECI Commercial Team were in attendance. Over two days there were almost 100 forty-five-minute breakout sessions to choose from, delivering training and sharing best practice in data preparation, data science, and visualisation.
There was a lot of discussion at the conference around how to set up an advanced analytical capability in a business, from the board level support required, to the pros and cons of the various routes that can be taken in terms of approach (i.e. do you have a crack team of hired in experts, or do you focus on “democratising” advanced analytics capabilities and bringing the entire organisation up a level?), how to integrate the capability within the existing organisation’s way of working, and how to recruit, retain and manage “genuine” data scientists.
From building capability to the ingredients for success for an advanced analytics project, a number of insights emerged from the conference:
Advanced analytics projects – key success factors:
1. Early analytical involvement – having someone with advanced analytics experience involved early in the project, during the scoping phase, can ensure the most appropriate and useful questions are being answered.
2. Outcomes and timelines – setting the approach and timelines early, reporting progress regularly and managing expectations can all help a project run smoothly.
3. Acceptance and actionability – it is important to set the acceptance criteria up-front, so that there is a shared understanding of what success for the project looks like and that the business can action the insights.
4. Start with a literature review – to avoid any duplication of effort, internal documents and academic publications should be reviewed at the start of the project to look for existing analysis or insights that can be tested with new data sources.
5. Reproducible methods – automate as much of the data preparation and analysis process as possible and avoid making any ad hoc changes to underlying data or processes, as these are more challenging to update and interpret.
6. Version control – keep track of all changes with robust version control, particularly where there is collaboration or frequent changes to the underlying datasets. Use comments to explain the steps you have taken to anyone who revisits your analysis in the future.
7. Parsimony – if two potential solutions are equally accurate, generally the simplest solution is best. The trade-off between accuracy and interpretability is something that should be considered in all projects.
8. Test everything – throughout the project all assumptions, datasets and models should be tested and challenged thoroughly. Data driven decisions need to be sense checked or the business is likely to reject them.
9. Citizen data scientist – propagate the knowledge and insights to help everyone in the business share the same analytical language set and avoid the “black box” trap which can lead to the business being suspicious of the results.
