“Algorithm development can seem like a boring technical detail – until you start recruiting specific segments for qualitative insights, or tagging people in your Brand Tracking and CX programmes.”
It’s very rare that a segmentation sits as a single, static information source. Brand and CX trackers, proposition tests, customer databases all benefit from having a segmentation lens applied.
Make sure that you have considered where the segmentation will live before confirming the segment solution. Need a very short algorithm to be added to a CX touchpoint survey? Need something that a sales force can pick up and use in their daily calls? This info might well have a fundamental effect on the segmentation we create.
Separately, tagging customer databases with segments using the Hybrid Segmentation approach needs a thorough understanding of the database before writing the survey or running any analysis. High-quality attribution requires large customer samples to take advantage of the most sophisticated Machine Learning algorithms.
Finally, algorithms themselves are not static entities. Most customer databases have gaps, and many companies are constantly upgrading systems or processes to create a better single customer view. We should look for ways to improve algorithms as the data quality and access improves too. Even outside of customer databases, we can fine-tune individual algorithms to make them work better across different markets or methodologies, or over time. If an algorithm isn’t working for whatever reason, there’s a good chance we can fix the problem and re-deploy.
