I’ve been wondering what you would call more advanced segmentation if web analytics products and even “analysts” refer to combinations of filters as “advanced segmentation”. I guess “algorithmic segmentation” might be a good term to convey that there is some, gasp, math involved. Plus, I like the idea of algorithmic trading from the investing world.
From what I can tell reading the recent tweets and blogs, many (but not all) mainstream web analysts are finally starting to learn about RFM (recency frequency monetary) despite decades of usage in catalogs and even a decade of posts from purely web analysts who’ve written about it already. RFM certainly seems like a fine start. It’s intuitive, easy to explain, you do not have to limit it to R, F, and M, and you can use business judgment to weight each factor. However, it’s not really a data-driven approach to segmentation as far as customer analytics go.
It’s like the old saying about OLAP-based vs. data mining based approaches to analyzing data:
- what data match this pattern? vs. what pattern can be revealed from this data?
Rather than getting bogged down in a micro-level discussion/decision about what manual judgment factor constitutes something like “engagement” - such as a mouse scroll - that may or may not be a significant independent variable in predicting a more useful metric such as profit or repeat purchase or response rate, it seems much more interesting to learn about more data-driven “algorithmic” techniques such as cluster analysis or decision trees. Just a quick note. More later…
