There is no one right method to use your customers’ transactional data to surface the differences and commonalities that will enable you to segment them. It really comes down to balancing your existing resources and expertise with external help in order to reap the benefits of segmentation in a timely manner.
This article is part of our series on customer segmentation. You can read about the benefits you get from using transactional data, or use our simple framework to determine which level of automation you need.
You have three levels of options
Manually
There’s a reason why transactional data hasn’t been extensively used until recently: raw transactional data is a noisy, unstructured dataset. So while you can do everything in-house, it means hiring developers and data scientists to categorize and transform the data to make it usable.
✅The pros: everything remains in-house, which is important for some businesses.
❌The cons: the opportunity cost, in terms of time, capital and other resources, might not be worth it. There’s a big chance this method requires you to move away from your core business.
Using categorized transactions
Some data aggregation providers are also able to deliver categorized transactions instead of just raw transaction records. This means your team will have a level of understanding for each transaction: the categories it belongs to. You then have to apply functions and move that data around to make it usable in the context of customer segmentation.
✅The pros: this method is considerably less expensive than doing everything manually, and can yield very good results if you know what you want to look at.
❌The cons: using categorized transactions can be very confusing to start with and still requires a fair bit of work from your team. The categories you receive from your aggregation provider means your data is clean and structured — but it hasn’t yet been transformed into useful insights. In other words, you need someone with a data science background to aggregate those categories together and produce insights.
Using insights from Attributes
When we set out to build our data enrichment tool Attributes, our goal was to go way beyond simple categorization capabilities. We wanted to output actionable insights about individual customers that our clients could use to feed their decisioning models: total income, spend for different categories of expenditures, minimum balance across detected accounts — the possibilities are endless.
As it turns out, these insights are also exactly what you need to segment your customers based on information extracted from their transaction history.
✅The pros: using Attributes, you take away the vast majority of the data science work — but you still own the actual segmentation process.
This doesn’t turn segmentation into a black box: you’re in charge of deciding what goes into understanding your ideal customers.
Attributes processes the raw transactional data and outputs data points that are relevant for you. The only work involved is to then take those values and integrate them into your segmentation process.
❌The cons: at present time, Attributes doesn't support the creation of new types of insights by our clients. While you get the insights at a much lower cost than doing everything in-house or working with categorized data, you will have to work with our team to create a customized set of insights that fits your segmentation goals.
Get started with Flinks today
