Saw an interesting article on e-commerce at E-commerce Best Practices Make Perfect by Marshall Lager. There were some great general purpose comments on e-commerce but a couple of things caught my eye. Firstly there were some great comments by Jackson Wilson, CTO of Proficient Systems who said
"Analytics isn't the same thing as reports, though the terms are often used interchangeably. Analytics is methodology and science. Reports are just output. One of the worst things we see is the assumption that Tool X will provide Y performance. It's not the tool that improves performance; it's the integration with your business."
Jackson is making several great points in this comment. Not only does the phrase "analytics" get used to mean reports but even "predictive analytics" gets used imprecisely. I blogged on the different types of predictive analytics before, but both kinds can be useful in this context. Firstly you can derive rules from reports, especially from predictive reports and those that help you segment customers. These rules can then be executed in a business rules engine to help improve the e-commerce experience for your customer or to help you better target them. Some rules products, like Blaze Advisor, allow you to import these kinds of analytically derived rules directly. The second kind of predictive analytics can actually be embedded into rules. Thus if I can build a model that predicts retention risk, say, for a customer then I can also build rules that use this information e.g. If Customer's RetentionRisk exceeds 75 then Customer's Promotions.add("Retention Offer"). This allows me to use even quite complex models in rules without having to make the rule writer understand how the analytic model works. In both cases, as Jackson notes, I must be able to constantly change as new data comes available. This is why the last letter of EDM stands for Management - you must be able to automate and improve decisions.
The article also had a great list of dos and don'ts, some of which seemed particularly decision-centric to me:
-
DO use customer data to give each individual a personalized shopping experience.
Note that having customer data, even having a 360 degree view, will not help unless you use it to improve their experience. In the rush to collect and manage and integrate customer data many organizations forget to think about how to use it. Evolving both the data you have and your use of it in parallel will work better than just trying to improve the data without a vision of potential use. -
DO be prepared to change your business rules and processes to meet emerging trends.
Build for change time - don't hard-code any of this stuff as it changes all the time. Your data tells you new things, you get smarter about your data, your customers change and your competitors change. Your system (for e-commerce or anything else) had better be able to change at the same speed. This is what business agility means. -
DO integrate online and offline channels, taking advantage of the best features of each.
One of the key tenets of EDM is consistency, especially across channels. This does not contradict this best practice as it means being consistent unless you mean not to be. So take advantage of each channel but if a customer would expect consistency across channels, make sure you deliver it. -
DON'T rely on new technology as a panacea. Tools don't improve performance by themselves; wise use of them does.
Could not agree more. Nothing to add here just thought this was such a good point it deserved to be repeated -
DON'T rely on automation and self-service exclusively. Sometimes, customers are people who need people.
True. But don't underestimate how much some people want to do themselves. If someone wants to self-serve, make sure they can. Don't make them call your 800 number just because you can't be bothered to automate the decision or process they need. Far too many e-commerce sites fail to automate key decisions in a way that lets the website and the call center respond quickly to customers. Anything that automatically gets referred to a supervisor 100% of the time should be examined with care. Likely 80-95% of these could be automated using rules and analytics so that the website and the most junior CSR could decide.
