How to choose the right tool for ecom personalization

There are a lot of tools for personalization out there, for search, notifications and many other features. So many service providers that in fact it happens quite often that the people we talk to are as much in need of such tools as in consulting to select the best ones for them.

So in that situation, how do you proceed? How do you select the best service when everyone promises indecipherable algorithms and unbelievable revenue effects? Let me share my personal 6-step advice for personalization success for your online business.

1. The elephant in the room: Do you really need personalization? We all tend to say yes, and with good reason, but let’s take a closer look: Today, everyone wants to do personalization, even though it has to be said that it actually is not live in too many stores today. I personally see it like in software development: Don’t optimize until you really have to. Many small stores probably should focus on products, building a brand and growing their revenue 10x, not improving margin by 20%. Later in the life of a successful store though, personalization is inevitable because of it’s vastly improved UX and better effectiveness regarding revenue per visit.
   One caveat however: I believe however that this “optimize later” logic will cease to be valid mid-term, once users are accustomed to a great personalized UX regardless of store size.
2. What do you want to achieve? Of course you know what those crucial KPIs for your store or your team are, but how do they relate to personalization? I’m talking about direct vs. indirect contributions to your goals: Assuming you are already able to track your key metrics in real time, then tools like Optimizely or Mouseflow are indirect contributors in the sense that their value proposition is to be a lever to raise your directly relevant KPIs. There has been a fair amount of criticism (see also this discussion of Hacker News) lately about indirect tools, and in our opinion that criticism is justified. They help you for instance to do the A/B testing job more effectively, but the real question is: Should you be doing A/B testing to achieve your goals? The answer is often unclear. That’s why at Free Machines, from day 1 we set our sights on directly contributing to your goals, whichever they may be: Improve add-to-basket rate? Great. Actually, it’s gross profit? Awesome. The average NPS? Fantastic. The number of frogs on a starry night in Barcelona? Absolutely. If you can measure it, we can train our algorithm towards that goal.
3. Make or buy? Buy! :-) I tend to have a strong opinion here. Of course, I’m not completely unbiased. Statistical optimization is work that is best left to specialists that can gather experience from working on many stores. If you’re not Amazon or Facebook then you probably don’t want to build up the teams over the time required to do so. Not because you can’t, but because it’s probably not the best use of your time.
   Small anecdote: An important part of the Free Machines office is to have a big inflatable circus horse. Why? Because we’re trying to help our customers to be “circus director” and focus on the strategic decisions which elements to personalize towards which goals, rather than do the testing and segmentation themselves. In our humble opinion, that’s best left to statistical algorithms.
4. Select a subset of tools to get started! I’m a big fan of agile/lean business processes. In a market like AI, where the direct contribution to your goals (see above) is not always proven beyond case studies before you at least do a PoC with a service provider, what does lean suggest? I would argue that it’s about the relationship you have with a service provider and if you are focussed on short-term gains or long-term growth. The best choice regarding relationship is a service provider that is a direct contact can be held responsible should the promised results fail to materialize. Regarding short-term, it’s probably best to go with market leaders if you need reasonable results quickly, and go with specialized cutting-edge providers if you’re focussed on the best possible long-term result.
5. Is this thing ROI positive? Ask about integration (in detail) and PoC duration. The only way that the whole project is guaranteed to be a ROI positive result is if all three of the following conditions are met: The service directly contributes towards your ROI goals, if it’s performance is measurable in parallel to what you do today (typically on a fraction of your traffic), and if it is cost-effective after service fee and your manual/integration effort. At Free Machines, we ensure all three with a pricing model that is tied to the actual gross profit achieved with our solution. We all know how painful it can be to spend money for raw revenue rather than gross profit — I’m looking at you, SEA — so you shouldn’t accept anything less.
6. Give it a go! Once you have a service provider and have settled the commercials, make sure you are in charge of controlling roll-out, for example via a setting in your shop add-on or your tag manager. If anything goes wrong down the road, this will help you figure out much quicker if an add-on is to blame. Anything else, like tracking your success, should at that point already be set up, and all the hard work should be left to the service provider. Congratulations! Your users will thank you for the great tailored user experience.

I hope that this guide was useful. Do you have comments, or additional hints on that? Do let me know via email or in the responses below.

Thanks for reading!