Author: Randall Malcolm
In brick and mortar stores, in line for the cash register, you often find rows of clever products you didn’t know you wanted. And what else do you have to do but look at them and wait?
Add this to the temperature in the stores, in-store branding, the sales people, where and how products are displayed on the shelves, and you have a matrix of strategy and research.
The million-dollar question: Why aren’t e-commerce sites built with that same data-driven, researched strategy?
The answer: Until now, there hasn’t been an easy way to turn ecomm data into data-driven, real-time user experience (UX) changes.
Why? Machine learning and a shift in focus: e-commerce was already on the rise, and COVID-19 dramatically accelerated the change.
So, the question: how do we direct visitors to checkout and help them add more to their carts along the way?
Dynamic user flows.
As we all know, visitors click/tap different elements to move from one page to the next.
These elements have four dimensions:
- content (the copy or the image),
- form (button, image, hyperlink, banner, etc.),
- location (above the fold, above the price, in the top left, etc.) and
- styling (colors, size, weight, spacing, alignment, etc.)
The trick is aligning these four elements with the users’ preferences or, in effect, making the path to conversion as easy and pleasurable as possible.
To begin, the landing pages should match the methods of acquisition–consistency of brand and content. For example, if the landing page features bright dresses and the digital ad features a drab pair of shoes, the bounce rate is higher.
Second, you need to track the acquisition source. Most sites use Google Analytics to do this and it’s a great tool. With the proper labeling, you can deduce why and how visitors arrive at your site. You understand their motivations.
And this, motivation, is the first clue (data point) as to what visitors want and/or need to see to make it from the landing page to the product pages.
With only this data point alone, you can begin to optimize category pages and related products (content and location).
When you begin to add in other data, such as the demographics of the visitors, on-site behaviors, and time of day, you can begin to change the form and styling of the elements.
For example, You may learn that men, between 35 and 45, after 10pm, in the midwest, prefer more saturated colors, free shipping, and bigger font sizes. All this keeps you on brand, but enhances certain elements.
Then, if all goes well, you have the ability to followup. You can track cookies and/or have user-specific UTMs in your retention campaigns.
This allows you to create personalized funnels.
You display the elements that have best funneled that visitor in the past.
In essence it’s a conversation just like one you’d have with a sales person. The visitor explains what they want by clicking on a link. The page adjusts to show different options. And each additional click and scroll, indicates what they want to see and how.
The site adjusts.
Well, really, the machine learning models have run all of the different variations for each scenario and delivers the that which is proven to have the highest probability of conversion.
To accomplish this, there are two machine learning models needed.
The first machine learning model has to do with content exclusively and is, by far, the hardest to set up. All the content has to be tagged standardized, and the delivery of that content has to be integrated with the site’s platform. It’s a big task that, but, if done well, pays off in spades.
The second machine learning model deals with the form, location, and styling of the elements. This set up takes about 30 minutes and some creative intuition. Once you label all of the elements you’re willing to change (10 minutes), you have to determine what the right adjustments might be (20 minutes). Then, you get to see which adjustments are working and which aren’t and you refine and refine and refine. This process not only constantly refines the funnels presented, but also allows you to adjust your acquisition and retention efforts according: a true value.
Obviously, both of these machine learning models can work in harmony and really make the site a personalized journey–the future of e-commerce.