At its most basic this could be implemented as a list of last n items that were explored by me on the last p visits. (n should ideally be a small number and p could be a reasonable historical period that would make sense for the customer. Going back too far back might not be useful as the possibility of the need for that item having expired would be higher.) This serves just as a simple reminder and a kind of a bookmark for easy access to the products.
Once past this hurdle there is some really significant value that can be created by making this memory intelligent. There are many possibilities here and I will only be scratching the surface with the ideas that follow. If I had more time, and a mandate to do this on an official basis, I could dream up more ways. So let’s see what can be done.
In the simple solution I suggested going back only a reasonable time back in history to get the customer’s products of interest. But that is a bad generalization. Especially for high value purchases. When I bought my camera I researched for almost 2 months to get all necessary information about camera bodies and additional lenses before deciding on what to purchase. So in case of showing recently viewed products this is something that needs to be kept in mind. So there is a trade-off between what category of product to show in this list between time since first view, last viewed date and relative value of the product.
I know I am being fairly subjective here, but then this is really a brain dump. So, there you go.
Further signals that can be used to build this list could include “changes” to the product. Of course the product itself doesn’t change but the information available about the product can easily change and be tracked. May be additional reviews have been posted, a friend might have bought it (social integration in e-commerce is another idea ripe in my mind and I will write about it in a few weeks), it may have been the darling of public social conversations recently or the it might have won an award. There can be scores of such signals that can be used to optimize the way this information is presented to the customer.
The challenge is obviously to identify the correct signals for this task. Furthermore the some signals are likely to work better with specific classes of products while there may not be any such valuable signal for some classes of products (e.g. a pendrive). The choices made at this signal identification stage will determine the complexity of implementation. Also to be ensured is that there is always a suitable fallback available in case the customer has not viewed any products recently or has only browsed those classes of products for which no signal is available.
Leave a comment to let me know what you think of the idea. E-commerce has so much to offer beyond just the recommendations being served up right now. It could really do with some intelligent memory for the time being. A socially intelligent memory will be even better.