Artificial Intelligence is becoming ubiquitous within e-commerce. Now is the time to make your site intelligent.
Contact us to understand how AI could work for you
Please share:
Share on facebook
Share on twitter
Share on linkedin
The proportion of sales that start with an online search is growing, the proportion of sales that are digital is growing, the expectations of the customers are growing.
There is nothing new there, those trends have been evident since, oh, the birth of the internet. What’s different is that these trends are accelerating and in particular the third point. Customer expectations are growing.
Here are 5 expectations or customer needs:
- Relevant content when I land on your site
- Site search that works and can recommend products as I type
- Recommendations that help me discover but also that match what I want
- Product information that is helpful and up to date
- Pricing and availability information that is transparent and competitive
There are many others, but these five help us to zero in on improvements that artificial intelligence can help with immediately.
Just to cover off some operational points, to get this work for your site is reasonably non-disruptive and trivial for you as a store. There is some code to put on your pages to begin with, but that is all. The code will help us to understand user journeys, granular engagements and content/product gaps. We would also do some work with your product file separately.
We can then start working with the data to surface the insights as to how your customers are using the store, what it is that they like and what puts them off, but even more importantly, where are the opportunities. If you like, this is the audit phase where we can really start to see how to prioritise the optimisations.
We have to start by running the initial models on the data to help us see the ‘unseen’, to see where we can use AI to better understand from a users point of view we are a hit or a miss.
1. Relevant Content when I land on your site
When a user lands on your site they are immediately looking for your hooks. Is this the right thing, does this site have what I am looking for. Over 50% of users will be
AI can help make decisions on what kinds of images work best. Not only what kind of image but also how they are captioned.
The same can be applied to text. It’s already common place to be able to customise the landing page to the ppc campaign, but what if you can start to change text based on a combination of what you already know about that customer as well as how they found you? If you don’t know that customer already, it is possible that you can make content optimisations based on similar users.
These patterns and the propensity of the recommendations that have previously been completed will mean that you can automate your optimisations. Most users search and research techniques are not particularly well-honed, AI can help to ensure we surface what they are looking for as soon as they land. Current AI makes recommedations and suggestions. But what if that AI could curate individual web experiences based on that user? So the images, the text, headlines, CTAs are all individual to what works best for that individual.
One good way to visualise this is to think of responsive design and how that works for different devices. Multiply that now by user behaviour, propensity and predictive modelling at an individual level and you start to get the picture.
The AI can create adaptive user interfaces that work for the individual. This also means that the individual has more control over the outputs. Giving them more time and bandwidth to engage with your content. It also starts to reduce accessibility issues that you may have.
This improves the experience immediately for the customer.
2. Site search that works and recommends products as I type
Search is vital to conversion. Search is much more useful to most Having a search function that uses AI to be able to predict (from the second character in some cases) what the customer is looking for puts you miles ahead.
Have you thought for example about Federated Search? This is when one search term can mean a number of things and the answer can lie within different categories or a combination of attributes.
This means you can configure results in a way that showcases these different destinations to help users. This can also be applied to query suggestions when the search term may be a little loose. The user will always know what they mean and the context for their search. This is a brilliant use case for AI, because you can learn so much from the different ways that users search and apply that to your content.
What’s really exciting is that different search terms and result formats will change for different users and you present that. Images and the different way that images can be presented could make search that more valuable, especially when users can use images and text to search your site.
But, say they uploaded an image of a yellow round neck sweatshirt with text and they added text that said French writing. We would be able to ‘read’ that image and translate that into a text search query to find closely matched items and present them back as images within the sarech results box. We’d also know if they had any French writing, or we could suggest other items if we did. What this means is that we can start to deduce semantic requirements from search and improve discoverability in that way.
These techniques can be used from DIY to building materials, to bird watching and recipes.
Some e-commerce sites report a doubling of site sales through intelligent search.
3. Recommendations that matter
Often recommendations are based on what others have browsed or bought on the website. Rule one, try to make that data as holistic as possible and use POS/ERP data too. Secondly, you can use Purpose data too. Purpose data is the reason why those customers made that purchase or that bundle buy. This is a crucial use of AI to be able to understand why your buyer types are buying, when and what they buy, but also for how much.
Being able to understand patterns and behaviours starts to answer the why question. If you are also holistic in your use of data, being able to also match unstructured data with customer data and clickstream behaviour, you are more likely to be able to understand the purpose of a purchase. Don’t be shy with your data sources, it may seem too difficult sometimes to release the data from their natural silos, but it is totally worth it in terms of payback.
We use dynamic re-ranking to make these suggestions optimal. Websites are dynamic and outside influences or campaigns or competitor activity can start to change how you would want to surface recommendations. By using AI they improve over time. As the use cases become more familiar your opportunities grow.
Recommendations also need a strategy and this can be controlled to be able to weight different business outcomes. Do we want to push own-brand, do we want to serve recs where we know we have competitive pricing, do we want the user to be able to bundle the purchase with other products to hit a discount tier?
AI can make these recommendations as dynamic as you want them to be, which is a long way from being served last years Birthday present for Aunt Hilda.
4. Product Information that is helpful and up to date
It’s always a bit gutting when you think you have found the right product but the product description isn’t quite what you thought it would be. This can happen if you are buying something technical or something fun or something in between. As we know, different buyers have different purposes, and some ‘customers’ want to use your product descriptions to help them make purchases elsewhere.
Product information and how you manage it is key to your overall SEO strategy as well your CRO. Having full product information and filling in te gaps, having the right descriptions, descriptions that are helpful in converting but also that rank well is a tricky business. But, the good news is, this is where AI can really help. Being able to use AI to make sure product descriptions are not only useful but helpful, that they make sense for search engines as well as searching humans and have the ability to be customised to help convert the sale, means you now have the means to optimise your catalogue.
AI can also help you classify your products and bundle products together in a way that you may not see yourself, but your users do. Users are often working around your navigation to find their solutions. That is the nub for us. How can we listed to these signals to help the customer find their solutions easier and how can we then help others in a similar way?
How you can then automate your processes to keep these files up to date is key. It’s not a set and forget strategy, it’s a set, automate, improve strategy.
5. Pricing and availability information that is transparent and competitive
Of course all this is fairly pointless if you aren’t pricing correctly or you are out of stock. Artificial Intelligence can help with both of those things. Competitiveness is key, there is a whole industry based on making sure prices are competitive and how you respond to changes in competitor strategies is important. We all know that supermarket executives spend a lot of time discussing the price of milk.
You may be in an environment where local pricing is important. What happens in different geographical areas needs to be recognised, can you do that online? How can you make your pricing dynamic based on demand or availability, do you want to do that? Scarcity drives prices (here’s looking at you PS5), but for your brand proposition you may want to compete on availability, while taking into account competitor pricing. AI allows you to constantly reevaluate these decisions to optimise them on a SKU by SKU basis.
Across a catalogue of 000s, with variations, this can be incredibly painful if you can’t automate it. it has been known for commercial managers to sit at their desks changing prices by fractions on a daily basis. This can be automated.
So pricing and availability can be optimised in real-time. With clear call to actions and the ability to short cut the check out process, this means that customers are experiencing less and less friction. It’s also important that you can be upfront about shipping and delivery times. Jut by being able to say “this will be with you by Friday, why not order this XYZ too?” you will grow AOV.
Takeaways
Artifical Intelligence in e-commerce means that you can do more with less. But, it means more than that. It means you can be more accurate, have a closer relationship with your customers, bring in new service levels and be less transactional. We haven’t touched on ‘bots’ in this piece because they need their own article. But you can see how control of the experience via AI allows you to then be more conversational with your customers through bots.
The future of e-commerce is about individualisation, how to adapt the digital ecosystem to make your relationship more human and more interesting. The responsibility for discoverability then moves off the user and their search and navigation skills, to one of the engine to provide these experiences.
