A humble bestseller product recommendation is an easy win. For a little effort and low data dependencies, it gives you strong revenue uplift.
Bestsellers’ high performance makes sense.
Your most popular products are popular for good reason. They exist at key intersections of value, features, desirability, quality, and trend.
Amplifying any of those signals to your audience is always going to make you money. It’s easy to understand why it’s one of the most popular types of product recommendations.
Job done, right? Why do you need AI?
After you’ve marketed your bestsellers for a time, you may hit a couple of challenges.
First, bestsellers favor your more established products. Unless new products are immediately successful, a recommendation can become self-reinforcing. It can be hard to get different products to breakthrough.
Second, you may notice that revenue uplift starts tracking new customer growth. At this point, the recommendation is saturating. Whilst it’s smart to hit your new customers with your best stuff, it’s now underperforming for existing customers.
Both problems warrant their own detailed discussion. For now, let’s look at a more nuanced strategy and how AI can help find new revenue in your existing customer base.
Unlocking inaccessible revenue
AI-powered product recommendations will identify new and unique customer/product relationships.
Finding these relationships at scale is where machine learning comes in. It analyses your products, orders, and web behavior data, so you don’t have to. It roots around the dark corners of your data to match products to customers.
Doing this analysis manually, even if you knew what to look for, would be an impossible task. Machine learning does it continually for you. Each time it trains on new data, it learns and gets more accurate.
My argument for this kind of big data approach to marketing is simple: don’t assume your established customer personas are the only truth. Until you use machine learning, you don’t know what you don’t know.
Winning with blended recommendation strategies
Bestsellers may always be your top performing recommendation. Talking to retailers, I’ve heard cases where a small set of products accounts for over 60% of sales. AI is unlikely to outperform against such massive numbers. (Unless you’re Amazon and have an enormous and diverse catalog!)
These retailers are aware of the risks of saturation. Not marketing effectively to their wider customer bases is a long-term challenge. Historically, there are easier battles to win that deliver nice returns.
Fortunately, technology is catching up to support retailers.
We’ve built Engagement Cloud product recommendations to support a blended strategy. You can combine so-called heuristics (like bestsellers) with hyper-personalized recommendations using AI.
The theme behind this strategy starts with covering known areas with broad sets of rules. Create non-AI product recommendations to match your known customer cohorts. You might focus on product categories, price points, seasonality, trends, or any other rules you like.
Once you have those, it’s time to infuse your campaigns with AI recommendations.
Here’s how to use different classes of recommendation:
- Set up multiple category-targeted best sellers for some big hitting recommendations;
- Find tomorrow’s best sellers with the most viewed recommendation type;
- Mix things up with the hybrid trending recommendation type (it blends best sellers and most viewed);
- Match your niche customers to their perfect products with AI-powered lookalikes;
- Use best next’s AI to let shoppers help other shoppers find products they didn’t even know they wanted.
With this approach, you’re casting the widest possible net to drive more sales. You’re building automated marketing around cohorts you know. Meantime, AI is finding new customer/product relationships you didn’t know you had.