3 Signs Your Business is Ready for a Recommendation Engine
Business + ManagementMachine LearningModelingrecommendation enginerecommender systemposted by Elizabeth Wallace, ODSC November 1, 2019 Elizabeth Wallace, ODSC
Data is in high demand, not just on the business side but for customer-facing solutions as well. When your business can fully integrate data into your customer journey and day to day experience, you become a more valuable tool to that customer. There’s a lot of noise out there; you need ways to personalize the experience so your products and services stand out. One way to drive that personalization is through the use of a recommendation engine. Amazon was just an online bookseller at one point long ago, but early adoption of a recommendation engine helped catapult Amazon into the behemoth it is today. Here’s how to know if it’s finally time to make the same leap yourself.
[Related Article: ML and Behavioral Economics for Personalized Choice Architecture]
You Have the Right Data
Recommendation engines run on data. If your business doesn’t have customer profiles established and hasn’t labeled attributes of products, your recommendation engine isn’t going to provide the kind of service your customers need to remember you.
If you’ve evolved past simple pie charts and Excel spreadsheets, you could be a good candidate for implementing a recommendation system. The first component of your data is product attributes. Your product inventory must be tagged thoroughly for the right attributes, and you should have a method in place for implementing your tagging strategies for new products.
Your customer histories and their interactions with the site is the second piece of your data strategy. You must have logical methods in place for storing your customer data safely and securely. Once these two pieces are in place – safely stored customer data and up to date product data – you could be ready to build your first engine.
You’ve Considered The Costs
Storing and securing your data isn’t the only cost to consider. You’ll also need the talent to see it through. If you’re going for an internal, proprietary system, you’re going to need a team of around six people:
- two data scientists who can maintain the quality and integrity of the data and build the algorithms
- two developers to process and develop the code
- two engineers to maintain the kind of infrastructure you’ll need to keep data up-to-date, secure, and running.
If you don’t decide to go that route – it could easily cost you in the millions of dollars – you can use external solutions or done for you engines. Even here, the monthly cost of implementation and support could still end up not cheap, depending on your company size. However, if you’ve considered both costs and understand the long term revenue you’re leaving on the table by not having your engine in place, it could be time.
You’re Already Scaling Operations
Recommender systems reduce the time customers spend looking for items and use real-time decision making to place items the customer is more likely to buy right in front of them. If you’re already scaling operations, whether product or customer reach, a recommendation engine could support that scale.
As your inventory grows, it’s harder for customers to come across all the pieces they might like. Even if your pieces are tagged well, the chances of a customer looking through ten pages of options are low. That recommender system takes its history and builds targeted suggestions, which could increase conversions and overall cart size.
Choosing a Recommendation Engine
If your business is large enough and you’ve got the capital, building an in-house system could help you get a recommendation engine that’s built precisely for your unique specifications. You don’t have to worry about moving outside your organization for maintenance and all your data remains securely in-house.
That option can come with a high cost, however. The cost of your data science and development team alone could run you nearly half a million dollars, plus the cost of storing and maintaining the security of the data itself. In return, you get a boost in sales and social credit, but if the startup is too high, there is another option.
Recommendation Engines as a service could offer a step in the direction you need. Outside organizations build your system and help you maintain it for security and quality assurance. You’ll still pay quite a bit in monthly maintenance and will have to outsource your support desk to the organization. In return, you’ll get a solution that’s faster to implement and less stressful on your team.
[Related Article: Designing Better Recommendation Systems with Machine Learning]
Regardless of what you choose, a well-done Recommendation Engine could help you scale up your sales and build a more dedicated customer base through personalization. It also kicks back data insights based on customer behavior that can help you make higher-order decisions for the direction of your company. If it’s time, it’s time. The return could be astounding.