/ Personalization

Should I buy or build a recommendation engine?

You've got a growing app or ecommerce store. You want to invest in a recommendation engine to use personalization to increase engagement and sales conversions. Should you buy or build?

We may be biased, but we think it is almost always better to "buy" by working with a recommendation engine provider.

Here are 5 reasons why:

1. Probably not your core business

If your core business is selling products, services, articles or other consumer goods, your engineering and data science resources are probably better spent elsewhere.

By working with the right service provider, you get a turn-key solution that works, leaving you time to focus your resources on more important features.

Caveat: If you are a business with large economies of scale and where recommendations can be your secret sauce, it might we worthwhile to build your own (e.g. Amazon, Netflix).

2. Performance

There may be open source systems out there that provide algorithms out-of-the-box. Some examples include Google Big Data Suite, Apache Spark and Prediction IO.

But a recommendation engine is not just about algorithms, collaborative filtering and content-based recommendations. It is about the business impact that personalization can bring to the customer experience.

For there to be business impact, there needs to be data integration, data cleaning, training, tracking results and numerous iterations. There needs to be enhancements to introduce elements of surprise, scarcity, social proof and urgency.

The algorithm is less than 5% of the effort in this process.

When you buy a solution, performance is a given. Otherwise, you can go to the next alternative.

3. Flexibility

For recommendations to perform in the business sense, they must be flexible enough to allow marketers to promote the products they want. They also need to handle a host of edge cases like unavailable products, missing product images and aesthetics.

Recommendation engine providers address these problems through working with many customers and are able to provide a production-ready solution.

4. Multiple channels

Having a recommendation engine does not mean you can profit from it. You have to be able to use the results of the recommendations across channels. This includes your website, emails, apps, Facebook retargeting ads etc.

Recommendation engine providers figure out the nuances of integrating and working with each channel so you can sit back and reap the benefits of better engagement and sales conversions.

5. Lower cost

Given the above-mentioned points, it usually costs a lot less to work with a service provider than building a recommender from scratch.

Related: Top recommendation engines in Asia

Justin Yek

Justin Yek

Partner & cofounder at Altitude Labs, creator of Metisa, former investment banker, public speaker, hobbyist musician

Read More