How to troubleshoot incorrect product recommendations

Recommended products not what you were expecting? Recommender seemingly not finding certain products? Recommendations not making sense...... No problem, there are a couple of simple ways you can double check everything is all configured correctly. This guide is here to help you investigate, identify and fix these issues to ensure your quiz is recommending the perfect products to each and every customer that uses it.

Tags

The first thing to check would be the tags. The Help Me Choose app uses the Shopify product tags to find and recommend products to your customers based on their answers to the quiz. Therefore, if there is an issue with the tags it will cause issues with the recommendations. The first thing to check is the spelling of the tags that you have added in to the quiz builder. The tags that are added to the quiz builder must be spelt EXACTLY the same as they are in your Shopify store. Start by just checking them in the quiz builder for any spelling mistakes or typos, if there is nothing obvious then you might want to cross reference the tags in the quiz builder with the tags in your Shopify store (you can find out how to view/add/edit Shopify product tags here). Check for things like spaces, hyphens etc. The tags are not case sensitive but must match the tags in your Shopify store in every other way.

Weightings

The next thing to investigate would be the weighting strategy you have used in the quiz builder. The Help Me Choose app uses a simple addition formula to recommend products based on the weighting you have applied to an answer and it's associated tag or tags. In order for this to work, first the tags must be assigned to products within your Shopify store, and then the weighting structure you use in the quiz builder must be logical. There are two main ways you can structure your weightings.

Equal Weightings

The first way, which is the most simple way, is to assign equal weightings to every answer (and subsequently the tags associated with the answers). In this scenario, the importance of all questions will be equal so the products that will be recommended at the end of the quiz will simply be the product in your store that has the highest score against its tags from the answers given

For example a product that has 5 tags assigned to it that meet the customers answers, will score higher than a product with just 4 tags assigned to it. The product with 5 tags will be recommended to the customer as it has scored higher in the weightings.

The Help Me Choose app uses a 1-100 scale for weightings and when using an equal weighting structure you can choose any number you like, as long as it is the same for every weighting.

If you have used an equal weighting structure and the app is producing random recommendations then go through the weightings in the quiz builder and just double check you have added the same number for every single weighting.

Incremental Weightings

The second method for weighting structure is incremental weightings. This is the best method to use if the answers to the questions you're asking have a varying level of importance when it comes to the products you want to recommend. For example, imagine you sell handbags and you want to identify what style of handbag the customer prefers and then what colours they like. The style of handbag is more important to the recommendations as you only want to show the customer their chosen style even if the colours might vary slightly. In this case a higher weighting would be given to answers/tags relating to the style question, 100 for example and then a lower weighting of 70 for example would be given to the colour answers/tags. As the Help me Choose app uses a simple addition method, products that score highest (by having the most relevant tags) at the end of the quiz are recommended.

If you have used an incremental weighting structure and you're getting strange recommendations, it might be worth double checking your structure and making sure it makes sense mathematically. You can experiment by changing the weightings and re-checking your recommendations. One thing to check is that if you have given equal weightings to answers/tags for multiple questions, and/or you have used checkbox question types, it is possible that certain products that you might not be expecting have scored higher than the ones you were expecting based on the answers given. So just double check what answers are being given and amend your weightings if required.

Still looking for answers? Get in touch with the team on [email protected]