Developing profitable partnerships for ecommerce SwiftERM

David Swift
4 min readFeb 17, 2021


Their search provider to implementing product recommendations, became a bundling and merchandising solution as well.

Ensure Your Partners Work Together

If you have dozens of providers, you’ll never have time to manage each one and know them in-depth. But if you have a handful of providers that talk to each other and work together, then you can really spend time working and understanding how they properly fit into your business, and how you can optimize each one.

These integrations could save you a ton of time, but more importantly provides an even better customer experience.

For example, a lot of our customers have reached out to Yotpo, and asked if they can add a new sort order, based on review ratings. In parallel, some people reached out to us and asked if we integrated with Yotpo, because they want to have a “star rating” filter, just like the one on Amazon. many solutions connect with Yotpo on joint solution:

Findify integrates directly with Yotpo, which allows us to pull all review and rating data. We use the information they provide (e.g., review ratings and velocity) to display smarter product rankings. Essentially products receiving high reviews, quickly, are ranked higher — because it essentially means customers love them, and they are trending. In addition, customers can filter products based on ratings, and sort them to find what they are looking for faster, either through search, collections or recommendations.

Without this integration, reviews would only be available on the product page, filtering by rating wouldn’t be available, and ranking wouldn’t be affected by reviews. Can you imagine trying to code that all on your own?

Augment Machine Learning with Human Touches

“After I install it, I don’t ever want to touch it. I just want it to be useful by itself.”

Sound familiar?

Like a lot of merchants, you probably want your extensions to be simple, so you can set it and forget it. That’s understandable, considering that you’ve got to manage many moving pieces with an ecommerce store.

Most machine learning tools, including ours, work well out of the box … up to a point.

That point is when your machinery needs additional information from humans.

For example, a machine can’t be expected to know that a new version of a product will be released in two months. As such, the algorithm won’t know to promote the product and sell it when it launches.

But when someone on your team provides very simple tools to tell the system about these upcoming products, it will then know to promote them to interested customers — resulting in better sales. For example …

  • If you have a product in high stock and you need to push inventory, you can arrange them to appear earlier
  • Or, you have a high priced product that could anchor pricing and increase your AOV. (We have quite a few merchants who apply this tactic, based on behavioral economics — a great topic to read up on in general)
  • There might be products that even though they don’t sell well, but act as a “brand maker” or “image creator” for the store; merchandisers like to place these products on top to communicate a certain message
  • And lastly, sometimes you have products that are very similar, but one of them has great margins; you can push those high margin products to the top, to optimize your bottom line

A huge benefit to machine learning is its ability to augment humans. That relationship is symbiotic, and in order for the machine to learn, it needs your team’s input, maintenance, and support.

So, you’ll need to invest a couple of hours a week, or a few hours a month, to properly calibrate the tools that you have at your disposal. This will really help maximize revenue and set up your perfect system.

Prioritize Ecommerce Data Hygiene

Garbage in, garbage out …

This adage rings true for anyone familiar with data analysis. If you don’t collect your data well, clean it, and ensure its integrity, it won’t reflect reality. All insights you draw from it will be invalid.

If you want to make the most of your machine learning software (and frankly any data-related application), data hygiene is essential. For example, ecommerce search recommendations can only be as good as your data. If it is corrupt or disorganized, you won’t be able to draw insights, and any changes you make won’t reflect a customers’ real-world experiences.

The need for data integrity applies to the rest of your store’s operations as well. It may seem obvious, but it’s still crucial. Do the common things uncommonly well:

  • Keep your inventory and product information up to date
  • Make sure that your product descriptions and details are current and accurate

Once you can trust your data, you can leverage the more advanced capabilities in your solutions. With Findify, for example, you can create hierarchical category nesting, or categories, based on any type of variable that you choose: skin type, material, length, delivery time, etc.

Originally published at on February 17, 2021.



David Swift

Founder & CEO of SwiftERM the personalization SaaS. Microsoft partners.