Margin Geek: How It Works

Are you looking for increased profit margins on Amazon? Or do you simply want to make your business more efficient? If the answer to either of these questions is yes, Margin Geek’s AI-driven product recommendation system allows you to increase your margins while saving time.

In short, we use AI to deeply scan product opportunities from your source lists. We are different because we can find product opportunities that other tools can't find, whether it's text-based matches, or a listing with no stated volume, but a ton of reviews generated in the last 3 months. In this article we break down how it works.

Example Output

To make it real, we've shared example output here and simplified output below (redacted for confidentiality):

Your output shows you a list of the best product opportunities for you, along with margin, return on investment, and predicted volume. All you have to do is review the top candidates.

How it works

Our deep scan match is based on allowing an AI to take a risk on matches it isn't sure about. We pair that with a human-in-the-loop quality check. Without this human-in-the-loop service, we would be forced to make less match predictions, and you would not get as deep a scan.

Step 1: Upload a supplier product sheet

Your business is built on finding high quality product sources, and you should be spending your time getting the best sources, not analyzing the lists themselves. Once you've identified your source, upload the source list in csv format.

Step 2a: AI Scanning Magic

We then apply our AI magic to scan your source products. We use many techniques to get the deepest scan

• Matching on product text
• Matching on any product code formats (UPC, EAN, GTIN, ISBN, etc.)
• Searching multiple queries for each product
• Accessing data from multiple sources
• Applying the latest research to fine-tune our matching models and get the best search (e.g., clean data practices, large language model fine-tuning, prompt engineering, pairwise reranking, the list goes on)

You are getting access to a scanning process we developed for profit share clients, where we were incentivized to never miss any opportunities. Using AI, we often find twice the number of matches compared to if we just searched using product code matches alone. (See our AI vs. Human trials)

Step 2b: Profitability and volume

All our recommendations include profit and volume. To get profitability estimates, we take the published fee structure from the ecommerce marketplace in question (e.g., Amazon), and we factor in all the variable costs for the product: referral fees, shipping to Amazon, FBA fees, storage fees, peak season fees, inbound shipping, low inventory fees, etc. We also incorporate your shipping fees and your fulfilled by merchant fees in our "Seller Costs" section. On pricing, we use 30 day historical pricing.

Beyond that, our algorithms intelligently perform base unit comparisons. If your supplier is selling a case, we’ll convert the cost of a case to the cost of an individual retail package.

Step 2c: Volume Estimates

If you don’t have robust volume estimates, you’ll purchase the wrong amount of product. If you don't have the data to estimate volume, you'll miss lucrative best-seller opportunities. This happens regularly in product categories like electronics, where other tools struggle to estimate volume because best seller rank (BSR) is not publicized for the major category.

To make sure you don’t miss any opportunities, we estimate volume three ways: a BSR-to-volume conversion (primary method), a backup reviews-based estimate, and using the X+ products sold this month tag, when available.  

Comparing our estimates to other tools, we tend to be the most conservative. For example, we adjust estimates for the number of sellers as well as the number of variants, which some other tools miss out on.

For one client, we found that our backup volume estimates allowed us to estimate volume for entire catalogs of products that were missing BSR information. Other tools had no volume estimates available for more than two-thirds of product listings in the catalog.

Step 3: Quality Checks

Our offshore virtual assistant team is available to perform quick manual checks on any scan to save you time before you go in to check the product list.  

Go / No-Go

Ultimately, the output is a go/no-go system recommendation. You get a priority list of the most profitable products, with all the data you need to make buying and quantity decisions.. 

You can use a conservative purchasing technique based on historical volume, or you can hire us to optimize your inventory automatically. We have purchased hundreds of thousands of dollars of product using optimization software that balances the opportunity cost of missing a high-margin sale to the risk of price wars, stale inventory, and write-offs. Using quality control and data management, we can put your purchasing on auto. Please contact us for details.

That’s a wrap!

In summary, our goal is to manage your product picking decisions, so you can focus on running the business. We hope we covered any questions you had about our service. Please contact us if you have any questions or any suggestions.