Skip to main content

Amazon FBA

Create an automated system to identify, buy, and sell merchandise on Amazon FBA (Fulfillment by Amazon).

Result was a system to analyze a high volume of products and determine which one's are profitable.

Overview​

Fulfillment by Amazon (FBA) is a service that allows businesses to sell products on Amazon and outsource order fulfillment to Amazon.

Businesses send products to Amazon fulfillment centers. When a customer makes a purchase, Amazon picks, packs, and ships the order.

FBA is attractive because there are 1,000s of niche categories that Amazon does not sell into.

The problem​

Determining which products to sell on Amazon is time-consuming because Amazon FBA is a unique market that requires a lot of analysis to be profitable.

The analysis is data-intensive, difficult to automate, and frankly, difficult to understand.

Financial success is based on finding profitable products. Researching products, one at a time, is time-consuming and available solutions do not provide enough information to make an informed (profitable) decision.

You need to determine the following for each product sold on Amazon:

  • Does Amazon offer the product and what are the variations (color, size, multiple products per package, shipping size and weight, etc.).

  • What is the current price and what is Amazon's FBA percent of the sale.

  • Analyze monthly sales volume; factor seasonality, popular variations, etc.

  • What is the historical pricing; is the current price too high or low.

  • How many sellers are competing for the "buy box," which determines who is credited with the sale.

  • How deep is the inventory; if the backlog is large (1,000) and sales are slow (200/month) you many never see a profit.

Sourcing products​

Sourcing products is the primary means of profitability. You need to either sell high-volume products with a reasonable profit or sell low-volume products that are highly profitable.

Sourcing products is just as important as determining profitability. Many items are bought from retailers, such as unique or close-out items from Target or Walmart, or procured from wholesalers and distributors.

Create a product - high risk and high reward​

Some manufacturers create unique products that can only be found on Amazon. They are the first, or early, entrant in a market and popular products can be successful. A few examples are premium yoga mats, baby merchandise, or unique kitchen gadgets.

Most products are designed in the US, manufactured in China, and ordered in multiples of 20-foot shipping containers. This is an expensive way to start a business.

The above is a small percent of Amazon's business. It's written about often enough that you would think the model was mainstream.

Creating a product can scale to hundreds of millions of dollars of revenue per year.

Wholesale - medium risk and medium reward​

Wholesalers often specialize in broad categories, such as clothes, kitchen wares, sporting goods, toys, etc. Some examples are niche kitchen goods or discontinued toys such as Legos.

Wholesalers represent a limited number of unique products (SKUs) and often require large, pallet size, minimum orders.

If you are able to find a product that's profitable and doesn't have much competition, then you may have struck gold.

Wholesaling can scale from 10 million to 100 million dollars of revenue per year.

Retail arbitrage - low risk and low reward​

Many people get started selling on Amazon by shopping for profitable products at retailers (Dollar stores, Target, Walmart) and shipping them to an Amazon warehouse for customer fulfillment. The most common FBA model.

For example, a friend often finds high-end close out plumbing faucets at local hardware stores. These products have low sales volumes (a few hundred per month) and have a built-in profit because the Amazon price is close to the MSRP (Manufacturer's Suggest Retail Price).

Retail arbitrage is time-consuming and does not scale beyond a small business.

Retail arbitrage can scale to 10 million dollars of revenue per year.

The wholesale problem​

Creating unique products or doing retail arbitrage is not, for most businesses, the path to financial success. Either too much risk or not enough reward, which often go hand-in-hand.

For many businesses, building a network of wholesalers is the route to success.

Having access to wholesalers that other Amazon FBA businesses have not found, is one part to the solution.

A wholesaler may have a upwards of 10,000 products they represent. Think of a wholesaler that has kitchen items and toys. Each category could have 1,000s of items. And each item has many variations (color, size, options, etc.).

Determining profitability, one product at a time, is laborious and impossible to do at scale. Consider how long and difficult it would be to analyze a spreadsheet with 5,000 products. You would give up before you got to the end.

The solution​

Software is required to automate the process - cherry-pick the winners, and automate replenishment's.

Software can evaluate 1,000's of products in minutes and identify profitable items with healthy sales volumes.

My software did just that - ingest spreadsheets from wholesalers, pick winning products, and determine order sizes based the Amazon price and backlog quantity.

The software​

Development of this software took more than a year. Most of that time was research, proof of concept, learning, and building relations with wholesalers.

Once the idea solidified I wrote about 100,000 lines of Python code that ingests spreadsheets and analyzes it for profitability.

A majority of the system is API calls to Amazon, which are rate limited. This is a huge problem because you need to collect lots of data to understand "the market" for each product.

Many people have resorted to an army of anonymous bots that pull Amazon data to feed their appetite for data, i.e., get around rate limitations.

Another challenge is physical products are messy...

  • ASINs (Amazon Standard Identification Number) are unique to Amazon and identify a particular item and one (of many) variations.

  • SKUs (Stock Keeping Units) are not exact and mapping a SKU to and ASIN is an art, not a science.

  • Amazon's catalog is not published; determining where a product resides in the catalog hierarchy, its sales rank, historic pricing, the number of items in the Amazon warehouse, and other attributes is forensic art.

  • SKU variations (red, large, 2-pack, etc.) are not well represented in the Amazon data. For example, 2- and 4-packs of toothpaste are popular. But there is a market for 10- and 20-packs of travel size toothpaste among travelers. That item wasn't available on Amazon but proved to be a popular and profitable item.

  • Bad data and typos are the norm, not the exception.

Two real-world examples​

Legos​

Amazon identifies low inventory items. One of those items, a Lego set, matched a product a wholesaler had in stock.

Pallets of Legos were ordered and shipped from the wholesaler to Amazon. Everything sold within a week.

Amazon deposited funds in the account before the wholesaler bill was due.

Toys​

I went to a trade show in Chicago that specialized in kitchen wares, seasonal goods, tools, toys, etc.

At the show, a toy distributor had a spreadsheet with SKU numbers, prices, quantity, etc.

Using my software, I was able to identify 50 products with high profitability and low competition.

Within an hour I placed an order at the show and a week later most of the merchandise landed at an Amazon warehouse in Los Angeles. As luck would have it, the distribution warehouse was 25 miles from the Amazon warehouse. A true drop-ship arrangement.

Within weeks, sales revenue from Amazon was credit to mt account. A month later replacements were ordered.

Technologies​

Deep understanding of the Amazon Selling Partner API (SP-API).

The back end was written in Python. The code base is about 100,000 lines of code.

MySQL is used for the database.