Bitcoin, dedicated servers, disruptive technology, and machine learning… What do these all have in common? Buzzwords of the modern internet, sure… but now they’re a part of xxxmultimedia.com.
Big Data; Small Business
Firsts things first, we’re talking big data. Yes, that big data. No matter the size of a store — there’s a lot of benefits to running a data-driven business.
We used machine learning algorithms to analyze hundreds of clip orders from our customers. Our hypothesis was by doing we’d discover a list a videos to recommend alongside any other video in our catalog. Through our research we found some underlying patterns.
From our data we learned that on average 15.1% of orders will contain multiple products. When we looked at those orders the trend revealed itself; similar items tended to be ordered together. That makes a lot of sense.
- People buying gender transformation videos tend to purchase more gender transformation videos… and so on.
- People tend to purchase videos containing the same pornstar on a single order.
- People more often purchase clips together which were released around the same time.
Quantifying this trend into useful data was the difficult part. How do we take all these different trends into account? Like this…
From the clip orders we analyzed, we learned learned a lot! With a quick bit of tuning we got a good idea of which videos our customers buy together. Below you’ll find an example of some of the results specific to our store.
You can now find these product recommendations around the site. Check them out!
To the Producers
If you own a clip store or e-commerce store, you can recommend videos that your fans are really looking for.
WordPress, Magento, SQL, CSV, etc. we’ve got you covered. Send us an email or contact us if you need professional help with analyzing this kind of data.
Please Note: This is recommended for stores with at least several orders per video; multi-clip orders are required.
For The Nerds
Our process mentioned in the Big Data; Small Business section was as follows.
Two spreadsheets containing the following data:
- Our entire video catalog on xxxmultimedia.com
- SKU (Each of our videos has a unique id to identify it across all stores.)
- ID (Unique to our store)
- A list of clip orders from our distributors containing the following fields in each row.
- Order Number
- Order Item Number
- SKU (The same SKU which appears in our catalog)
Using this data we grouped together all individual clip orders by their order item number. This gives us a list of all orders, and the items on those orders.
Then we filtered out all orders which only contained one item.
Using confidence and support parameters we we’re able to factor out orders which didn’t seem relevant. After that we manually prune any outliers from the results. This was used as our final list of product recommendations.