“We know our customers and products better than any third party,” they told us. What they were unsure about was how to handle the volume and deploy an enterprise solution. Millions of people used the company’s website per day, and the new product recommendation engine needed to run at 100% uptime with sub second performance. In short, they needed an in-house product recommender capable of enterprise-grade performance.
There were 3 challenges the new product recommender had to meet to replace the enterprise software. It needed to be–
If these technical challenges weren’t enough, the company's license for its commercial recommendation product was set to renew in a handful of months. If we didn’t get the new software finished and implemented quickly, the company would have to shell out millions of dollars to keep their product recommender on their website.
The Solution: An Enterprise-Grade Product Recommender
Working in close collaboration with client subject matter experts, we deployed 40+ unique models to recommend products to customers. These models took into account data such as:
Personalized to each customer, the models would allow the software to recommend products it determined the customer would like. The result was a tailored user experience that both increased click-through rates, purchase conversion and customer satisfaction.
As indicated by these increased conversions, our recommender made even better predictions than the commercial software.
In order to meet the demands of an enterprise online presence, we had to make sure the new platform was fast! We were aiming for sub100 ms response time to present recommendations to the user. When we benchmarked our recommender, it performed 10 to 20 ms on average, making it even faster than the commercial software and well within the requirements to support this enterprise online sales channel.
Since our client was a global enterprise with annual sales in the billions of dollars, the client’s website was handling millions of visits around the clock. With that volume of traffic, even a few seconds of downtime could end up costing the company thousands of dollars. Given KIS’s experience building enterprise software, we design for scalability and reliability in mind from the start! The recommender we built has been running reliably for 2 years now, no unexpected downtime from technical issues.
This was our first project implemented on the Google Cloud Platform. While it was challenging to go into production with such an aggressive schedule on a cloud with which we had little familiarity, we met the challenge head-on and completed it without any major setbacks.
With a tight 3-month deadline looming, we had no time to waste. KIS took a team of 4 talented engineers with different skillsets and got straight to work. We brough an enterprise architect with 20+ years of experience building world class software, a senior full stack developer, a machine learning specialist and a front-end engineer. We believe that small, focused teams of talented individuals can delivery faster and with better quality. This project was a textbook example of exactly that.
In a matter of days, we had a working POC; and few months and many iterations later we had an enterprise class product recommender that we knew was ready for prime time!
From there, it was all about time management and working with other departments in the company to implement the new product recommender in time.
This took a great deal of coordination, supplying clear and firm dates to the website engineers at the company that needed to integrate with our product. We needed to clearly set expectations, provide documentation and follow-up with them regularly to make sure they were keeping pace. Good technology is nothing without successful execution. At KIS we have learned these real-world lessons and we focus as much on planning, communication and logistics as we do on technology!
In the end, all parties involved we able to hold firm to the schedule and the team delivered the production-ready software to our first market in time for a fraction of the cost the client to renew the enterprise software license.
Writing custom software is hard. Large companies often assume that building software is more risky than buying software, and in many cases this is true. Many organizations have seen failed projects which sour them to the idea of building custom solutions. It is hard for large companies to hire and retain top tier technical talent so they feel less confident that they can build something. This project is a perfect example of where KIS’s expertise can be leveraged for a client.
This client made the brave decision to write something to replace an off the shelf tool. In most cases a talented team can build something as good if not better than commercial software, often for a fraction of the cost. This is exactly what we helped this client do!
You bring the idea, and we can build it!
Enterprise-grade product recommendation engine built and deployed to first market in ~3 months.
The client has been using our enterprise-grade product recommendation engine for over 2 years now. In that time, the software has served over half a-billion recommendations and is on track to roll out in ~24 countries by 2024.
The recommender has also contributed to an increased click-through and conversion rates: ~36% of website visitors clicked on a recommended product, while ~20% of those customers ended up purchasing one of the recommended products.
Of the items sold through the client’s website, ~24% came from our recommendations this year, generating millions of dollars in revenue.
By implementing our software before the client’s enterprise commercial licensing was set to expire, KIS saved the company millions of dollars in licensing fees. Supporting this custom solution will cost money. However this in-house labor will be at least 1/10th the price of the commercial software licenses.
Our new custom-built recommender outperformed the commercial product in virtually every category – especially speed. Our recommendations are served consistently clocked in at 10 to 20 ms.
Many companies look at software writing as a risky endeavor. Why take the chance of paying someone to write software when similar software already exists? To avoid this perceived risk, companies end up paying expensive recurring licensing fees instead of hiring trusted professionals to create the software for them at a fraction of the cost.
In this case, we built enterprise-grade recommendation engine that outpaced and outperformed an expensive piece of commercial software for a fraction of the cost, saving them millions in licensing fees.
tech strategy & consulting.
Empowering companies, investing in IT, tech training, scalable squads, and delivering the right value.
Data migration, conversion, analysis and integration.
Database consolidation and management.
machine learning models.
Use AI algorithms and data to improve the accuracy of your analysis and make better predictions that impact the decision making processes.
web services development.
From discovery to delivery, we solve issues that go beyond the basics.