Fetcherr: Harnessing the power of deep learning and Google Cloud to optimize travel fares
About Fetcherr
Founded in 2019, Fetcherr aims to disrupt traditional rule-based (legacy) revenue systems in the aviation industry through its proprietary AI-powered goal-based enterprise pricing and workflow optimization system.
Tell us your challenge. We're here to help.
Contact us
About Sela
With 450 experts and 30 years of proven experience, Sela helps customers around the world with their architectural, infrastructure, development and automation work.
Fetcherr uses BigQuery and Vertex AI to support an AI-centered, real-time price optimization platform for the airline industry.
Google Cloud results
- Drives revenue for airlines with a real-time, AI-powered, plug and play pricing solution that minimizes technology costs
- Onboards customers with legacy infrastructure in weeks instead of years with a web application and powerful data pipelines
- Recruits and retains top talent across the world with high security, low friction remote working overseen by Google IAP
Automatically prices millions of fares in real-time
Every flight is unique, whether it’s a business class trip or a summer holiday getaway. But how does an airline know how much to charge for each flight? Pricing airfares involves monitoring thousands of routes for each airline and adjusting prices based on many different factors like the time of year, weather conditions, and the cost of fuel. In a high-cost, low-margin industry like retail aviation, finding the balance between profitability for the airline and what a customer is willing to pay is key to staying one step ahead.
Fetcherr makes this process easier, faster and better, with a platform that uses cutting-edge proprietary AI to predict demand and provide comprehensive, continuous price recommendations in real-time. "We describe Fetcherr as 'Enterprise as a Service'. Pricing is such a fundamental part of the modern airline industry and this solution is effectively an enterprise system," says Robby Nissan, Co-founder and Chief Strategy Officer, Fetcherr. "Our clients get two major advantages from us that they wouldn’t get elsewhere. Firstly, they have a pricing solution built with the very latest deep learning technology. Secondly, they can do this in real-time."
Fetcherr’s offering has seen them go from a proof of concept in 2019, to working with two major airlines, responsible for millions of airfares across Latin America and Europe every year. In order to make Fetcherr as powerful as possible and still keep it plug and play, the company chose Google Cloud as the core of their infrastructure.
Although it launched in 2019, the technology that powers Fetcherr has been years in the making. One of its four co-founders, Dr Uri Yerushalmi forged a successful career pioneering AI for algorithmic financial trading. And Shimi Avizmil was developing real-time financial trading platforms before he became co-founder and CTO of Fetcherr. "Real-time is in our DNA," says Avizmil. "I was working down to the millisecond, while Uri was working at the nanosecond level."
Although Fetcherr has its roots in financial trading, the core principles of the technology are industry-agnostic. Fetcherr is, at its heart, a predictor of demand. The co-founders chose to apply Fetcherr to the airline industry because they saw a volatile market desperately in need of a solution to an age-old conundrum.
Disrupting traditional airline pricing models
In a typical airline, there might be a few dozen people setting the price ranges for millions of tickets every year. Naturally, they cannot take into account all the available data 24 hours a day, so they make their calculations periodically, based on what is hopefully a representative slice of the data.
The aviation industry tends to focus its research and development efforts on planes and avionics rather than back-end infrastructure. "Most of the airlines we see can’t incorporate big data analytics or the latest AI tools without incurring massive infrastructure costs," says Nissan. "They might generate huge amounts more data than they did 10 years ago, but with many still using legacy on-premises infrastructure, they are unable to convert this data into useful insight."
In early 2020, when the pandemic led to a global shutdown, few industries were hit harder than aviation. Airlines had to cut costs more than ever and Fetcherr’s solution could not have come at a better time. "All of a sudden they found that their traditional, rules-based pricing models had no idea how to price flights and they needed a solution fast," says Nissan.
Building a demand predictor
From the very outset, Fetcherr chose Google Cloud for its infrastructure. "Google Cloud was the natural choice for us," says Avizmil. "Firstly, it was vital we had the best security for us and our clients. Secondly, we wanted a platform that could integrate a Kubernetes-based environment. Finally, it provides us with all kinds of opportunities for advanced analytics and artificial intelligence at scale."
After building its proof of concept, the company teamed up with an airline to develop its first commercial release. Realizing that it would now be working at a much bigger scale, the co-founders chose to work with Google Cloud Partner Sela. "They knew the Google Cloud landscape very well and found exactly the right solutions for our use cases," says Avizmil. "They were particularly helpful with billing, allowing us to scale up while keeping costs to a minimum."
Fetcherr’s platform is simple at its core. It is designed to take in a huge amount of data from a wide variety of sources, analyze it, then use the results to predict the optimal pricing strategy for any given situation.
"Google Cloud was the natural choice for us. Firstly, it was vital we had the best security for us and our clients. Secondly, we wanted a platform that could integrate a Kubernetes-based environment. Finally, it provides us with all kinds of opportunities for advanced analytics and artificial intelligence at scale."
—Shimi Avizmil, Co-founder and CTO, FetcherrWhen Fetcherr onboards a new client, the first challenge is to find and ingest the airline’s own data. In legacy infrastructure, data sources can often be siloed and Fetcherr takes the necessary time to ensure that all the relevant information is made available and with minimal effect on the client’s servers. The datasets are then transferred from the client to Google Cloud Storage buckets via SFTP. With Dataflow, Fetcherr has built powerful pipelines that transfer these data files into the company’s data warehouse, built with BigQuery. In addition, there are several pipelines for external data sources such as weather reports, currency exchange information, and political risk assessments.
Once all of the data is in BigQuery, Fetcherr passes it through its proprietary deep learning models, many of which were built with Tensor Flow, the open source machine learning technology developed at Google Cloud. The models then use the collected data to form predictions at a much more granular level. While the models themselves are proprietary, Fetcherr deploys them with Vertex AI. "Our developers loved working with Vertex AI," says Avizmil. "It’s a great tool to integrate all our models onto one environment and unify our whole process."
The entire infrastructure is designed around microservices, orchestrated with Google Kubernetes Engine for maximum scalability along with Google Cloud TPUs for improved speed with machine learning processes.
"Our developers loved working with Vertex AI. It’s a great tool to integrate all our models onto one environment and unify our whole process."
—Shimi Avizmil, Co-founder and CTO, FetcherrShifting a paradigm
With Google Cloud at its heart, Fetcherr is in a position to revolutionize pricing in the airline industry. Its ability to ingest and process huge amounts of data along with its deep-learning technology means that, for the first time, airlines can look at the totality of information available and achieve a granularity that simply isn’t possible with traditional pricing models. That has led to significant revenue gains of up to 40% in specific cases. Fetcherr’s existing customers can redefine the post-pandemic landscape with the best possible information at their fingertips.
Thanks to its SFTP protocols and data pipelines, Fetcherr can offer its solution without leaving a big footprint in the airlines’ IT operations. "It’s completely plug and play for our customers," says Nissan. "If this was an on-premises solution, it could take up to two years to integrate with their infrastructure. Because we are a web application, we can get them up and running in a few weeks."
With the highest levels of regulatory compliance with Google Cloud, Fetcherr can ease its customers’ worries about data security wherever they are in the world. The company also makes extensive use of Google Identity-Aware Proxy (IAP) for remote working for a zero trust environment with low friction user experience. This allows Fetcherr to recruit some of the world’s best talent, and keep them. The company has a 99% retention rate for technical staff. "We hire the people we want to hire and distance isn’t a barrier for us," says Avizmil.
Smart and effective use of Google Cloud has helped Fetcherr stay in control of the costs associated with scalability at speed. "Google Cloud lets us optimize our usage so we can operate at the scale we need to but still keep costs minimal," says Avizmil. "We often have to ingest 200 gigabyte files in seconds, so we can add resources on demand. But then we can scale back down."
That ability to manage its resources means that Fetcherr has been able to focus on perfecting its platform and expanding its customer base. In the second half of 2022 the company plans to bring onboard new airlines that will dramatically increase the demands on the platform. "We have some very big customers joining us and our work at the moment is about making sure we have a highly scalable environment," says Nissan. "Then in the longer term, we’re looking forward to expanding into new verticals like hotels, cargo logistics, and even finance. We like what we’ve done with Google Cloud and we’re confident that we can meet whatever challenges lie ahead."
"Google Cloud lets us optimize our usage so we can operate at the scale we need to but still keep costs minimal. We often have to ingest 200 gigabyte files in seconds, so we can add resources on demand. But then we can scale back down."
—Shimi Avizmil, Co-founder and CTO, FetcherrTell us your challenge. We're here to help.
Contact us
About Fetcherr
Founded in 2019, Fetcherr aims to disrupt traditional rule-based (legacy) revenue systems in the aviation industry through its proprietary AI-powered goal-based enterprise pricing and workflow optimization system.
About Sela
With 450 experts and 30 years of proven experience, Sela helps customers around the world with their architectural, infrastructure, development and automation work.