“If we had already been live last year in aviation, there were many indicators that would have screamed something was wrong.”
With no end in sight to the Covid-19 crisis, it may be a long time before air travel resumes. Still, eager jetsetters are already speculating: “What will it cost?”
Will airfares jump to make up for lost time and lower capacity? Or will they plunge to incentivize reluctant tourists? The same question applies to hotels and attractions.
Israeli startup Fetcherr claims to have cracked the pricing code. Using artificial intelligence that considers hundreds of data sources, Fetcherr’s founders say their tool can predict prices up to a year in the future with 90 percent accuracy.
Well, that’s the promise at least. So far, Fetcherr has run its algorithm on 1.5 million randomly selected items on Amazon.com and found that its price predictions were correct when they checked back two weeks later.
Fetcherr, which was founded in 2018 and launched its technology late last year, simply hasn’t been around long enough to predict prices a year down the road.
But what’s remarkable about Fetcherr’s potential is that it can work its pricing magic even in the presence of an unanticipated “black swan” event — like a global pandemic that grounds most of the world’s air traffic.
“If we had already been live last year in aviation, there were many indicators that would have screamed something was wrong,” the company’s CTO Shimi Avizmil tells ISRAEL21c.
News into numbers
He points to a study published by Harvard researchers that detected a noticeable rise in vehicles parked outside hospitals in Wuhan, China, between August and December 2019, coinciding with an increase in searches for possible coronavirus symptoms like dry cough and diarrhea.
Fetcherr’s AI, Avizmil said, can spot these kinds of correlations human eyes often miss. “We turn news into numbers.”
While China has disputed the Harvard findings, it nevertheless illustrates how Fetcherr’s prediction tools uniquely take into account unexpected types of data.
For example, the weather.
If Fetcherr’s “Deep Price Neural Network” notes that a hailstorm is predicted in May that could damage cars, it understands that families in the affected areas may have lower travel budgets for the coming summer season.
Airlines would thus do better to drop prices a bit “to not miss customers who are price driven,” Fetcherr CEO Roy Cohen tells ISRAEL21c.
Fetcherr similarly tracks conventions, cultural and political events that may increase demand for a specific time frame, which would allow an airline to raise prices. It looks at competitor pricing, demand curves and the availability of inventory.
Fetcherr is currently conducting a pilot with an unspecified airline in Asia and has demonstrated that “had the airline used our system in 2019, they could have added 2% to their revenue by boosting prices by a few dollars here and there on certain routes, Cohen explains. The airline wound up leaving $1.6 million “on the floor.”
Predicting the price of anything
Fetcherr is not airline specific; its engine can predict the price of anything, from toasters to trousers. A “price” can be insurance rates, credit scores, used car values or even the price of bail based on the AI’s estimate of how great a flight risk a particular suspect is.
“Our system is unbiased by subconscious judgments. We eliminate the emotion” in complex decision making, Cohen explains.
Fetcherr is starting with the travel industry in large part due to Covid. Companies “don’t have a clue what’s going on,” Cohen says. People can’t react “as fast as computers. We provide certainty in uncertain times.”
Fetcherr’s technology is based on the field of “algo-trading” – that is, high frequency, automatic trading by computer algorithms on the stock exchange. CTO Avizmil worked with financial trading technology for much of his career before co-founding Fetcherr with Cohen.
The company is working with “more than five clients” today, says Cohen. The investors in hospitality chain Ramada put money into Fetcherr, which has raised a total of $1.4 million and employs 15.
Fetcherr’s biggest competitor is US-based Data Robot. Cohen has nothing but praise for that company, which has raised over $300 million. But large companies like Data Robot “need 18 months to build a team to work with your company,” Cohen says, claiming that Fetcherr can do the same in three to six months.
No black box
Going forward, Fetcherr’s goal is to build the world’s first “downloadable AI toolkit” akin to Microsoft Office, Cohen says, that you can run yourself and that provides transparency into how it comes up with its insights.
“Companies don’t want to buy a ‘black box’ that tells you to do this or that without explaining why,” he says.
Fetcherr’s simple interface for customers has customizable feeds for pricing history, revenue, news, competition and big-picture correlations.
Ultimately, the goal is for a company’s IT department to be able to install Fetcherr’s software independently.
Fetcherr can work in two ways with its customers: automated or manually.
In automated mode, when Fetcherr says a price should go up or down, the change is implemented without human intervention.
Most companies prefer manual – either because they prefer hands-on control over critical decisions or because their computer systems won’t allow automated pricing changes.
Fetcherr’s Amazon demo was just for internal purposes – no Amazon vendors asked for it (although they might now). Robby Nissan, Fetcherr’s chief business demand officer, told us the Fetcherr team would have gladly continued beyond the 1.5 million products analyzed, but it took place before Fetcherr had raised money “and the server costs got too high.”
The idea behind Fetcherr was hatched at a kids’ birthday party in Netanya, where Fetcherr’s founders all live and where the company is now located. Cohen and Avizmil had taken a break away from the noise and started talking. Fetcherr’s customers are now reaping the benefit of the kindergarten chaos.
So, what does Fetcherr say about taking those kids on a holiday abroad if and when the pandemic subsides? Will it cost an arm and a leg?
For now, Fetcherr’s analysis shows that prices are down as airlines “have to show revenue and boost demand,” Cohen says. When flights return to pre-pandemic levels, Fetcherr will be there, too, forecasting the future and watching out for more black swans.
For more information, click here