Everything was going well for innovative entrepreneur and restaurateur Haruki Odajima — or so it seemed until the coronavirus pandemic swirled across the country earlier this year.
His artificial intelligence business, Ebilab, which helps restaurants predict how many customers they will get with more than 95 percent accuracy, was flourishing. That success was driven by his long-established restaurant Ebiya in Ise, Mie Prefecture. He had been using the Japanese restaurant to demonstrate how his system can more than triple productivity, grow profits fivefold and reduce food waste by more than 70 percent.
Then came the coronavirus. With businesses asked to close and people told to stay in, the restaurant industry has chalked up unprecedented losses as bankruptcies sweep the country. And Ebilab’s invention, which uses over 100 different data sets to project customers up to 45 days in advance, was no panacea.
But Odajima hasn’t given up hope. In fact, he says the pandemic could precipitate an overhaul of the industry by forcing it to abandon its low-margin business model.
“We need to make decisions based on statistical data instead of intuition and experience as in the past,” he says.
Ebilab’s attempts to project daily foot traffic remain rare in the industry, though an increasing number of chain eateries and retailers are taking advantage of smartphone apps to issue specially tailored marketing coupons based on big data, said Shun Tanaka, a senior analyst at SBI Securities.
For Ebilab’s Odajima, the need for data-driven management was highlighted when he decided to temporarily shutter his restaurant on April 7 after the system drew attention to an alarming decline in foot traffic in the vicinity, which is usually crowded with tourists drawn by the Grand Shrines of Ise. The decision made him an outlier in the area as most other restaurants stayed open until business suspension requests were issued on April 20.
“We are good at collecting data, such as the number of pedestrians, by installing various sensors in the city,” Odajima said in an interview in Tokyo. “There’s a strong correlation between foot traffic and the number of dining customers at our restaurant. We devised our own algorithms for (deciding on) closure by taking into account which of the two options would lead to smaller losses: staying open or closing and using the state subsidies for retaining employees.
“We decided to close the restaurant on April 7 when the daily foot traffic in front of the store fell below 2,000 per day,” he said.
Odajima also relied on the AI program to select his reopening date of May 27.
“After the central government lifted the state of emergency for most of the prefectures including Mie in mid-May, most restaurants reopened too early and suspended business again after customers didn’t come back fast enough,” he said.
“We reopened after foot traffic topped 1,500 per day. This is because the government had raised the ceiling for the employment adjustment subsidy to ¥15,000 per day per employee from ¥8,330, so we decided to lower the bar for reopening.”
Inquiries about Ebilab’s AI services are robust, and the company now has about 160 to 170 subscribers. In cooperation with Persol Holdings, which invested ¥100 million in the company, Ebilab aims to have 10,000 subscribers by the end of 2022.
Another service Odajima launched during the two-month shutdown is a system that tells you how crowded a restaurant is in real time using in-store digital signage and a smartphone or web browser.
Odajima has become a poster child for entrepreneurs after transforming Ebiya — a restaurant more than a century old that was still relying on abacuses when he inherited it from his father-in-law in 2012 — into a restaurant armed with cutting-edge data.
The biggest question the nation’s several hundred thousand restaurants now face is how to survive the coronavirus.
Restaurant sales plunged 39.6 percent in April, the sharpest year-on-year decline since 1994. Though the sales drop eased to 32.2 percent in May and narrowed further in June, SBI’s Tanaka says there seem to be large differences in the rate of recovery, depending on the business model.
Overall, he says the fast food segment appears to be faring better, though there is no guarantee sales will rebound to pre-pandemic levels.
Odajima, however, says the industry is focusing too much on cheap food underpinned by low profit margins and high turnover.
“We must take a fundamental look at the existing business model. In that sense, we are lucky to have made use of data, as it has shored up our profitability,” he said.
In addition to reducing food losses and optimizing preparation work for the next day, Odajima says AI-based management can raise sales per customer significantly based on constant improvement.
For example, such incentives as a QR code-based questionnaire filled out in return for a free rice cracker show data-driven techniques allow for objective decision-making and demonstrate that failure-proof investments are possible, he said.
In Japan, the number of restaurants taking advantage of today’s digital transformation, dubbed “DX,” to create new value account for less than 1 percent of the total, meaning countless opportunities have been missed to generate growth, says Ryuji Tokiwagi, Ebilab’s chief technology officer and evangelist.
The results of this high-tech push can of course be found at Ebiya, which is employing both student discounts and rainy day discounts to shore up sales amid the pandemic. The former produced a jump in customers, he said.
“Before we offered the 10 percent discount, students accounted for only 4.6 percent, but the ratio jumped instantly to 18.4 percent after we offered the student discount,” Odajima says.
Asked whether the discounts would erode profitability, he said: “If the number of customers increased to 150 from 100, then we would be fine with 30 percent discounts.”
Despite the lifting of the national state of emergency in late May, tourism in Ise was bound to remain low. So Ebilab launched another service at the same time that lets people virtually shop at its adjacent souvenir store and interact with employees by web camera. It also recently launched a virtual reality tour service that takes people on simulated tours of Ise shrine.
“The AI can be used effectively for improving efficiency or marketing, but it cannot handle everything,” Odajima says. “But I know from my own experience that many solutions to various challenges come when we combine all elements of technology, creativity and personnel.”