It all started with a traffic jam. In 2009, Charlie Davies was driving to an important appointment on the roads of London. He had looked up the shortest route before he left home, but as he waited in bumper-to-bumper traffic that day, he realized that he didn’t care how far away his destination was; he cared about how much time it would take him to arrive.
Even then, it was common for people to use a point-A-to-point-B search tool before going anywhere. But Davies recognized that there was a gap in the market, and that gap could be filled by a tool that let people search for content based on travel time, not travel distance.
So, Davies called up a former work colleague, Peter Lilley, and pitched him the idea.
“It’s an incredibly intuitive thing,” Lilley says. “Whenever someone tells you a distance, in your head you convert it to time. We just wanted to take that guesswork out of the process.”
Together, they came up with an idea for a search engine that would let people search for various destinations based on the time it would take to get there. For example, if you live in New York and you’re looking for a restaurant, it’s not very helpful to search within a five-mile radius because five miles in one direction might take 15 minutes, while 5 miles in another direction could take an hour. Plus, it all depends on what mode of transportation you use.
Davies and Lilley named their company iGeolise. And after many iterations, they settled on a B2B model in which they would license their search platform to other companies. But before they could scale, they needed the right technology to help them return search results instantly and ensure that their database could be updated with new travel information quickly. For that, they turned to Oracle.
Time for Something New
The first iteration of the iGeolise TravelTime platform was a consumer-based search. The problem with that B2C approach was that it would put iGeolise in competition with the likes of Google, Yahoo, and other large companies with extensive resources and deep pockets. Even smaller companies such as hotel chains already have their own search tools on their websites. It didn’t seem likely that iGeolise would be able to draw traffic away from every site that provided its own search.
That’s when Davies and Lilley decided on a B2B model. They would build a platform that any other website or mobile app could link to or embed, and allow their end users to search by travel time and mode of transportation.
“By building a platform that websites could link to via an API, we would let them use their own content for their existing audience,” Lilley says. “That was fantastic for us because it meant we didn’t have to build up a bunch of content and attract our own audience.”
They worked on a shoestring budget to build the platform, forgoing investor capital and just focusing on making a great product. But when it was done, the engine took several seconds to return results. For consumers who would be using the tool, that was simply too long to wait. After all, according to an oft-quoted 2008 study, a one-second delay in page load time can lead to a 7% loss in conversions. And in the intervening 10 years, consumers have no doubt grown more demanding.
So while Davies and his team of developers worked on speeding up the platform, Lilley got creative and began work on a separate revenue stream that would keep the company funded.
Deep Data Analysis
While a few seconds was too long for a consumer to wait for search results, Lilley realized there was another user base that wouldn’t mind that short wait at all: business users who would otherwise have to crunch this data manually.
Large businesses regularly need to access data about travel times to make important decisions. For example, a company’s HR team might be tasked with finding a new office location within 45 minutes of most current employees’ homes.
“If someone had to figure that out by plugging addresses into Google Maps, one by one, it would take forever,” Davies says. “It might take our platform a few seconds, maybe even a full minute. And while that’s too long for someone on a consumer website, it’s not for an HR team. They could just go grab a cup of coffee in the kitchen, come back, and the results would be ready.”
For those users, the early iteration of iGeolise’s platform was already plenty fast, so the company started marketing it directly to them. They quickly amassed a base of customers that included HR teams, as well as retailers whose leaders needed help planning where to put their next locations. For example, imagine that a coffee seller wanted to be sure that all people in a certain area live within a 15-minute walk of the company’s nearest location. The iGeolise platform could quickly show the company’s planners where walkers have gaps in coverage—or where there are too many stores for a certain geography to support profitably.
iGeolise’s analytics business grew quickly because it helps clients save significant amounts of time, and soon after its launch it was providing plenty of revenue to help Lilley and Davies scale. In the meantime, developers were hard at work to make the search tool fast enough for consumers so they could license it to other companies—and that work was about to pay off.
Oracle Technology Speeds Processing
One of the biggest obstacles to a lightning-fast search engine was simply getting the data into the database in the first place. When iGeolise launches in any country, it must first spend an enormous amount of time finding roads and walking and bike paths, as well as collecting data on the country’s public transportation systems.
Developers produce complex algorithms that can predict traffic density at certain times of day. They also have to take into account factors like cultural transportation norms. For example, in Amsterdam, it’s common for commuters to use a combination of bike and bus or bike and train. In the UK, it’s common for commuters to drive to a train station and complete their journey by rail.
But that data doesn’t necessarily stay the same. A city might revise its bus timetables, for example, so iGeolise’s developers update the database every week for every country. Lilley likens it to a clown with spinning plates.
“It would take us maybe a whole day just to update the UK,” Lilley says. “But we’re in 26 countries now, so we haven’t got just one spinning plate; we’ve got 26 of them, and we have to keep them spinning all the time. That was becoming a real barrier for us because it would take us a week’s worth of processing time to update our system, and that clearly doesn’t scale very well.”
To speed up processing, iGeolise switched to Oracle Cloud Infrastructure Container Engine for Kubernetes with an integrated Docker-compliant container registry as a fully managed service.
“It cut our parsing time from well over a week, every week, to well under a day,” Lilley says. “That means when there’s an error in the data we’re bringing in, we can correct it and just reparse it the next day. Kubernetes and Docker on Oracle has made a significant difference to our ability to scale.”
The iGeolise platform also uses a custom-built geocoder on Oracle VM to convert human-readable postal addresses into latitude and longitude coordinates and back again.
“A lot of our clients spend quite a lot of money on geocoders, and now we can provide our geocoder within our API to our existing customers, which saves them a bunch of money,” Lilley says.
Finding the Best Route
Next, Davies and his developers decided to build a field sales routing tool that would help a company’s sales reps figure out the best route to visit the most customers in a given day.
“There are lots of field sales routing tools, but they’re always routed by road,” Lilley says. “We thought that was a bit weird because half the world lives in urban areas, and in an urban area you don’t drive if you can avoid it.”
So the iGeolise development team built a routing application that uses all modes of transport, including public transport. Initially, the application took about a day to process results for a single client. But using Oracle Cloud Infrastructure, the iGeolise platform can now process results in less than an hour.
The potential benefits are extensive. Imagine that a healthcare provider has to make 100 home visits in a day and has only 12 community health nurses in the area. The iGeolise platform can allocate patients among those 12 nurses according to the nurses’ starting addresses and put them in the optimum order so travel time is minimized.
Ready for Prime Time
Today, the iGeolise TravelTime search tool can return 100,000 ranked and sorted results within 300 milliseconds. Clients such as Zoopla, a property search website, report a 300% increase in conversions. Jobsite, an employment search tool popular in the UK, reported that location relevance had increased by 60%. Many other customers report a benefit to the bottom line—and both Davies and Lilley say they could not have accomplished these results as easily without Oracle.
“We have very, very intense CPU requirements, and Oracle Cloud Infrastructure has helped us increase the speed of searches,” Davies says. “Oracle VM enables us to scale, so that if a client starts doing 10 times as many searches tomorrow as they were yesterday, we can happily and easily autoscale into that.”
Now, Davies and Lilley are free to chart a course for the future, which will include ever-faster search times and scaling the sales side of the business. The next focus will be on expanding their reach among retailers and hotels.
“Our field sales routing tool, our geocoder, and our fast parsing—we could not have done those three things at the speed we do without Oracle,” says Lilley.