They have been the darling of the investment community, and are certainly proving to be a life-saver for many a consumer. We are talking about on-demand services – often accessed via apps – promising that users can dial up a car, meal, handyman or even a massage in minutes.
It seems there is no services segment that on-demand apps won’t touch, as they race around town to meet consumers’ delivery, help and groceries needs. Last year, on-demand small businesses picked up $6.5bn in investment funding – ten times more than 2013, according to CB Insights.
With a flood of new entrants coming in, and with every second counting, what systems make an on-demand provider tick, and how can they move fast enough to keep customers happy?
You’re in the last-mile business
First thing’s first – this is harder than you could imagine. Many businesses think they can crack on-demand with just an effective user interface and some drivers on the books.
Yes, user experience is important. But on-demand services are not in the user interface business: they are in the last-mile business. On-demand culture thrives on the coordination of pickup and drop-off times to suit users’ needs. The ‘product’ here is a person at the customer’s door at a time of their choosing. In other words, logistics.
You don’t get to white-board this business. Once your app reaches a certain level of orders each day, there comes a breaking point. That’s why on-demand services must invest in underlying logistics technology, or risk customer disappointment.
Punctuality is a maths problem
From day one, one of our main KPIs had been punctuality. Now we needed help. So we switched our entire tech focus aware from the pretty stuff, our app and website, to all the things that customer can’t see – developing logistical excellence. That meant hiring two mathematicians to our core backend team and developing our own algorithm to automatically route drivers between customers and laundromats.
Why an algorithm? Well, because although the principles of allocating drivers in required parts of the city are the same as a human operator would process, at a certain point of scale a human’s brain would simply explode with the complexity. For example, whilst coordinating five drivers is easy, coordinating 50 drivers is not. Now our punctuality has hit 95%, you can see how crucial facilitative technology is for running our business successfully.
Only custom logistics are logical
For any services business looking to plan routes for drivers, there are certainly off-the-shelf software solutions to help. OnFleet targets delivery solutions, Viamente helps plan routes for fleets with multiple stops, and Google Maps’ API can be harnessed by other software.
But these only get you so far – literally. When service users have expectations of making an on-the-fly booking with a one-hour arrival, a single new order can throw a driver’s day into chaos. In a city like London, where dozens of new bookings take place every few minutes, systems like these are too rigid to adapt.
That’s why many such services are turning to custom-built solutions.
Demand prediction is the secret sauce
When your favourite grocery or eating service has cracked getting to your door, the next logical question it will ask is: How many drivers do I need? This is a question that puts many under pressure. But driver volume is also the vital side of the supply-demand curve that avoids customer disappointment. So, how can demand be accurately predicted to ensure the correct supply of wheels? The answer is found in data. With enough inputs, demand for services can actually predicted, so that services resource themselves adequately on any given day.
Factors like weather can play a surprisingly large part in demand forecasting. When it’s raining outside and that plan for a romantic restaurant dinner goes out the window, the likes of Foodora can expect a 20% hike in numbers calling for a delivered meal. It’s not just food – even services like ours find that poor weather negatively impacts people’s desire to do just about anything they do in town. But it doesn’t stop there. Others are tapping event listings, for example, to expect high pizza demand following a soccer match, and first-party historical data that can even better inform outcomes.
Of course, harnessing inputs to predict demand in this way involves a lot of data processing. So would-be on-demand services should arm up on big data and computing power.
Take expansion steady
When fast matters, slow and steady can win the race. When on-demand services take off but are constrained to large metropolitan areas, word spreads, demand builds and those in surrounding cities grow eager for their arrival. But smart apps don’t jump to expand too soon, even when competition looms.
The service economy is not like e-commerce, and the same rules of growth don’t apply. Businesses have been able to massively scale their e-commerce activity since the 1990s by capitalising on the drop-shipping construct – they don’t have to build any new infrastructure because someone like DHL or UPS handles logistics for them.
But, for an on-demand service app, expanding to a new city, as Uber recently did in Cardiff, things aren’t so simple. They need to have adequate numbers of both drivers, suppliers and customers in place before even considering it. Far better to become clear number-one leader in a single market first, then roll out large from a strong position.