A new startup formula seems to have gripped a whole generation of entrepreneurs: give consumers what they want, exactly when they want it.

On-demand delivery companies have cropped up to meet nearly every need, from mattresses to marijuana (where it's legal). Of course, if you've been around startup culture long enough you know that this isn't a new formula at all. You might be wondering if it's 1999 all over again. 

But there's at least one key difference between the Kozmos, Webvans, and Pets.coms of ages past and the delivery startups of today: data.   

GPS phone tracking has made easy work of matching the right customer with the nearest courier. And with more people connecting to the Web on smartphones than ever before, the higher density of users (and potential users) puts scale for new startups within reach. These startups can see where and when there's the most demand as well as the most efficient routes to deliver the goods. 

This is the kind of data retailers in the early 2000s would have killed for, says Daphne Carmeli, founder of the crowdsourced delivery service Deliv, especially since "he who owns data wins." But it's not as simple as getting your hands on the data: If there are going to be winners, they will be the startups that analyze the data in ever-more sophisticated ways. 

Getting More Efficient--Down to the Minute 

Phillip Krim is learning this firsthand as he works to streamline the delivery process for Casper, the mattress startup he founded last spring in New York. "At first we thought it was on-demand delivery people wanted," he says in an Inc. interview, but after surveying customers he realized they didn't want their bed right away so much as they wanted a specific window in which to have it delivered. The solution, then, was to optimize the process, using data he had on hand. 

Today when customers place an order with Casper, they specify when they need the delivery and the startup pulls a bed from the inventory near the time frame. The delivery window has also been whittled down, from three to four hours to one to two. As a result, Casper doesn't need as many on-demand couriers, and the companies themselves can cut costs by pooling customers around certain times. "They can pool our order data with other companies and schedule multiple deliveries at 6 p.m.," says Krim. "That makes us more efficient." 

Being able to predict when a customer wants a delivery is indeed very powerful, says Deliv's Carmeli, who says one-hour delivery service isn't as important to customers as the Webvans of the early 2000s originally thought. "It's more valuable to say, 'I'd like it between 2 and 3 p.m.'" versus "in the next hour." "Predictability and convenience trumps speed every day of the week." 

Yet Keith McCarty, founder and CEO behind Eaze, the Bay Area startup dubbed "the Uber for weed," says speed is exactly the point. "Even though I like the community model of Lyft better, Lyft would have a 20-minute delay [when it was just starting out], whereas Uber would have a five-minute delay," he recalls. Which did you think he chose? To win the game of being on-demand, he says, you truly have to be on-demand. 

In San Francisco, Eaze's average medicinal marijuana delivery takes roughly 10 minutes. But McCarty says it could be even faster. "We want to get our operations almost automated and the only way to really do that is through technology," he says. Currently, his team is tracking behavior among patients to determine how often they make purchases. Soon they'll receive reminders to refill their stash, and eventually those orders will automatically be refilled without the customers lifting a finger. "Once we start to get into that, we can say why request? We know you consume about an eight of an ounce, let's schedule a delivery. It's predictive modelling based on previous transactions." 

Using Data to Diversify 

Knowing customers' purchase behaviors at any given time or day of the week is also empowering in an age when on-demand startups don't house their own inventory. Unlike Webvan or Kozmo, who went under doing just that, McCarty can see an influx of requests coming from Manhattan's Financial District at 5 p.m., for instance, and assume it's from people coming home from work. Then he "can position drivers on the supply side to make faster deliveries," and "make more deliveries in the area per driver." And since the drivers work for dispensaries Eaze partners with, the data is a value-add to potpreneurs curious about what customers want. 

The data extends even further, says McCarty who hopes to eventually share his data with pot farmers. "They're growing the medicine based on what people want," he says, "the data points that they have are much more limited. We're capturing a much broader set of data." 

John A. Deighton, a Harvard Business School professor who wrote a case study on Webvan, isn't quite so convinced data is the cure-all for what ails small startups. Sure, a startup like Eaze can double as a market research source and provide insight into what customers want, "but you're up against the big players," he warns. "Google is interested, and Amazon's interested. They're stronger in the upstream end of the business and can aggregate that information across many product categories. They'll have a really tough time fighting the people with the really broad footprint like Amazon." 

The Power of the Crowd 

The most promising delivery startups will be the ones that successfully tap strong communities within the sharing economy to get a meaningful volume of data, Deighton says. "That's why Uber's being talked about as an $18 billion company," he quips. "If a lot of people combine their observations, then suddenly you see a big picture that's very interesting." 

It's this kind of community keeping audacious startups like Instacart, which offers one-hour grocery delivery in 15 cities, ahead of the curve. Instacart's customers rate the company's personal shoppers on how fast their order arrived and the quality of the items selected. At the same time, Instacart keeps track of data on the stores, such as availability of items. All of this data works on multiple levels, from creating product listings to serving as a crystal ball that can determine when a particular store location will run out of bananas. The data is so detailed that "if a city has three Costcos, we can decide which locations in any given order would best fulfill a customer's order," says Max Mullen, the mastermind behind Instacart's data analysis and the company's co-founder. "When you have a large scale of demand and supply, these problems dissolve themselves."

That said, even Mullen admits "it's easy to forget that the most important thing is what the customer's experience is, and whether or not they have the demand for the thing you're offering." And if you aren't tweaking as you go along, you may grow too quickly--and quickly flame out.

"The reason we began this capacity planning system is because we were growing the company," says Mullen. "One night, we got more orders than we could possibly fulfill and went around delivering them around midnight. The customers were upset, so we said, okay, we have to predict demand," something Webvan never did, notes Deighton. 

"If you're going to [launch an on-demand startup], don't do it the way the U.S.S.R. did things," building everything at once, he says. "It's much better to be tentative and experimental. Build another warehouse if the first one's being utilized. Be very gradual. Very tentative."  

Published on: Sep 9, 2014