Nearly 86 million Americans shop on their smartphones, according to comScore. However, an estimated 40% of shoppers use three or more channels before making a purchase—up from just 10% in 2002. That multichannel path to purchase makes it tough for marketers who are trying to attribute which sales are linked to which channels—especially the increasingly popular, thus important, mobile channel.

“There are different ways to talk about attribution. First click, last click, etcetera,” says Jeffrey Barnes, director of marketing and predictive analytics at digital agency Huge. “Marketing-tech vendors are starting to be really open about the algorithms they use for attribution.” Barnes says that willingness to test varying attribution models is proving to be necessary for marketers who want to accurately attribute sales, signups, and other desired actions to mobile.

But some marketers are erring on the side of caution. “I’ve spoken with a lot of advertisers, and many of them aren’t totally comfortable with mobile because it lacks a true attribution model,” says Chuck Moxley, CMO of mobile advertising technology company 4INFO.

Even so, Juniper Research reports that businesses will spend $7.4 billion in mobile messaging by 2017. As brands continue to invest in mobile, effective attribution for the channel will rise in priority. One challenge with mobile attribution is that cookies—though the industry standard for tracking customers online—are ineffective in the mobile channel. “Mobile doesn’t really have a persistent identifier like a cookie,” says Mike Fyall, VP of marketing at mobile linking company URX. “Without cookies, the focus for mobile [attribution] has been on the [mobile-app] install.”

A mobile app download or install is, at best, a dubious success metric, according to 4INFO’s Moxley, who suggests that an install is about as telling as a page view or an email open. “You have to distinguish between vanity metrics and useful metrics,” he explains. “In the end, if you get one million clicks, that’s great and all, but did those clicks drive sales?”

Users often download apps simply to test them. But almost as often, users will ignore or remove the app soon after downloading it. “A lot of companies are set up to track where their app installs come from, but the focus is starting to shift toward reengagement,” Fyall says.

While some marketers track direct activities like installs, others measure activity through deeplinks to see where app users are earlier in their purchase cycle. “Look at the way [taxi service] Uber uses deeplinks to drive people from Google Maps to [its] own app,” Fyall says. “Uber can see that these people came from Google Maps.”

Although deeplinking offers new opportunities and solutions for mobile attribution, the practice has yet to provide the same level of data as the all-powerful cookie. “I don’t know how deeplinking will solve the cookie problem,” Moxley says. “If you’re logged in as a Google user then the publisher will know who you are, but if not, then how do you know who the phone belongs to?”

Some experts say a holistic look at the customer journey will give marketers the best insight into mobile attribution. “There aren’t different answers for mobile,” says Emily Wengert, VP of user experience at Huge. “There isn’t a moment where we only talk about mobile. It doesn’t make sense to be so channel specific.”

Whatever the future of mobile attribution may hold, marketers must keep the customer’s journey in mind.

“We’re all very fixated on attribution, but once you have an end-to-end process in place that tracks these comprehensive customer journeys, it’s not much further to look at your customers’ lifetime value,” explains Mark Floisand, VP of product marketing at marketing software company Sitecore. “You have to look at how you marry attribution, customer acquisition costs, transaction value, and customer lifetime values to get metrics that reward more than just the initial transaction.”