Adobe has today released its 2016 Digital Insights Shopping Predictions for the upcoming holiday season, and while the data shows some outstanding growth in online sales, one statistic shows just how fragile the market is.

Adobe has an extensive data pool that has allowed it to predict consumer spending accurately. In fact, the company has been within two percent of the actual results over the past two years, making Adobe’s predictions among the most accurate in the industry.

Before we get to the most surprising part of the study, let’s take a look at the headlines.

Adobe expects an extended holiday season with consumers shopping earlier and later in the season than usual. It expects that over 53 consecutive days (57 calendar days in total), online sales will exceed $1 billion. That is in contrast to last year, where the season lasted 31 consecutive days.

Cyber Monday is expected to be the largest online shopping day of all time, generating $3.36 billion in sales. That is equivalent to 9.4 percent growth on 2015. And sales on Thanksgiving Day will increase 15.6 percent year-over-year (YoY) to $2 billion, although Adobe expects it will not reach the 25 percent YoY increase of 2015. Black Friday is predicted to grow 11.3 percent YoY to $3.05 billion.

But the data-point that grabs all the attention is the percentage of shoppers that drive online sales, and the sizeable piece of the American apple pie they control.

“The most surprising results include the fact that 5 percent of customers drive 35 percent of overall revenue for the average retailer,” Becky Tasker, managing analyst at Adobe Digital Insights, told me. “Imagine if this number decreased to 4 percent and the potential revenue impact it could have on a retailer.”

Tasker is right. A small swing in the online sales drivers would cause a massive shift in the $38.5 billion attributed to ecommerce and would affect smaller retailers immensely.

That isn’t the only surprising result. Mobile will overtake desktop for the first time as the device of choice for online sales, but it turns out that conversions are falling short.

Poor mobile shopping experiences will only drive 34 percent of sales, a 19 percent gap between visits and purchases. That isn’t surprising considering the poor customer satisfaction across the majority of retail apps, most of which are failing their users.

“In terms of mobile, we specifically looked at smartphone versus desktop performance in terms of conversion (sales) rates, cart conversion rates, and average order value (AOV),” Tasker said. “We see several indications as to why smartphones lag. First, conversion on a desktop is 2.7X higher than a smartphone. Second, we actually see that once a consumer reaches the cart page on a desktop, nearly 30 percent of those carts turn into an actual order, compared to just 19 percent of carts on smartphones.”

While many conclusions can be drawn from this report, and others that point to issues with the mobile purchasing experience, two points of interest stand out immediately.

“There is an inherent challenge with getting customers to purchase on a smartphone: Therefore, mobile optimization and cross-device experiences should be considered,” Tasker said. “And customers are currently spending less on a smartphone, so retailers need to determine strategies to increase the amount of spend.”

Tasker believes that loyalty can play a big part here.

“We looked at the consumer behavior of price matching across retailers to determine how many products were bought at a lower price, the same price, or a higher price in the holiday season,” Tasker said. “While we do see the amount of products purchased at a lower price increases to 41 percent during the holiday season, we also see a consistent 25 percent of products being bought at a higher price. This indicates that customers are willing to pay a higher price for some products, and are likely doing so with a retailer or brand that they are comfortable with, assured of their product quality, and know it will arrive in time. Retailers must focus on strategies that build this loyal customer base — not only by offering the best price, but having a reputation for great quality and service.”

In addition to loyalty, customer service, and correct pricing, retailers need to be aware of several factors on their mobile websites and in native apps to ensure the online shoppers increase in number and not reduce, especially with so much at stake.

“Ultimately, the best apps are resonating with consumers, but a cohesive experience must be created to attract customers,” Tasker said. “We surveyed more than 1,000 U.S. consumers and found that small screens, being shown advertisements, unable to use the ‘pinch and zoom’ functionality within an app, slow performance, poor navigation, irrelevant push notifications, and difficult checkout and payment processes were common frustrations among all age groups, most of which can be enhanced through ongoing optimization efforts.”

The full report, which predicts the top gifts, best dates for deals, mobile shopping data, online sales by growth, and more — all of which has been underpinned by machine learning, processing billions of points of data flowing through the Adobe Marketing Cloud — is available today from the Adobe website.