The Coffee Shop Mystery: How a Sales Dip Sparked a Data Revolution

The rain of a Seattle Tuesday in October 2024 didn’t just wet the sidewalks of Pike Place Market—it seemed to seep through the floorboards of Bean There’s flagship store, where Maya Alvarez stood behind the espresso bar, staring at the tablet in her hand as if it had personally betrayed her. The tablet’s screen glowed with the same red number she’d been avoiding for three weeks: -8% year-over-year sales decline for the second consecutive quarter. Five years ago, she’d opened this 10-square-meter store with a secondhand espresso machine and a dream of bringing third-wave coffee to the Pacific Northwest, lining up out the door every morning by 7 a.m. for her signature lavender oat milk latte. Now, at 2 p.m. on a rainy Tuesday, the shop held only three customers: a tourist taking photos of the “Seattle’s Best Coffee” sign (she’d politely corrected him twice already), a remote worker camped out in the corner with a single cold brew, and a regular who’d been coming every Wednesday for three years—today, he’d ordered a tea, the first non-coffee drink she’d ever seen him buy.

Maya’s fingers, stained with the faint brown ring of a half-finished flat white she’d made an hour ago, tapped the screen again, as if the number might change if she refreshed it fast enough. She’d built Bean There on gut instinct: she’d known her regulars by name, knew which neighborhood preferred darker roasts, knew that a free cookie with a large latte on rainy days would boost sales by 12% (she’d tested it once, on a whim, in 2020). She’d expanded from that single store to 12 locations across Seattle and Portland, with 120 employees, a local roasting facility, and a cult following for her seasonal pumpkin spice latte that sold out every October. But for the last six months, something had felt off. The lines were shorter, the regulars fewer, the online reviews mentioning “long wait times” and “overpriced” more often than “best latte in Seattle.”

The bell above the door jingled, and Raj Mehta, her head of operations, walked in, his neon-yellow rain jacket dripping onto the hardwood floor. Raj was a former McDonald’s ops manager who’d joined Bean There three years ago, when Maya was struggling to scale her supply chain to keep up with demand. He was practical to a fault, trusted spreadsheets more than he trusted Maya’s “coffee shop vibes,” and had been nagging her for months to cut costs. “I just got the inventory report,” he said, slumping into the chair behind the counter, not bothering with small talk. “We’ve got 2,000 pounds of pumpkin syrup left over from the fall launch. That’s $18,000 in dead stock. And the Q4 forecast says we’re going to have 30% more unsold inventory if we don’t cut the menu.”

Maya sighed, setting her tablet down. “I know, I know. The pumpkin latte didn’t perform. But the seasonal drinks are what bring people in. If we cut them, we’ll lose even more customers.”

“Maybe the problem isn’t the menu,” a quiet voice cut in from the doorway. Lila Chen, Bean There’s newly hired data analyst, stood there holding a stack of printed spreadsheets, her glasses slipping down her nose. She was 24, fresh out of the University of Washington with a master’s in statistics, hired two months earlier at the insistence of Maya’s cousin, who’d convinced her that “even coffee shops need data people these days.” Maya had hired her mostly to make pretty charts for her investor updates, assuming a data analyst would just tell her what she already knew.

“What do you mean?” Maya asked, leaning against the counter.

Lila walked over, spreading the spreadsheets out on the table. “I spent the last week pulling data from our POS system, delivery platforms, and membership database. The 8% sales decline you’re seeing is the *revenue* number, which includes the 15% commission we pay to Uber Eats and DoorDash. If you strip that out, actual sales only declined 3%.” She tapped a cell on the spreadsheet. “But the bigger issue is foot traffic: we’re down 12% across all stores, and our repeat customer rate is down 18% from last year. New customer acquisition is up 7%, but those new customers spend 40% less per visit than our regulars. So we’re spending more to get customers who don’t stick around, and our loyal customers are leaving.”

Raj frowned, leaning over the spreadsheet. “How did you get those numbers? I’ve been pulling sales reports every week, and I never saw that repeat customer number.”

“The data’s all over the place,” Lila said, tapping three different spreadsheets. “Our POS system exports sales data with dates in MM/DD/YYYY format, the delivery platforms use DD-MM-YYYY, and the membership system uses Unix timestamps. The weekly sales report you get, Raj, is pulled by an admin who manually copies and pastes data from all three sources into one Excel sheet. She’s been doing it wrong for three months—she’s been matching delivery orders to the wrong dates, so the sales numbers for the last quarter of 2023 are inflated by 5%.”

Maya’s face flushed. She’d been making decisions based on those wrong reports for months. She’d cut the marketing budget for the loyalty program in August, assuming it was a waste of money, because the reports showed repeat customer rates were flat. Now she realized she’d cut the program that was keeping her regulars coming back.

“So what do you suggest?” Raj asked, crossing his arms. “We just… ask people why they stopped coming?”

“No,” Lila said, pushing her glasses up her nose. “We need to do structured analysis. First, we align all our data into one source, so everyone’s looking at the same numbers. Then we explore the data to find patterns: which customer segments are leaving? Which stores are hit hardest? Is it a product issue, a pricing issue, a service issue? Then we test hypotheses, not guesses. For example, your theory that the menu is too long—we can test that by looking at sell-through rates for each item, not just by guessing which ones to cut.”

Maya looked at the two spreadsheets in front of her, then at the red -8% on her tablet. For five years, she’d run Bean There on instinct, and it had worked—until it didn’t. She thought of the regular who’d just ordered tea, the one who used to bring her homemade scones every Wednesday. She’d never thought to ask him why he’d stopped drinking coffee. She’d never thought to ask *any* of her regulars why they’d stopped coming. She’d just assumed it was the new Starbucks on the corner, or the recession, or the fact that people were drinking less coffee these days.

“You have one month,” she said finally, looking at Lila. “If you can figure out exactly why our sales are down, and give us a concrete plan to fix it, we’ll implement it. If you can’t, you’ll spend a week working the floor at the flagship store, learning how to pull a shot before you try to tell us how to run our business.”

Lila nodded, not flinching. Raj looked skeptical, but he didn’t argue. He knew as well as Maya did that they were flying blind right now, making decisions based on half-remembered conversations and misaligned spreadsheets.

---

As the three of them left the store an hour later, the rain had let up, leaving the streets glistening under the gray Seattle sky. Maya walked to her car, her mind racing. She’d always prided herself on being a “founder who knows her customers,” but the last hour had made her realize she didn’t know anything at all. She had 120 employees, a roasting facility, a supply chain that stretched from Colombia to her store shelves, and she was making decisions based on the same gut instinct she’d used when she had one 10-square-meter store and 10 customers a day.

That’s the core tension this book is built to resolve.

If you’re reading this, you’re likely an intermediate business practitioner: you know your way around Excel, you’ve built pivot tables, used VLOOKUP (or its modern successor XLOOKUP) to merge datasets, created basic charts to show your boss why Q3 missed target. You know what mean, median, and sta