David dreaded the last Friday of every month. He'd arrive at the office at 7 AM, coffee in hand, knowing exactly what the next ten hours would look like. Five browser tabs open to different systems. A PowerPoint deck with forty-seven slides. And a spreadsheet that linked to three other spreadsheets, two of which had broken formulas he would need to rebuild by hand.
He was the finance director at a regional manufacturing firm. His monthly board report was essential. It was also a monster. The process consumed his entire last Friday of the month — forty hours of labor spread across one very long day, two weekends of prep, and the occasional panic call to the warehouse team when their numbers didn't match his.
Nobody questioned the process. The report had existed for six years. It predated David. It predated half the leadership team. Like many manual rituals, it was accepted as a fact of life — expensive, painful, and inevitable. Until an FDE asked a simple question: what would it take to automate this monthly reporting?
The Friday Night Ritual
The report followed a choreography David could perform in his sleep, which was fortunate because he often did. He'd start by exporting Shopify sales data into a CSV. Then he'd pull Stripe transactions into a different CSV. The warehouse API gave him inventory numbers, but only if he remembered to refresh the token first. The CRM held customer acquisition costs in a dashboard that didn't export cleanly, so he screenshotted the totals and typed them in by hand. And then there was the Google Sheet — the master spreadsheet that three departments updated independently, sometimes on time, sometimes not.
By noon, he'd have all five sources in one place. By 3 PM, he'd have reconciled the discrepancies. The Stripe total never matched the Shopify total because returns flowed through different systems. The warehouse count was always two days behind the sales count. The CRM attribution model used last-click while David's report used first-click, a difference nobody had explained to him in six years.
The PowerPoint deck itself was a work of art. Thirty slides of charts, five slides of commentary, and twelve slides that nobody had ever asked a question about. David included them because his predecessor had included them, and his predecessor's predecessor had included them. By 6 PM, he'd email the deck to the leadership team. By 7 PM, he'd be home, exhausted, already dreading next month.
The Five-System Problem
Our FDE didn't start with code. She started with observation. She sat with David on a Thursday — not a Friday, because Fridays were too frantic for learning . and watched him work. She noticed the pattern. Five systems. One report. Zero integrations. The human body was doing what software should have been doing years ago.
Systems themselves weren't the problem. Shopify, Stripe, the warehouse API, the CRM, and Google Sheets each did their job well. The problem was the space between them — the glue that consisted entirely of David's patience and typing speed.
The problem wasn't a technology failure. It was an integration failure, plain and simple: five systems that refused to talk.
A typical mid-size firm loses fifteen to twenty hours per week to manual reporting across all departments. That's nearly a thousand hours annually — half a full-time employee . spent copying numbers between systems that could talk to each other if someone built the connection. The cost isn't just time. It's the lag between event and insight. By the time David's report reached the leadership team, some of the data was a week old. Decisions were being made on history, not current reality.
Week One: Mapping
The FDE's first week was entirely non-technical. She interviewed David, the warehouse manager, and the marketing lead. She asked the same question repeatedly: what decisions does this report actually drive? The answers were revealing.
Of the forty-seven slides, only nine were regularly discussed in the board meeting. The rest were insurance — included in case someone asked a question that hadn't come up in three years. The warehouse variance slide, the customer cohort slide, the regional breakdown slide . all beautifully crafted, all rarely referenced. David kept them because removing them felt riskier than maintaining them.
The FDE proposed a radical scope cut. The new dashboard would cover the nine essential metrics first. The historical slides would be archived, not deleted — available if needed, but not maintained. David was skeptical. His predecessor had warned him never to drop a metric. But the FDE showed him the math: twenty hours of manual work to produce thirty-eight slides that nobody read. Even David had to admit that wasn't sustainable.
This mapping phase is where most automation projects fail. Teams try to replicate the entire manual process in software, which means they automate the waste along with the value. The FDE's discipline here — understanding what mattered before building anything . saved weeks of scope creep. If you want to see what happens without that discipline, read our story on when scope destroys a project timeline.
Week Two: The Pipeline
With scope defined, the build was straightforward. A Python script connected to all five sources using their APIs — except Google Sheets, which required a small workaround through the Sheets API. The script ran every morning at 6 AM, pulling fresh data into a Postgres database. A lightweight dashboard built with Streamlit displayed the nine essential metrics in real time.
The technical implementation took four days. The Shopify and Stripe integrations were standard. The warehouse API was older and crankier, but it responded to polite HTTP requests. The CRM export was the trickiest part — the vendor didn't expose the attribution model through the API, so the FDE reverse-engineered it from the dashboard and validated the numbers against David's manual calculations. Getting attribution right matters more than most teams admit; a wrong model makes every downstream metric meaningless.
That first automated report generated itself on a Tuesday at 6:17 AM. David opened his email and found a link to the dashboard. The numbers looked right. He checked them against his spreadsheet. They matched within rounding error. He stared at the screen for a long moment, then forwarded the link to the leadership team with a single sentence: "The monthly report is live."
Tuesday marked the end of the Friday ritual. David never spent another last Friday of the month copy-pasting into PowerPoint. The forty-hour report had become a twenty-minute review — ten minutes to scan the dashboard, ten minutes to write commentary.
The Human Review Gate
The FDE insisted on one feature that seemed unnecessary at first: a human review gate. Before the dashboard was considered final, David had to log in and click a button confirming that the numbers looked reasonable. This twenty-minute sanity check caught anomalies three times in the first six months.
Once, the warehouse API returned duplicate records after a weekend maintenance window. The pipeline loaded them faithfully, and the inventory numbers doubled. David caught it during his Monday morning review. He flagged the issue, the FDE added deduplication logic, and the problem never recurred. Without the review gate, that error would have reached the leadership team.
Fully unattended automation is a fantasy. Even the best pipelines encounter edge cases — API changes, data quality issues, timing mismatches. A human review gate doesn't slow you down. It catches the mistakes that code can't predict. The twenty minutes David spent reviewing were a fraction of the forty hours he used to spend building, and they gave him confidence that the numbers were right.
This pattern of building something small that becomes permanent is more common than you'd think. Many of the best FDE engagements start exactly this way. For a deeper look at how prototypes evolve into products, read about the prototype that became permanent.
What the Savings Funded
The real payoff wasn't the forty hours saved. It was what those forty hours bought. David's company reinvested his time and the FDE's freed capacity into three subsequent builds. A real-time inventory alert system. A customer churn predictor using historical sales data. And an automated invoice reconciliation tool that saved the bookkeeping team another twenty hours monthly. The pattern is common: one reliable pipeline makes leadership hungry for the next.
This engagement, which cost roughly $18K for the reporting automation alone, funded itself in the first quarter through reclaimed labor. But the bigger return was strategic. Once leadership saw that automation could deliver reliable data without the Friday panic, they became willing to fund more ambitious projects. The reporting pipeline was the gateway drug.
That pattern defines the best FDE engagements. One unglamorous automation — often a report, sometimes a data sync, occasionally a simple workflow . proves that embedded engineering can deliver measurable value quickly. That proof builds trust. Trust builds budget. Budget builds the next thing.
If you want the full arc of how these engagements unfold, read a full FDE engagement story covering three builds, one near-disaster, and the measurable wins that justified the program. David's report was just the beginning. The dashboard that changed his Fridays was the foundation for everything that followed. For another example of boring automation with outsized returns, see how invoice processing automation saved a bookkeeping team twenty hours a month.
David still checks the dashboard every Monday morning. It takes him twenty minutes. Sometimes he finishes his coffee before he's done. The rest of the day is his. And the last Friday of every month? He takes it off.