of hands-on work to complete a full roster of roughly 20–27 reports.
Case Study · Beta · Marketing Operations / Automation
OneSentence in.FinishedReports out.
Rebuilding a complex weekly real estate reporting routine as a tested, restart-safe system that pulls from five live sources, regenerates charts, updates Canva, and leaves the team with a clear handoff.
20reports in a full roster week
5live sources per report
47maverage once stable
60+discrepancies resolved in testing
Beta status
The core workflow is working and validated in live runs. Reliability, throughput and team operating patterns continue to be refined as the automation encounters new production conditions.
The assignment
Automate the week without automating away judgment.
The report looked like a template update. In practice, it was a chain of decisions spread across five systems, two regenerated charts, variable-length tables, archival rules and source-specific exceptions.
The target was not simply to produce a report. It was to produce the right report without someone watching every step, and to make the workflow clear enough that another teammate could launch it from their own computer.
The operator's role shifts to setup, exceptions, review and delivery instead of building every page.
The reporting system
Five sources. One controlled output.
Data collection stays source-specific. Validation and rollover logic sit in the middle. Canva becomes the controlled delivery surface, not the place where the system has to remember what happened last week.
01FlexMLS
02Zillow
03BrokerBay
04ShowingTime
05Canva
Orchestration layerPull · calculate · rebuild · verifyRules live once. Every output is checked at the moment it is placed.
Weekly outputClient-ready Canva report
FlexMLS
Listing activity, portal views, saves, shares and roster context.
Zillow
Owner-dashboard traffic totals and exact daily chart data.
BrokerBay
Showing requests, totals and contact-history records.
ShowingTime
Target-market analysis by zip, price band and date window.
Canva
The client-facing report, rolled forward and updated page by page.
The redesign
The breakthrough was separating what the machine could make exact.
The original SOP had grown to roughly fifty pages because every new failure added another warning. The redesign divided the workflow by what each surface could do reliably.
Direct design operations
Copy, rename, archive, replace, position and verify elements by ID, removing stale coordinates, duplicate layers and page drift from the path.
Live systems and table cells
Use signed-in source systems for the data they own, and keep only the table work Canva cannot expose through direct design operations.
One source for rules
Measurements, formulas, folder references and guardrails live in the automation package, while the human SOP is generated from a maintained source.
Reliability by design
The incident log became a product requirement.
Live testing exposed the difference between “the automation produced a report” and “the automation produced the right report.” Each discrepancy was traced to a cause, resolved, and written back into the system.
Validate the evidence, not the interface
Every exported report is checked against the criteria printed inside the file. Chart values are read from the underlying data and reconciled to the visible total.
Make partial work unmistakable
A report stays marked work in progress until its final page passes validation. A restart resumes it instead of silently treating it as finished.
Isolate failure to one listing
Transient issues retry; blocked pages, listings or agents are logged and queued without sacrificing the rest of the week.
Turn every incident into a guardrail
The test log captures the symptom, root cause and permanent rule. Repeated mistakes are designed out of the system instead of merely warned against.
Stress-test evidence · 3–4 August 2026
Measured under live conditions, not a polished demo.
The twenty-two-hour test did two jobs at once: it produced real weekly reports and rebuilt the workflow whenever it met a new edge case. Once the system stopped changing, the last three reports landed in 36, 43 and 51 minutes.
What the test proved
Seven reports delivered, every client-facing figure checked against its source, and no number corrected after delivery.
2,461separate actions across the live run
Browser actions1,079
Direct Canva edits274
Reading, checks + file work1,108
36mreport 05
43mreport 06
51mreport 07
47mstable average
Relevant evidence · Attached report
Weekly Reports: Stress Test
The full account of the 22-hour live run: what shipped, what interrupted it, which blockers were closed, the measured production rate and the delivery cadence the evidence supported.
- Extent
- 7 pages
- Test window
- 3–4 August 2026
- Evidence
- 7 reports delivered · 2,461 actions measured
Click the preview to inspect the complete report.
Operational truth
Automation made the work predictable. It did not make the constraints disappear.
A full twenty-report week requires roughly 16–19 hours of machine time. The reports must include weekend activity, and the shared systems impose real session and access constraints.
That is different from the former 12–18 hours of hands-on production for a 20–27-report roster: the gain is reclaimed operator attention, controlled accuracy and predictable recovery, not a claim that complex work consumes zero time.
The outcome was a more honest service cadence: start after the weekend data closes, let the system run in planned sittings, and deliver Tuesday at noon, on time, during working hours, with a repeatable handoff.
New dependable delivery cadenceTuesday · 12:00 PM
The handoff
A system the team could run and understand.
The final deliverable was more than the automation. It included the operating model, installation path, evidence trail and communication needed to make the workflow usable beyond its builder.
Automation package
Self-contained workflow, specifications, browser guardrails and chart scripts.
SOP v2
A 38-page human-readable operating model generated from a controlled source.
Team run guide
Five pages covering setup, permissions, weekly use and expected timing.
Test-run ledger
Twenty-eight pages of live discrepancies, root causes and resolved rules.
Progress + stress reports
Decision-ready summaries of capability, measured throughput and constraints.
Delivery communication
A clear team brief aligning access, schedule and the Tuesday-noon cadence.
Outcome
From a fragile routine to an owned operating system.
The finished workflow can restart safely, skip completed work, isolate failures, keep its archive clean, validate each page as it lands and hand one clear summary back to the team.
The automation remains in beta: capable enough to produce live reports, with its incident ledger and stress-test cycle still actively shaping the production release.
Operational credentials, property-level data and internal links are intentionally omitted from this public case study. Results shown are drawn from the documented live stress test.