The Spreadsheet Problem
A distribution company with 45 staff was running operations on 12 Excel files. Sales data, inventory levels, delivery status, supplier lead times, each lived in a separate file, owned by a different person, updated at different times.
Every Monday, the operations manager spent two hours compiling the weekly report. By the time the report was ready, some of the data was already outdated.
This is a common story. Excel is a remarkable tool. It's also the wrong tool when four different people need the same data to make decisions in real time.
What a Custom Dashboard Actually Replaces
Before building anything, the team mapped out what decisions the dashboard needed to support:
- Which product lines are below minimum stock?
- Which deliveries are late, and by how much?
- Which suppliers are causing the most delays?
- What's the revenue run rate vs the target?
These were questions that took hours to answer. The goal was to make them answerable in seconds.
The Migration Process
Phase 1: Data audit (Week 1)
We inventoried all 12 spreadsheets. Which ones were actually used? Which ones fed into other ones? Where did the data originate, manual entry, exports from other systems, or both?
Result: 4 of the 12 spreadsheets were duplicates or rarely used. The real data sources were 3 main systems.
Phase 2: Data model design (Week 2)
Before building a dashboard, you need to agree on definitions. What counts as a "delayed delivery"? Is it delayed at dispatch, at expected delivery, or when the customer complains? These questions surface assumptions that different teams have held for years without realising they disagreed.
Phase 3: Build and connect (Weeks 3–5)
We built a data pipeline that pulled from the 3 source systems, standardised the formats, and fed a central database. The dashboard pulled from this central database, live, refreshed every 15 minutes.
Phase 4: Rollout (Week 6)
We ran the dashboard alongside the spreadsheets for two weeks. Any discrepancy triggered an investigation. Most were data quality issues in the source systems, surfaced for the first time because the data was now visible.
What Changed
The Monday report that took 2 hours now takes 10 minutes. The dashboard is live, so the data is current when the meeting starts, not two days old.
More importantly, the team asks different questions now. When data is hard to access, people stop asking. When it's instant, they ask more, and make better decisions.
Before and After: The Numbers
| Metric | Before (Excel) | After (Dashboard) | Improvement | | ------------------------------- | ------------------------- | ----------------------------- | ------------------ | | Weekly report preparation | 2 hours | 10 minutes | 92% time saved | | Data freshness | 24–48 hours old | Real-time (15-min refresh) | ~100x faster | | Error rate in reports | 3–5% (manual copy-paste) | <0.1% (automated pipeline) | 97% fewer errors | | Time to answer ad-hoc questions | 30–60 minutes | Under 30 seconds | 60–120x faster | | Staff hours on data management | 15 hours/week across team | 2 hours/week | 87% reduction | | Annual labour cost on reporting | ~$39,000 | ~$5,200 | $33,800/year saved |
What It Actually Cost
The project took 6 weeks and was significantly cheaper than the commercial BI tools the company had evaluated. Those tools required ongoing licences plus consultants to customise them, which brought the total cost of ownership close to the custom build anyway.
The difference: the custom dashboard connects directly to their specific systems, uses their terminology, and can be changed without a new consulting engagement.
Cost Comparison: Dashboard Options
| Option | Upfront Cost | Annual Cost | Timeline | Customisation | | ---------------------------------- | --------------- | -------------------------------------------- | ---------- | -------------------------------- | | Commercial BI (Tableau/Power BI) | $0–$5,000 setup | $15,000–$40,000/year (licences + consulting) | 8–16 weeks | Limited to platform capabilities | | Custom dashboard (traditional dev) | $40,000–$80,000 | $5,000–$10,000/year (maintenance) | 3–6 months | Unlimited | | Custom dashboard (AI-powered) | $8,000–$20,000 | $2,400–$6,000/year (hosting + maintenance) | 3–6 weeks | Unlimited |
For the 45-person company in this case study, the AI-powered custom build cost roughly $12,000 and runs on $100/month cloud hosting. The commercial BI tools they evaluated would have cost $25,000–$35,000 per year in licences alone. The custom dashboard paid for itself in under 5 months.
