How to Build a Custom BI Dashboard: From Data Integration and Real-Time Visualization to Access Control (With Real Costs and a 90-Day Roadmap)

"Every month-end I spend three days exporting data from five systems into Excel and manually pasting a report for the boss." That is the SMB pain we hear most. A 40-person trading company spends 20+ hours a month just consolidating orders, inventory, payments, and ad data, and the numbers are stale the next day. A custom BI dashboard automates this once and for all: data flows in by itself, charts update themselves, access controls itself.

When It Fits vs. When It Doesn't

A custom dashboard fits when:

Data is spread across 3+ systems (ERP, e-commerce, payments, ads, CRM) and you need one screen for the whole picture.
You manually assemble reports weekly/monthly, high-repetition and error-prone.
Different departments need role-specific views of the same data.
Leadership wants real-time numbers, not month-end-only.

Skip it (for now) when:

Data lives in one system whose built-in reports suffice.
Data volume is small and viewed once or twice a month; Excel pivot tables are cheaper.
The underlying data is messy and unmaintained; fix the source first, do not rush pretty charts.

Alternatives Matrix

Option	Pros	Cons	Cost band
Custom dashboard (Laravel + chart library)	Fits your process exactly, flexible permissions, extensible long-term	High upfront build, needs maintenance	From NT$120,000
BI SaaS (Looker Studio / Power BI / Metabase)	Fast start, many chart templates	Limited complex permissions and customization, data-residency concerns	Free to NT$300-800/user/mo
Excel / Google Sheets + auto-import	Near-zero cost, everyone knows it	Manual upkeep, error-prone, not real-time	Near free
System built-in reports	No extra build	Cannot integrate across systems, fixed format	Included in existing system

Full Process Breakdown (Tools and Deliverables)

Week 1 | Requirements and metric definition: interview departments, define 8-15 KPIs, map data sources. Deliverables: a metric dictionary (Notion) + data-flow diagram (Figma/FigJam).
Weeks 2-3 | Data integration and cleaning: pull each system's data via API/schedule into a database (PostgreSQL/MySQL); handle formats, dedupe, align time. Deliverables: ETL schedule + table schema.
Weeks 4-5 | Dashboard development: Laravel API backend + front-end charts (Chart.js/ECharts) with filters, drill-down, export. Deliverable: an operable dashboard beta.
Week 6 | Permissions and acceptance: set role layers (boss sees all, departments see their own), launch training. Deliverables: production launch + user manual.

Full Cost Breakdown

Development: base version NT$120,000-180,000 (3-5 sources, 10-15 charts).
Data-source API licensing: some systems charge extra for API access (NT$0-30,000/yr each).
Hosting and database: VPS or cloud roughly NT$800-3,000/mo.
Scheduling/auto-refresh: real-time needs a background worker, raising hosting cost slightly.
Ongoing maintenance: for upstream API changes and new metrics, a monthly retainer of NT$8,000-20,000 is advisable.

Hidden costs most often show up in data-source API licensing and re-connecting after upstream changes; inventory how each source is obtained before signing.

Reality vs. Client Imagination

Imagination: connecting data is fast; charting is the hard part. Reality: 80% of the time goes to data cleaning and alignment; charting is 20%.
Imagination: build once, done forever. Reality: when upstream changes, the dashboard must follow; maintenance is the norm.
Imagination: more metrics is better. Reality: 15 charts nobody reads are useless; 5 charts that change decisions matter.

Common Traps and How to Avoid Them

Trap: charting before defining metrics → hold a requirements meeting to define KPIs first.
Trap: a source has no API, only manual export → confirm each system's data-retrieval method before starting.
Trap: no permission plan, everyone sees company-wide numbers → define roles and data scope before launch.
Trap: real-time need not stated, then a background refresh is demanded later → confirm how real-time in the requirements phase.
Trap: the underlying data is wrong → dashboards only amplify errors; clean the source before visualizing.

Success Metrics + 90-Day Roadmap

Day 30: did reporting hours drop from 20+ to under 2? Are departments actually using it?
Day 60: based on usage feedback, cut charts nobody reads, add the metrics truly needed.
Day 90: check whether any decision changed because of a number seen; that is the ROI proof.

Decision Checklist

☐ Is my data spread across 3+ systems?
☐ Does monthly reporting take over 8 hours?
☐ Can I clearly list 8-15 metrics to track?
☐ Does every source have an API or export path?
☐ Do I need cross-department permission layers?
☐ Do I need real-time, or is daily refresh enough?
☐ Is the underlying data clean and maintained?
☐ Do I have budget for monthly maintenance?
☐ Will someone own reading this dashboard and act on it?
☐ Can I tell apart nice-to-see vs decision-changing metrics?

FAQ

Why not just use Power BI or Looker Studio?

If your permissions are simple, sources few, and cloud data acceptable, BI SaaS is the fastest choice. Custom pays off when you need complex permission layers, deep customization, data kept on your own servers, or embedding into an existing system.

How long does a dashboard take?

A base version (3-5 sources, 10-15 charts) takes about 5-6 weeks. Time goes mostly to data integration and cleaning, not charting.

Will my data leak?

A custom solution keeps data on your own servers or private cloud with fully self-controlled permissions, a key advantage over some BI SaaS.

Call to Action

Want to know if your data is worth a dashboard? We offer a free 30-minute consultation to inventory your sources and metrics. ScriptWalker's custom BI dashboard service starts at NT$120,000.