Building a Shift Scheduling and Attendance System: Rostering Engine, Geofenced Clock-In, and Overtime Calculation

1. One Roster Change, Five People Recalculating Payroll

A three-store beverage chain employs 42 people: 12 salaried, 30 hourly. The manager spends 2.5 hours every Sunday building next week's roster in Excel and handles about nine last-minute swaps a week. At month end HR reconciles three sources - the punch-clock CSV, the Excel roster, and leave requests in a LINE group - burning 11 hours. From January 2026 Taiwan's minimum wage is NT$29,500 monthly and NT$196 hourly; under-counting 10 minutes per person per day across 30 staff is roughly NT$16,000 of disputed pay a month. A scheduling system is not about drawing a grid. It is about wiring rules, evidence, and payroll into one line.

2. Good Fit vs Poor Fit

Custom build makes sense when:

25+ staff with rotating shifts or cross-store coverage
Hourly and salaried staff mixed, with different overtime rules
Scheduling depends on skills or certifications (barista-qualified, forklift licence)
Statutory staffing ratios apply (clinics, kindergartens, security)
Attendance must feed payroll, labour insurance brackets, or ERP

Use off-the-shelf tools instead when:

Under 15 staff on fixed day shifts; a spreadsheet plus a punch app is enough
Only one shift pattern and no cross-location dispatching
A reorganisation is coming within three months and rules are still fluid
Nobody owns skill tags and shift parameters

3. Alternatives Matrix

Option	Annual cost NT$	Pros	Cons
Excel roster + punch clock	3,000-15,000	Zero learning curve	No compliance guardrails, manual reconciliation
Taiwan HR SaaS (NUEIP, femas)	30,000-120,000	Vendor maintains legal updates, payroll included	Fixed rostering logic; skill and staffing-ratio rules do not fit
Global scheduling SaaS (Deputy, When I Work)	50,000-200,000	Mature rostering UX	No grasp of Taiwan rest-day or labour-insurance rules
Custom build (ScriptWalker)	Year one 280,000-450,000	Your rules become the system rules	Higher investment, 10-14 weeks

Decision formula: (monthly rostering hours + reconciliation hours) x manager hourly rate + retroactive pay + swap-coordination cost, multiplied by 12. If that exceeds 60% of year-one custom cost, build it. For the chain above it was NT$380,000 against NT$320,000.

4. Delivery Breakdown (10-14 weeks)

Weeks 1-2 | Rules and compliance audit: Interview managers and floor staff, turn invisible rules into parameters, and map legal limits on consecutive days, rest between shifts, and daily hours. Tool: Notion. Deliverables: shift dictionary, overtime rules matrix.
Weeks 3-4 | Interface design: A drag-and-drop weekly roster in the back office, a three-second clock-in on mobile. Tool: Figma. Deliverables: clickable prototype, swap approval flow.
Weeks 5-9 | Rostering engine and clock-in: A rules engine covering demand coverage, skill tags, compliance checks and a conflict lock; GPS geofenced clock-in with QR fallback; Flutter on mobile, Laravel on the backend, jobs following the official Laravel Task Scheduling documentation. Deliverables: roster API, concurrent-swap test report.
Weeks 10-11 | Payroll pre-calculation: Convert attendance into overtime tiers, rest-day premiums and lateness deductions, recorded to the minute and retained five years per Article 30 of the Labor Standards Act. Deliverables: pre-calculation report, labour-insurance export files.
Weeks 12-14 | Pilot and handover: One store runs in parallel for two weeks, computing payroll both ways, monitored with UptimeRobot and Sentry. Deliverables: training recordings, acceptance checklist.

5. Real Cost Breakdown

ScriptWalker Shift Scheduling and Attendance System development starts at NT$150,000. One-off items:

Rules and compliance audit: 20,000-40,000
UI/UX (roster board + mobile clock-in): 25,000-50,000
Rostering engine and compliance checks: 60,000-110,000
Clock-in module (GPS / QR / NFC): 30,000-60,000
Overtime and payroll pre-calculation: 35,000-70,000
Role-based permissions and reporting: 20,000-40,000

Hidden costs most buyers forget:

App store listing: Apple US$99/year, Google Play US$25 one-off
Swap and lateness SMS: about NT$0.8 each, roughly NT$2,400 for 3,000 a month
Hosting plus five-year attendance archiving: NT$1,500-5,000/month
Annual maintenance and regulatory updates: 15-20% of build cost
Internal rollout labour: training and parallel reconciliation, 40-60 hours

Three budget tiers:

Entry NT$150,000-250,000: single store, one shift pattern, web clock-in
Standard NT$280,000-450,000: multi-store and multi-shift, Flutter location clock-in, overtime pre-calculation
Advanced NT$500,000-850,000: certification-aware scheduling, demand-forecast rosters, ERP integration

6. Client Expectation vs Implementation Reality

Clients assume	What actually happens
The system auto-generates a perfect roster	It produces a 70-80% draft; the value is blocking illegal and conflicting shifts first
App clock-in ends time fraud	Buddy punching and disabled location persist; you need geofence radius, device binding and anomaly reports together
Full rollout in two weeks	Parallel running takes at least two weeks; keep the old process until the first payroll close
Old punch data imports cleanly	Most exports carry no shift or leave type; historical hours are reference only, never a payroll basis

7. Common Traps and How to Avoid Them

Rules live only in people's heads: unwritten norms surface at go-live. Fix: capture every rule as a toggleable parameter in week one.
One overtime formula only: hourly and salaried staff derive their hourly base differently. Fix: run both rule sets on the same day and print the formula used.
Records kept only to the hour: breaching Article 30(5) carries NT$90,000-450,000 penalties under Article 79. Fix: store timestamps to the second, display to the minute.
No audit trail on edits: nobody can prove who amended a punch. Fix: write every change to an audit log with editor, original value and reason.
Swaps bypass approval: two staff trade and one works eight days straight. Fix: run compliance checks at submission and block with a stated reason.

8. Success Metrics and the First 90 Days

Day 30: clock-in adoption at 95%+, manually corrected punches under 5%, rostering time down from 150 minutes to under 45. This phase is only about data hygiene.
Day 60: payroll reconciliation down from 11 hours to under 3, average swap handling under 10 minutes, and a clear picture of what the compliance engine blocks so you can tune parameters.
Day 90: labour cost as a share of revenue split by store and time band, overtime hours down 10-15%, and audit-ready attendance reports.

9. Decision Checklist

☐ More than 25 employees?
☐ Both hourly and salaried staff?
☐ Three or more unplanned swaps per week?
☐ Weekly rostering over 90 minutes?
☐ Month-end reconciliation over 5 hours?
☐ Retroactive overtime or a dispute in the past year?
☐ Attendance already recorded to the minute?
☐ Cross-store dispatching needed?
☐ Certifications or staffing ratios affect scheduling?
☐ Attendance must feed payroll or ERP?
☐ One internal owner can sign off shift rules?

Six or more ticked and a custom build usually pays back. Three or fewer, run on SaaS for a year.

10. FAQ

Can rostering be fully automatic?

It can produce a draft, but not a launch-ready roster. The engine generates 70-80% from demand, skill tags and legal limits; the manager tunes the human factors. Automation usually fails on inaccurate demand forecasts, not on the algorithm.

Is mobile clock-in legally valid in Taiwan?

Yes. The Labor Standards Act accepts computerised attendance systems. What matters is verifiability, daily records to the minute and five-year retention. In practice: device binding, location verification, an audit trail on edits, and letting staff view their own records.

We already use HR SaaS - do we still need a custom build?

If the SaaS payroll module works, customise scheduling and leave payroll in the SaaS, pushing attendance across by API. That keeps cost around NT$180,000-300,000 and avoids rebuilding compliant components.

What are the ongoing annual costs?

Budget 15-20% of build cost for maintenance and regulatory updates, plus NT$18,000-60,000 a year for hosting and backup and around NT$30,000 a year for SMS. Standard-tier ownership runs NT$80,000-130,000 from year two.

11. Next Step

If your rosters live in Excel and payroll is reconciled by hand, multiply last week's rostering and reconciliation hours by the manager hourly rate. The number is usually larger than expected. We offer a 45-minute free assessment with a shift-dictionary template and a budget range sized to your headcount. No pitch, no lock-in.