Two approaches to dashboards: real-time and scheduled.
Real-time dashboard updates every minute or hour. Operational dashboard for ops team watching uptime. Ops sees error rate spike in real-time. Ops responds immediately. Cost: expensive infrastructure to compute metrics constantly.
Scheduled dashboard updates once daily at 6am. Executive dashboard. CEO sees revenue at 6am. Sufficient for decisions made daily. Cost: cheap, simple infrastructure.
Choose based on decision frequency and cost tolerance.
When real-time is needed
Operations dashboards
System can go down. Ops needs to know immediately. Delay of one hour is too late. Build real-time.
Incident response dashboards
Problem happens. On-call engineer needs to see current status. Stale data is dangerous. Build real-time.
Trading dashboards
Seconds matter. Data that is five minutes old is useless. Build real-time.
Customer-facing dashboards
Customers want to see current system status. Showing status from six hours ago is misleading. Build real-time.
When scheduled is fine
Executive dashboards
CEO checks once daily. Data from 6am is fine. Decision is made once daily. Cheaper to compute once daily.
Marketing dashboards
Campaign data refreshes daily. One day delay is not critical. Team makes decisions once daily.
Financial dashboards
Month-end data is final. Daily updates are not needed. Weekly or monthly is fine.
Sales dashboards
Sales manager checks quota daily. Data from yesterday is fine. Decision is made based on recent data, not real-time.
Cost comparison
Real-time dashboard: infrastructure cost one hundred thousand dollars per year. Maintenance cost fifty thousand per year. Total: one hundred fifty thousand annually.
Scheduled dashboard: infrastructure cost five thousand dollars per year. Maintenance cost five thousand per year. Total: ten thousand annually.
Difference: one hundred forty thousand dollars per year. Huge. Only use real-time if you truly need it.
Real example: three dashboards at different refresh rates
Executive dashboard (scheduled daily at 6am)
Revenue, churn, acquisition cost, expansion revenue. Updated once daily. Cost: low.
Sales dashboard (scheduled hourly)
Pipeline, quota, forecast. Updated hourly at top of each hour. Ops team checks at start and end of day. Cost: moderate.
Operational dashboard (real-time)
Uptime, error rate, response time. Updated every minute. Ops team watches continuously. On-call engineer uses for incident response. Cost: high.
Strategy
Three dashboards with three different refresh rates. Each is optimized for use case. Do not make all dashboards real-time (too expensive). Do not make all scheduled (miss real-time problems).
Hybrid approach: real-time alerts plus scheduled dashboards
Dashboard updates once daily (cheap). Alert fires if metric crosses threshold (expensive but precise). Example: revenue dashboard updates daily. But alert fires if daily revenue drops fifty percent (real-time threshold check). Alert is rare but real-time when needed. Saves cost while maintaining responsiveness.
Frequently asked questions
How do we know if we need real-time versus scheduled?
Ask: what is cost of one-hour delay. If cost is one thousand dollars per hour, real-time is justified. If cost is zero (decision is not made until tomorrow anyway), scheduled is fine. Example: ops dashboard monitors checkout. Checkout is down one hour. One hundred customers cannot buy. One hundred transactions at fifty dollars each is five thousand dollars lost. Real-time is justified. Marketing dashboard shows campaign performance. Campaign is running for one month. One-day delay in data is not critical. Scheduled is fine. With WEMASY's website builder, you can set this up directly on your website.
What is minimum real-time update frequency?
One minute is practical minimum. Updates faster than one minute create noise (data is constantly changing). Alerts fire constantly. Team is in constant response mode. One-minute updates are suitable for operational dashboards. Five-minute updates are more practical (less noise, less infrastructure). Balance responsiveness and noise.
How do we handle real-time dashboards when source data is not real-time?
Use near-real-time. Real-time dashboard shows data from five minutes ago (latest available). Better to show slightly stale data than no data. Example: database updates every five minutes. Dashboard shows five-minute-old data as current. Label it last updated five minutes ago. Transparency prevents confusion.
Should we have separate real-time and scheduled versions of same dashboard?
No. One dashboard with two refresh options. User chooses: refresh every minute (real-time mode) or refresh once daily (scheduled mode). Different users have different needs. Same dashboard, flexible refresh. Simplifies maintenance.
How do we prevent real-time dashboards from creating alert fatigue?
Set high thresholds. Real-time alert should fire only when metric is significantly wrong. Threshold: error rate above two percent. Not error rate above zero point two percent. High threshold equals fewer alerts equals less fatigue. Alert should signal real problem, not normal variation.
What is the best way to communicate when a dashboard is real-time versus scheduled?
Label clearly. Real-time dashboard header: Updates every minute. Scheduled dashboard header: Updated daily at 6am. Do not assume user knows refresh rate. Make it explicit. User should always know: how fresh is this data.