"Make everything real-time" is a budget trap. The skill isn't streaming more — it's knowing which KPIs change an intraday decision and paying for low latency only there. Here's a practical shortlist of the operational Genesys Cloud KPIs worth streaming, why each one drives action, and where the line is.
The test: does this number change what someone does in the next 15 minutes?
If yes, it's operational — stream it. If it only informs a weekly review, an interval-based read is fine. Apply that test and the list almost writes itself.
The operational shortlist
1. Service level & ASA (by queue/skill)
The classic. When SL dips or ASA climbs right now, a supervisor opens overflow, pulls agents from offline work, or re-skills. A 20-minute-old SL is a post-mortem; a live one is a lever.
2. Queue depth & longest wait
Leading indicators that move before SL breaches. Longest-waiting interaction is the single most actionable number on a wallboard — it tells you who's about to abandon.
3. Agent state & availability
How many agents are on-queue, in aux, in after-call work, by skill. This is the supply side of the intraday equation; without it, you're flying the demand side blind.
4. Real-time adherence (RTA)
Is the agent doing what the schedule says now? Feeds both coaching and the WFM intraday loop. Seconds-fresh by nature.
5. Abandonment rate (live)
Rising abandons are a fire alarm. Live abandonment, paired with longest-wait, tells you whether to act before SL formally breaches.
6. Offered/handled volume vs forecast
Intraday actuals against the plan. When volume runs hot versus forecast, planners reshape breaks and overtime — but only if they can see the gap in near real time.
7. Occupancy
The pressure gauge. Sustained high occupancy predicts burnout and shrinkage spikes; it's the metric that justifies pulling people off the phones.
8. Callbacks / queued tasks
For estates using callbacks or digital tasks, the live backlog is as operational as the voice queue — and just as capable of silently growing.
What does not need to be real-time
Be honest about this — it's where the savings are:
| Stream it (operational) | Interval is fine (analytical) |
|---|---|
| Service level / ASA | Daily/weekly SL trends |
| Queue depth, longest wait | Historical volume seasonality |
| Agent state, RTA | Monthly adherence scorecards |
| Live abandonment | QA/CSAT trend analysis |
| Volume vs forecast (intraday) | Long-run forecast accuracy |
The architecture point
The reason teams can't stream the operational list is usually that each metric is trapped behind a polling job or a vendor dashboard. Compute them once from the Genesys event stream and they become available everywhere at once — wallboards, WFM, CRM and BI — on the same heartbeat. That's what the reactive engine does, and why a live dashboard stops disagreeing with the WFM feed.
How to start
- List your KPIs and apply the 15-minute test.
- Tag each operational or analytical.
- Put the operational set on an event-driven path; leave the rest on interval reads.
- Make sure every consumer reads the same computed operational KPI — no re-derivation.
FAQ
Isn't more real-time always better? No. Low latency has a cost, and most analytical metrics gain nothing from it. Spend the latency budget where it changes a decision.
How fresh can these be? On a streaming path, operational KPIs reach consumers in under a second; the floor is how fast Genesys emits the underlying event.
Where do I start if I can only do a few? Service level, longest wait and agent state. Those three cover the demand, the early warning and the supply — enough to run the floor.
Want help separating your operational KPIs from the analytical ones? Book a discovery call — we'll do exactly that before recommending any architecture.