Classic blended advice says “inbound is king.” That slogan protects answer-speed SLAs — until it becomes a blunt instrument. If every inbound ring freezes outbound for minutes, collections and sales books miss their own SLAs, callbacks age, and supervisors start fighting the dialer instead of coaching agents.
Pressure, not panic
A better model scores Inbound Pressure (IP) and Outbound Pressure (OP) continuously.
Inbound pressure rises with queue depth, oldest wait, arrival velocity, abandon risk, and occupancy — and falls when free capacity appears.
Outbound pressure should not equal “number of leads left.” A healthy book of 100,000 leads can contribute almost zero OP. What should raise OP: humans already waiting to bridge, overdue callbacks, campaigns behind KPI, predictive load that is about to miss service targets.
Hysteresis stops floor thrash
If IP is slightly higher than OP for one second, flipping share immediately creates oscillation: agents bounce between inbound and outbound, wrap times explode, and neither queue stabilizes. Production systems require a sustained gap (for example Δ ≥ 15 for ~10 seconds) and cap how far share can move per cycle.
Safe moments only
Share shifts should land on safe lifecycle points — call ended, disposition completed, recovery buffer done, agent available — never mid-sentence on a live conversation. Diverting a pre-bridge outbound answer to inbound when IP dominates can be valid; tearing down a connected call is not.
ASA should burn outbound capacity
When the oldest inbound wait exceeds target ASA, the unified pace brain should cut outbound dial capacity (VoxLink scales cuts roughly 20–70% by aggressiveness). That is how blending becomes SLA-aware instead of slogan-aware: inbound service level has a mathematical claim on dialer fuel.
Why this ranks as “best practice data”
Operators in South Africa, India, and global BPOs share the same question: *Can we prove the platform protected customers in queue without silently destroying outbound revenue?* Pressure scores, share timelines, divert events, and ASA-linked pace cuts are the evidence pack. They turn blending from folklore into auditable operations intelligence.