The Problem With External SLAs Without Internal OLAs

When I joined AudienceView in 2018, the support organisation had customer-facing SLAs but no mechanism for holding internal teams to the commitments that made those SLAs achievable. A P2 issue requiring a hosting team intervention had no internal response target. Engineering escalations had no defined ownership handoff. Platform queries could sit unactioned for days without anyone formally accountable.

The gap was visible in the data. P2 response times exceeded 20 days — not because support agents were slow, but because the internal dependencies had no formal structure. Fixing the customer-facing number meant fixing the internal accountability layer first.

"An SLA without an OLA is a promise made without a plan. You can commit to a customer response time without an internal agreement — you just can’t keep it consistently."

Designing the Framework

01
Define Severity Tiers
P1 (business-critical, revenue-impacting), P2 (significant, workaround available), P3 (low-impact). Definitions specific enough to remove ambiguity in classification.
02
Map Internal Dependencies
For each severity tier, identify which internal teams were required to respond and in what capacity. This mapping became the foundation of the OLA commitments.
03
Negotiate Commitments
OLAs are agreements — they require the other party to actually agree. Working with hosting, engineering, and infrastructure leads to establish targets that were achievable given real capacity.
04
Build the Escalation Path
When an internal team misses their OLA commitment, who is notified, by when, and through what channel? Without this, missed OLAs produce frustration rather than resolution.
05
Formalise and Report
Written OLAs signed by relevant team leads. Monthly OLA compliance reporting distributed to all parties. The reporting was the mechanism that made the agreements real.

The Negotiation Challenge

The hardest part of building an OLA framework isn’t the design — it’s getting other teams to commit to targets they’ll be held accountable for. The approach that worked: frame the OLA as a two-way commitment. Support would provide better quality escalations — clearer problem statements, reproducible steps, relevant logs already gathered — in exchange for faster internal response. Both sides were changing, not just engineering being held to a new standard.

Before and After

20+ days
P2 response — baseline
2 days
P2 response — OLA target
0
Formal internal commitments — before
4
Teams with signed OLAs — after

What We Learned

SLA performance is an internal process problem, not a support team problem. When P2 response times exceed 20 days, the instinct is to look at support. But if support is waiting on internal teams, the problem is the internal commitment layer.

Negotiation produces better OLAs than imposition. An OLA the hosting team helped design is one they own the outcome of. An OLA imposed from above gets routed around in practice.

The escalation path is more important than the target. What happens when the OLA is missed? A defined escalation path gives the agreement teeth without creating an adversarial dynamic.

OLA FrameworkProcess DesignSLA ManagementCross-functional AlignmentEscalation DesignStakeholder Negotiation