Almost every outsourcing relationship that goes bad had a perfectly good contract. The metrics were defined, the price agreed, the scope written down. What was missing was the mechanism for deciding what to do when reality didn't match the annex: who raises the issue, which meeting it belongs in, who authorizes the change, and when someone checks that it worked.
That's contract governance. It isn't bureaucracy — it's the only part of the agreement you use every week.
Three levels, three different conversations
The most common mistake is one monthly meeting that holds every topic. Forty minutes go to a single case that arrived by email that morning, and the decision that has been pending since last month never gets made. The fix is to separate by level.
Operational: daily or weekly
Between the provider's supervisor and their counterpart on the client side. Open queue, shift absences, cases stuck waiting on information, unexpected spikes. Fifteen-minute conversations with no minutes and no slides. Their job is to keep small things from piling up into a monthly problem.
Tactical: monthly
Between operations managers. Trends, not individual days. Quality, staffing, process changes, misses with their cause attached. This is where adjustments that fit inside the contract get approved: shifting a coverage window, rebalancing the team across queues, updating the knowledge base.
Strategic: quarterly
Between the people who signed. Scope, pricing model, automation priorities, growing or shrinking the team. It's the only meeting where it makes sense to ask whether the contract is still the right one. If it turns into a review of last month's numbers, it has been wasted.
Who can approve this at 5 p.m. on a Friday
Responsibility matrices usually stop at "who owns the process." That's rarely the question you need. The useful one is narrower: when a decision comes up, who approves it and how fast. It's worth writing down for the decisions that actually occur:
- Overtime or extra shift coverage during an unplanned spike.
- Additional hiring or a reduction in assigned headcount.
- Changing a step in the process, including who documents it afterwards.
- An exception to a service commitment during a contingency.
- Access to a new system or data set.
- Use of a tool other than the agreed one.
For each: who decides, who gets informed, and within what window. A contract that doesn't answer this forces improvisation exactly when improvisation is most expensive, and it always ends up in the inbox of someone who had no context.
The monthly meeting that's worth holding
A fixed agenda beats a good-looking dashboard. The order matters:
- Commitments from the last meeting. Met and not met, no discussion yet. Ten minutes.
- Metrics against commitment. Not loose averages or a new chart every month: the same indicators, compared to what was agreed.
- Exceptions. The cases that fell outside the range, with their cause. This is where anything gets learned.
- Risks for the coming month. Seasonality, client campaigns, system changes, vacation coverage.
- Decisions made today, each with an owner and a date.
If a review meeting ends without decisions that have an owner and a date, it wasn't a meeting: it was a report read out loud.
Item three is the one most often skipped and the one that pays best. A metric inside its range needs no explanation; a metric outside its range with no identified cause will fall outside again next month. On what to look at in those reviews, we wrote separately about what to measure from month one.
Escalation: routes, not loose names
An escalation matrix works when it says three things: which situation escalates, to whom at each level, and how long before it moves up if nobody answers. Without that last interval, every escalation depends on somebody happening to look at their phone.
It also helps to separate operational escalation from contractual escalation. The first resolves a case: a system is down, someone has to decide whether to switch to manual logging. The second opens a different conversation: the service commitment has been missed three months running, and that is no longer a night-shift problem. Blending them means the structural issues get handled in the daily chat and never reach the table where they could be fixed.
The four living documents
Almost no operation needs more paperwork than it already has, but four things do need to be current:
- Escalation matrix with names, shifts and step intervals.
- Process change log: what changed, who asked for it, who approved it, effective when.
- Incident log with root cause and action taken, not just date and duration.
- Decision log: one line per decision, with owner and date.
The decision log looks like the least important and prevents the most expensive argument there is: whether something was agreed or not. Six months later, when a manager on either side moves on, it's the only document that explains why the operation runs the way it runs.
Signs governance is failing
- Decisions happen over email or chat, and nobody knows which version is current.
- The monthly meeting gets cancelled whenever the month went well.
- The same topics come back three months in a row with no owner.
- The client escalates straight to agents, skipping the supervisor.
- Nobody can say what changed in the process last quarter, though everyone agrees something did.
- The service commitment is being renegotiated in practice, with nothing written down. If that's happening, the problem may be in the design: it's worth revisiting how to write an SLA you can meet.
Not every contract needs the same machinery
A three-person operation running a stable process doesn't need a quarterly committee or formal minutes; a daily channel and a short monthly review will do. An operation with several processes, extended shifts and third-party personal data does need all three levels and current documents. Governance gets sized the way staffing does: by complexity and risk, not by contract value.
What doesn't change is the principle: decisions need somewhere to be made. When they don't have it, they still get made — just late, and with nobody behind them.
Where smartBPO fits
When we stand up an operation, we settle before day one who talks to whom at each level, what gets decided in each meeting, and what the escalation intervals are. We keep the change log and the decision log as part of the service, not as an annex filled in the week before the quarterly review. It doesn't show up in a sales proposal, but it's the difference between correcting something within the week and finding out about it in the monthly report.