By Yair Knijn · February 13, 2026
The dashboard is green because the VIPs scream: the quiet accounts breaching SLA in silence
The COO opens the monthly operations review and sees 97% SLA attainment across the desk. Green. The director nods, the QBR slide writes itself, and nobody asks the next question. That number is an average, and an average is a place where the worst cases go to hide.
The wrong assumption is that one figure describes the portfolio. It describes the loud middle. The three accounts that don't email the account manager every time a clock slips are sitting underneath that 97%, absorbing the breaches that keep the headline clean.
Why a single SLA percentage is a misleading vanity metric
A blended attainment number is a weighted vote, and the vote is rigged toward whoever files the most tickets. If a high-volume account submits 600 tickets a month and a strategic account submits 40, the strategic account's misses round to zero in the aggregate. You can breach every ticket they open and still report 95% if the rest of the queue runs clean.
IBM's own framing of SLA metrics splits multi-level agreements into customer-based, service-based, and corporate tiers precisely because one corporate number cannot represent obligations that differ by contract. Reporting only the corporate roll-up isn't a summary. It's a way to not look at the customer-based layer where the money and the renewals actually live.
Optimizing for noise: how loud customers train the queue
Triage is a learned behavior, and most desks learn it from feedback. The accounts that escalate, CC the director, and reopen tickets get worked first because the cost of ignoring them is immediate and personal. The agent feels it. The account that opens a P3, says nothing, and waits politely teaches the queue that it can wait. Over a few quarters the desk has quietly trained itself to serve volume of complaint, not value of contract.
The result is a desk that is responsive and wrong at the same time. It clears the noise, the noisy accounts are satisfied, and the dashboard rewards the behavior. Meanwhile the silent strategic account is having the worst experience in the book and generating no signal the desk can see.
Per-account, per-tier SLA attainment and the churn signal it reveals
Segment the same data by account and tier and the picture inverts. Churn analysis keeps surfacing the same lesson: the patterns that predict a lapse never appear in the aggregate rate. They show up when you split it. A strategic account at 82% resolution attainment for two consecutive months, with no open escalation, is not a happy account. It's a quiet one that has stopped expecting better and started shopping.
What to put in front of the COO instead of one percentage:
- Attainment per account, broken out by
responseandresolutionagainst that account's actual contracted targets, not the corporate default. - A trend line per strategic account, because two declining months is a churn signal even when the absolute number still looks acceptable.
- Breach concentration: what share of this month's misses landed on the top five accounts by contract value.
Re-weighting triage toward contract value, not volume of complaint
The fix is structural, not a pep talk about caring more. Tie priority to the account's tier and contract value at intake, so a strategic account's P3 outranks a commodity account's P2 by policy rather than by who shouts. Set distinct response and resolution clocks per tier, and make a strategic-account breach page someone with the authority to act before the customer notices, not after the credit hits the invoice.
OpsDesk runs per-account SLA clocks inside each customer workspace, so attainment is segmented by tier and value from the first ticket rather than reconstructed in a spreadsheet the week before the QBR. The breach you want to catch is the one the customer hasn't mentioned yet. See how the per-account SLA tracking works.