Field note ·
Reconciling OSS counters with probe samples
Counters and probes disagree for honest reasons. A monthly pack that pretends they are the same number will collapse in the review.
OSS performance-management counters count what the vendor’s software counted. A probe or a crowd sample counts what a handset or a box on a desk experienced. Both are useful. They are not interchangeable, and a quality pack that overlays them on one axis without a note is asking for a wasted hour.
A typical UK 4G retainability counter excludes some abnormal releases by design. A probe call that failed during a train journey through a cutting will still show as a failed attempt. If the SLA is written on the counter, the probe chart is context, not the contract number.
We keep two ledgers in the monthly pack. Ledger A is the contractual KPI, copied from the dictionary freeze. Ledger B is the probe or drive overlay, drawn on a separate page with its own sample size. Where they move together, the narrative can say so in one sentence. Where they diverge, we spend the paragraph on geography and sample, not on blame.
Probe fleets also cluster. A box that lives in a warehouse in Leeds will never see the climb out of the Holme Valley. If your enterprise customer’s staff work from villages around Emley, a warehouse probe is the wrong witness. We will say that in the pack rather than average it away.
Bring both extracts if you have them. Do not ask us to “blend” them into a single percentage for a slide. That blend cannot be audited six months later when the clause is disputed.