Remote monitoring SLA checklist

Which response time are you comparing?

An SLA becomes meaningful when it identifies the measured event and the clock used to measure it. An incident can have a camera timestamp, an event-generation timestamp, a platform receipt time, a review start, a decision, a call attempt and an acknowledgement. Different intervals answer different questions. Comparing a receipt-to-review target with an event-to-contact target can make two unlike services appear directly comparable.

Write the start and finish in ordinary language alongside the metric name. Ask which system supplies each timestamp and how missing or disputed timestamps are handled. The agreement should identify whether figures describe a target, a contractual commitment or an observed report. A percentile, an average and a maximum are different measures. Do not convert one into another or interpret an advertised number as the likely duration of every incident at your site.

Separate service dependencies from silent exclusions

Some work depends on equipment and people outside the monitoring provider's control. A failed connection can prevent an event from reaching the reviewing system. A site contact may not answer. A local responder may have its own acceptance and attendance conditions. These dependencies should be stated before the target is assessed, together with the provider's duty to recognize and communicate the resulting gap.

A long exclusion list without an outage procedure leaves the buyer with a promise that can disappear at the exact moment it matters. Ask how a missing feed is recorded, who is notified and whether the incident remains open while a dependency is unavailable. The SLA cannot make an external dependency reliable, but it can make responsibility and reporting clear. Distinguish an agreed suspension of a duty from an unexplained absence of service.

What should the monthly SLA report contain?

A useful report shows the population behind each figure. Ask for the count of eligible events, excluded events, events with missing timestamps and events outside the target, with reasons recorded consistently. Include the reporting time zone and the period covered. If only a small number of events occurred, avoid drawing a broad conclusion from a single percentage. Read individual exceptions as well as the aggregate.

The review should produce an action, an owner and a checkable completion condition where a problem is found. A report stating that a camera was unreliable does not show whether the camera was repaired or the service resumed. Link the service exception to the technical work order or site decision without copying passwords, personal contact numbers or footage into an unrestricted spreadsheet. Decide how disputed classifications are reviewed and who can amend the final record.

Work through one timeline before comparing the promise

Use an explicitly fictional timeline to check definitions. Suppose an event occurs at 22:00, the receiving system records it at 22:02, review begins at 22:03 and the first call is attempted at 22:05. Receipt-to-review is one minute, occurrence-to-call is five minutes and receipt-to-call is three minutes, provided the time sources are genuinely aligned. None of these values establishes acknowledged ownership or attendance.

The example contains no performance claim. It demonstrates why a shared number can describe different duties. If the camera clock is uncertain, the occurrence-based interval should remain unmeasured until its relationship is supported. If the recipient never acknowledges the call, the report should not invent an end time for accepted response. Missing evidence is part of the result, not an invitation to choose the most favorable interval.

Ask the provider to map its own target and report onto this sequence using the same terminology. Preserve the stated clock starts, stops, exclusions and denominator. Once definitions agree, the business can decide which intervals matter and which records are practical to obtain. That is a stronger basis for service review than comparing a pair of headline response-time claims without their measurement rules.

A fictional comparison: two attractive five-minute promises

Consider two hypothetical proposals using the same five-minute phrase. The first starts its clock only when an event arrives in its platform and ends at the first call attempt. The second describes the time from an observed event to a human decision, but excludes communication to the site. Neither statement tells you when a person takes ownership of the response. The figures cannot be compared as if they measure attendance or resolution.

In the comparison sheet, preserve each original definition. Ask both providers to map the same staged incident onto their timeline and identify the records available at each point. If an interval is not measured, mark it unmeasured. If a responsibility is not included, mark it outside scope. The purpose is to expose differences before agreeing a service, not to produce an artificially precise league table from incompatible metrics.