Godwit AI Labs Talk to us

HomeInsights › Managed IT

Your SLA probably promises a reply, not a fix

Response time is when somebody acknowledges you. Resolution time is when you can work again. Most support contracts guarantee the first and are silent on the second.

Published
4 September 2026
Reading
5 min
Topic
Managed IT
The distinction
Reply, not fixonly one is usually promised

Read a managed support contract closely and one thing tends to be missing. There will be a table of severities with times against them, and the times will be response times: thirty minutes for a P1, four hours for a P3. What will usually not be there is any commitment about when the thing gets fixed.

Both parties then sign, understanding different things.

One incident, and the part the contract covers A 30-minute response SLA, met in full, on a seven-and-a-half-hour outage 0h 2h 4h 6h 7h 30m Raised → acknowledged00:00 – 00:30 · the SLA → work starts00:30 – 01:15 → workaround01:15 – 03:00 → fault fixed03:00 – 07:30 Promised, measured, and met in full — the thirty minutes at the top The other seven hours are not mentioned in the agreement. A credit for a missed response would have been a percentage of one month’s fee.
Intervals are illustrative · the distinction between response and resolution is not

Three different clocks

Response is when a human acknowledges the ticket. It is easy to measure, easy to automate, and easy to hit. It is also the one thing that is nearly always promised.

Workaround is when you can work again, even if the underlying fault is still there — the failover, the manual process, the reboot that holds until Friday. This is the clock a business actually cares about, and it is rarely in the table.

Resolution is when the fault is gone. Frequently it depends on somebody who is not party to your contract at all: an ISP, a hardware vendor, a SaaS provider, the author of an application nobody has patched since 2021.

That last dependency is the honest reason resolution is hard to promise. It is not a reason for the contract to be silent about it.

Who decides it is a P1

Worth checking, because in practice the provider does — and the provider is the party whose numbers are measured against the severity that gets assigned.

The fix is not a clause about good faith. It is a severity table with examples from your estate, written down at the start: the billing run cannot complete is a P1; the Mumbai office cannot reach the internet is a P1; one person cannot print is a P3. Ten lines like that resolve almost every argument you would otherwise have at the worst possible moment, and writing them costs an hour.

What business hours means, in India, specifically

A four-hour response inside business hours can mean the next morning if the ticket lands at 6pm. Fine, if that is understood. So pin down: which hours, which days, whose holiday calendar, and what happens on a state holiday that is a working day elsewhere in the country. Diwali, Holi, and a regional new year are all foreseeable, and all of them have sat in the gap between two contracts.

If you need cover beyond that, price it deliberately as after-hours cover with a named rota, rather than hoping that a 24×7 label in the header means somebody is awake.

Service credits are not a remedy

The standard remedy for a missed SLA is a credit: a percentage of one month’s fee, often capped at that month’s fee, and usually claimable only if you notice and ask within a window.

Set that against a day of a warehouse not shipping, or a billing run that misses a cut-off. The credit is not compensation and was never designed to be — it is a signal, a way for the contract to register that something went wrong. Treat it as a thermometer, and put your energy into the clauses that reduce the number of bad days instead of the ones that price them.

What to ask for instead

  • A severity table with your own examples in it, agreed before the first incident.
  • A workaround or resolution target, even a soft one, even one that explicitly excludes third-party dependencies. A provider willing to write a target for the part they control is telling you something; one who will only commit to a reply is telling you something too.
  • An escalation path with names and elapsed times. After two hours it reaches X, after four it reaches Y. Not a mailbox. A person, and when.
  • A monthly report you did not have to ask for, listing tickets, times, and any breaches. If breaches only surface when you go looking, you will stop looking, which is roughly the point at which the SLA stops working.
  • Whatever the exit looks like. Documentation handed over, credentials transferred, and a notice period you can actually live with. The best time to agree how a support relationship ends is while both sides still want it to work.

The tell

Ask what they will not commit to. A provider who says plainly that they cannot promise a resolution time on an ISP fault, a vendor RMA, or a SaaS outage — and who will commit to a first-line target on everything inside their own control — has read their own contract and expects to be held to it.

A table of numbers with no exclusions anywhere in it has usually not been read by anybody who would have to meet them.

Need somebody to run it?

Monitoring, patching, backups and the phone answered when something breaks, on a retainer sized to the estate rather than to a rate card.

How running it works