Skip to content
Interface language: English
Imperium / Signal

“We replied in an hour.” The case is still open: what does the promised service time mean?

Distinguish first response from resolution, business and calendar time, pauses and handoffs before promising a service time.

Article language: English

A seated woman beside a glass sphere with dark reflections.
In this article

“We reply within an hour” could mean a customer receives a case number, a question from an adviser or completed help. Each might happen after sixty minutes. The word “reply” on a website does not explain which one the company promises.

Imagine a service maintaining an online shop. A customer reports that they cannot download an order report. The company registers the message, asks which report is affected, investigates and returns with a result. This fictional sequence helps define service time: which event starts the clock, which stops it and what happens in between.

A stage name should identify an event

An acknowledgement says a message reached the company's channel. It need not mean an employee has read it. An automatic number may be useful, but a customer should distinguish it from a response addressing the problem, particularly when both appear under the heading “we received an answer”.

Taking up a case can mean an initial assessment or the start of a specific check. If a company promises that action, it must be able to recognise it in its workflow. Assigning a name to a ticket does not always mean work has begun. Conversely, a phone conversation may start an investigation before any response appears in message history.

Zendesk's documentation on first reply time describes measurement from ticket creation to an employee's first public reply. It also distinguishes calendar time and business hours, with separate channel rules and exceptions. Even a tool field's name needs to be read alongside the definition used in the configuration.

Resolution needs a different condition. In our example it might be providing a way to download the report, restoring the function or the customer's confirmation of the result. The company must choose the moment appropriate to its service. Sending “it should work now” is not automatically the same as a checked result.

Do not try to settle every stage with “response time up to one hour”. You can separately state the first response and how the subsequent plan is determined. SLA, or service level agreement, is an agreed service standard covering specified events and times. The abbreviation alone does not clarify an ambiguous promise.

One case, two clocks

For this example only, assume service hours Monday to Friday, 09:00 to 17:00, in one time zone, with no holidays or time change during the weekend considered. A request arrives Friday at 16:30. An automatic acknowledgement goes out a minute later. An adviser replies Monday at 09:15.

From case creation to the human reply, forty-five minutes elapsed within the assumed service hours: half an hour Friday and fifteen minutes Monday. In calendar time it is sixty-four hours and forty-five minutes. Both numbers describe the same sequence; they differ in which period is counted.

If the promise concerns an hour of service time, the reply deadline in this fictional arrangement is Monday at 09:30. If the site promises an hour without that explanation, the customer may expect contact on Friday. Setting the right calendar in a tool is not enough. The recipient needs to understand the promise before using the service.

Specify which team's hours apply. A company may receive messages around the clock but reply in a limited window. It may also have a separate urgent-support scope. An available form does not mean a specialist is always on duty. When working across countries, state the time zone instead of assuming “Monday morning” begins simultaneously for both sides.

GitLab's own support levels state that its SLA times concern first response and should not be treated as expected resolution time. This is one provider's explicit separation of promises.

A pause needs a reason and an end event

At 09:15 Monday our adviser asks which report is affected. At 10:00 the required information arrives. Assume the company has agreed to pause a selected measure while waiting for such data. The pause is forty-five minutes, but the whole case remains open.

Atlassian's SLA-condition guidance describes setting the start, optional pause and end of a measurement. Waiting for a customer response is one example of a pause. The company chooses which wait justifies a pause in that measure.

Define what warrants that state. Is the question about information actually required for the next action? Can the information already be found in the material received? In our example it makes no sense to pause for a missing identifier if the customer supplied it in the first message and support overlooked it.

Restart also needs a recognisable event. The 10:00 reply might contain the required data, a request for clarification or only a signature. If every new message restarts the clock, check that behaviour against the agreed rule. If an employee decides after assessing the content, account for how and when that assessment happens.

Work waiting for a customer (in Polish) still needs an owner and a condition for further action. Excluding time from one measure should not remove the case from view. A customer may also be waiting for clarification of a question the company considers a sent request for data. The label “with customer” does not yet describe the whole situation.

Aeneia conceptual photography

What ends the measurement when help has arrived?

Suppose at 11:00 the adviser provides a checked way to download the report. Since Friday 16:30, sixty-six hours and thirty minutes of calendar time have elapsed. Within the assumed service hours it is two and a half hours. Excluding the stated pause leaves one hour and forty-five minutes. These are three different measurements, not three possible estimates of the adviser's effort.

None directly states active working time. Between receipt of data and the result, the case might have waited for another person. If the company needs effort, it must identify that separately. Time a case spends in service and time spent performing actions do not become the same quantity by subtracting one pause.

At 12:00 the customer confirms receipt of the required result. The company should know whether its standard ends at 11:00 or 12:00 and retain both facts if needed for later assessment. The condition should not depend on which time looks better in a particular report. It must follow the previously agreed meaning of completion.

Zendesk's metric list separately defines time to first resolution and time from creation to the last resolution. If a reopened case is resolved later, the second measure includes time to that later resolution. Before comparing reports, check which end event the selected number concerns.

If a customer returns with the same problem, decide how to preserve the connection to the earlier case. Otherwise the first resolution may appear successful while further service looks like a wholly new need. Not every return must be combined automatically. A new question about a different report may need a separate ticket. The basis for the choice should be recognisable.

Handoff to another person does not create a new beginning

Suppose at 10:15 the adviser passes the problem to a specialist. The second employee starts their own stage, but the customer's need arose earlier. The company can measure both teams' times separately, while preserving the whole case history. Replacing it with time since the latest assignment answers a different question from the customer's: how long have I waited since reporting it?

Changing the case type works similarly. The team first classified it as a usage question, then recognised a problem needing a functional check. Decide whether the standard changes and how to treat elapsed time. Merely moving the ticket to another category will not settle it. The decision-maker should understand the effect on the promise and on the report's interpretation.

With two channels, also choose the event from which one need is counted. The customer first wrote and later called about the same problem. If the system stores the contacts separately, linking them should preserve the original start chosen for the whole case. Otherwise the combined case can look newer even though the customer has waited since the same moment. Call duration can remain a separate useful measure.

The event in the tool must meet the promise to the customer

Match the agreed standard to information the company actually has. If a tool stores the time of the first public message, you cannot call that number a reliable time when analysis began without another check. Analysis may have started earlier. The message may instead have only asked to route the case to another department.

This matters especially for an AI-prepared reply. A message may provide useful help or merely acknowledge receipt. Define which stage the company considers fulfilled by that message and how it checks the later effect. Do not assume every automatic message is a resolution, or that every human message is one.

In metrics that produce unwanted behaviour (in Polish), the issue is a measure's influence on how people act. The earlier task here is to establish whether the fact measured is the same one promised. Even without pressure on the result, a measure can describe a different stage from the word “response” in an offer.

For priority in a service queue (in Polish), name which time the preference should shorten. Earlier urgency assessment and earlier start of appropriate help are distinct events. If a case passes to another team after assessment, assigning priority does not yet confirm availability of the needed person.

Service booking and releasing a place (in Polish) also distinguishes a message from its effect. A request for a new date may need agreement. A quick acknowledgement should not sound as though the customer already has a confirmed new booking.

Accept the standard against one case history

Before publishing a time promise, ask two people independently to identify the start, pauses and end in the same history. Add an automatic message, an after-hours reply and a return with the same need. If results differ, find why: different understanding of the rule, a missing event or tool configuration. Each problem needs a different correction.

Also test from the customer's side. Read only the sentence in the offer and the acknowledgement after contact, without internal instructions. Do those materials say when a reply can be expected and what it concerns? If the team needs five minutes of oral explanation for one sentence, improve the sentence before launching another advertisement.

Information about a delay (in Polish) is useful when a standard cannot be met. That differs from defining the time itself. The company should distinguish a missed promised stage from further work whose deadline has not yet been set, and tell the recipient what happens next.

If AI is part of the support process, discuss with Aeneia Audit whether the current report, configuration and reproducible histories can inform an agreed review. Before reducing a promised hour to half an hour, establish exactly what is meant to happen within it.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear