What a service level is and is not
Every marketing team that serves other people, whether a central team serving regional offices, a small team serving sales and product, or an operations function serving the rest of marketing, already has service levels. They are just unwritten. Requesters carry a guess about how long things take, the team carries a different one, and each missed guess becomes a complaint on one side and a grievance on the other. Writing the levels down turns the guesses into an agreement both sides can plan around.
A service level is a statement about a type of request: when work of this kind arrives complete, it is usually delivered within this time, and the requester gives feedback within that time. It is set from what the team actually delivers, agreed rather than imposed, and reviewed when the work changes. The campaign workflow guide makes the same point about review turnaround in a paragraph: a reviewer who knows the expected time can plan. The comparison of centralised and decentralised operations says a hub must treat its spokes as customers with a request route and a service level. This page builds the levels.
It is not a prioritisation tool. A service level says how long accepted work takes, not whether work is accepted, and the decision about which requests come first stays with the guide to prioritising work and capacity, including its route for genuinely urgent requests. Nor is it a personal target, a contract with penalties, or a promise to do everything faster. A team that sets levels it cannot meet has simply written its complaints in advance.
| A service level is | A service level is not |
|---|---|
| A typical turnaround per request type, from the team's own records | A benchmark copied from another team |
| Conditional on a complete request | A promise that applies however little the requester provides |
| Two-sided, with duties for the requester | A one-way commitment by the marketing team |
| A measure of the system, reported for the team | A KPIA metric tied directly to an objective. for each designer or writer |
| Reviewed when work or staffing changes | Fixed once and defended for years |
Before you start
- Have an intake route. Requests need one front door, as the campaign workflow guide and the prioritising guide both describe. Without it, there is nowhere to publish the catalogue and nothing to measure.
- Have stage history. Turnaround is set from the time recorded in each stage. The work management guide lists the cycle-time report by work type and stage that this tutorial reads. If stages are not yet recorded, start recording them and set provisional levels until there is enough history.
- Know who you serve. List the teams, offices or roles that send work. If the team has a written marketing operations charter, its list of internal customers and services is the starting point.
- Read the agency guide's service levels. The guide to selecting, briefing and reviewing agencies sets levels on both sides of an agency relationship, including how quickly the client gives feedback. The same two-sided logic applies inside the business.
- Find a sponsor. Someone senior on the requesting side who will sign the levels with you, so they are an agreement rather than marketing's terms.
The steps
List the request types the team serves
Go through recent requests and group them into types that behave alike: a social post from approved assets, an email from an approved template, a new landing page, a sales presentation, a printed leaflet, a short campaign, a report or dashboard change. A type is useful when items within it take broadly similar effort and pass through the same stages; split a type when its items vary widely, and merge types that never differ. Keep the list short enough to read on one page. Anything that does not fit a type is quoted individually, which is itself a level: "bespoke work is scoped and dated at triage".
Write what each type needs before work can start
For each type, write the entry conditions: what must arrive with the request for the work to start without a chase. For a campaign that is a signed brief, written to the standard the campaign brief guide sets; for design work, a Creative BriefA short document telling the people making creative work what it must achieve, for whom, with which message and within which limits., for which the creative brief template gives a format; for a printed piece, final copy, quantities and the delivery address. Keep each list to what has caused delays before. These conditions are the most useful part of the catalogue, because incomplete requests are where most service levels are lost.
Measure how long each type takes today
Pull the cycle time for each type from the work system, broken down by stage, for a recent period that reflects normal work. Separate the time the team spent working from the time it spent waiting, and note where it waited on the requester: for a missing input, a decision or a round of comments. This is the BaselineAn estimate of what a metric would have done without a given change, campaign or promotion, against which the effect is measured. the levels are built on. Read it with care: a handful of items per type shows a range, not a reliable average, and reading small numbers explains why one unusual item should not set the level. Never fill a gap with a figure from another organisation.
Agree a turnaround for each type with the people who request it
Take the baseline to the people who request each type and agree a level together. Start from what the records show the team usually achieves when the request is complete, not from the fastest item or the requester's wish. Where the requester needs something faster, discuss what would make it possible, such as a template, an earlier request or a lighter review, rather than promising it. Write the level as a typical turnaround for complete requests in working time, and say plainly that work outside the level goes through the urgent route or is scoped individually.
Write the requester's side of the agreement
A service level the requester can break without consequence is not an agreement. For each type, write what the requester commits to: a complete request with every entry condition met, feedback within an agreed time, one consolidated set of comments from all their reviewers rather than a sequence of individual ones, a named person who can approve, and notice of large requests ahead of time. These mirror the client-side levels in the agency guide. They usually matter as much as the team's own turnaround, because waiting for feedback is often the longest stage in the records. When that wait is pushing campaigns past their dates, the diagnosis for campaigns that keep launching late times recent launches stage by stage to show which side of the agreement the days were lost on.
Define when the clock starts, pauses and stops
Write the rules that decide what counts. The clock starts when a request meets its entry conditions and is accepted at triage, not when the first email arrives. It pauses while the item waits on the requester, such as for a missing input or for comments past their agreed time, and the pause is visible in the work system as a stage of its own. It stops at a defined point: delivered for approval, approved, or live, whichever the level names. Changes to the request after work has started reset the clock or are scoped again. Without these rules every miss becomes an argument about the facts. Customer support teams face the same choices about clocks and pending statuses, and the guide to customer service metrics settles them once for first replies, resolutions and repeat contacts. Just as an automatic acknowledgement does not stop the clock on time to first replyThe time from a customer's first message to the first reply from a person; read beside repeat contacts, as a quick reply that solves nothing still scores well., a message confirming that a request has arrived does not stop a marketing service level's clock.
Publish the catalogue at the front door
Put the catalogue where requests are made: on the intake form, in the request portal, at the top of the request board. Requesters should see the entry conditions and the level for a type at the moment they ask for it. Name who agreed each level and who owns it, using a RACI matrix so that one role is accountable for each level, one is consulted on it and the rest are informed. Tell the requesting teams what has changed and why, and point them to the urgent route for what the catalogue does not cover.
Report against the levels and discuss misses
Report, for each request type, how much work met its level, how much missed and where the time went for the misses: waiting on inputs, waiting on feedback, reworkThe share of items sent back from a review or launch gate to an earlier stage; each return carries a reason, so the pattern points to the step that failed. after review or time in the team's own queue. Report it for the team as a whole, never by person. Goodhart's law explains what happens when a measure becomes a target for individuals: items are closed early, feedback rounds are skipped and hard work is avoided. Discuss misses with requesters at a regular review, and treat a pattern of misses at one stage as a process problem to fix rather than a performance problem to manage.
Review the levels when the work or the team changes
Set a review date when the catalogue is published, and also review it when something material changes: a new request type, a change in the team, a new approval step, a new tool or a sustained shift in demand. Revisit any level that is almost always met by a wide margin, because it may be too loose to help anyone plan, and any level that is regularly missed for reasons inside the team, because it may promise more than the team can deliver. Record each change and the reason, so requesters can see that levels move with the evidence rather than with pressure.
A service catalogue on one page
The catalogue is the artefact the steps produce. It works best on a single page, laid out by request type, so that a requester can find their request and read what they must provide and what they can expect.
| Column | What it holds | Notes |
|---|---|---|
| Request type | The name requesters recognise | Match the names used on the intake form |
| Examples | A few typical requests of this type | Helps requesters choose the right type |
| Entry conditions | What must arrive before work starts | Links to the brief, template or input standard |
| Typical turnaround | The agreed level for a complete request, in working time | Set from the team's own records |
| Requester commits to | Feedback time, consolidated comments, a named approver, notice | The other half of the agreement |
| Clock stops at | Delivered for approval, approved or live | One definition per type |
| Owner | The role accountable for the level | From the RACI |
| Not covered | What goes through the urgent route or is scoped individually | Prevents disputes at the edges |
Two notes belong at the foot of the page: when the catalogue was last reviewed, and where requesters can see how the team is doing against it. Analytics questions usually have their own intake with different entry conditions, and the analytics request process covers that route; a marketing catalogue can point to it rather than absorb it.
A hotel group's central team agrees its levels (illustrative)
Skelvanor Hotels agrees service levels with its hotels
Skelvanor Hotels, an invented group of 14 hotels with a central marketing team of 6, received requests from every hotel by email, chat and phone. General managers complained that artwork took "forever"; the central team complained that briefs arrived without room rates or dates.
The operations lead grouped a quarter's requests into 5 types and pulled their stage history from the work system. Local event artwork, the most common type, took a median of 8 working days from request to final file. Of that, the design work itself took about 2 days. The rest was waiting: on average 2 days for missing details from the hotel and 3 days for a second round of comments, which usually arrived after a general manager had shown the first draft to a colleague who had not been asked to review it.
The team and a panel of general managers agreed levels for each type. For local event artwork, a complete request, with dates, prices, approved photography and a named approver, would be delivered for approval within 4 working days. Hotels committed to one consolidated set of comments within 2 working days, with the clock paused while the team waited for them. A second round of changes reset the clock unless the first draft had missed the brief.
After a quarter, the report showed 81% of complete local event requests delivered within the level. Most misses still came from waiting on a second round of comments, at 3 hotels in particular, so the review that followed added a line to the request form naming everyone who would see the draft. The team's own design time had not changed.
Common mistakes
- Copying someone else's turnaround. Another team's levels describe another team's work. Set levels from your own records.
- One-sided levels. If only the marketing team has commitments, every delay on the requester's side becomes a miss on marketing's. Write both sides.
- Starting the clock at the first email. The clock starts when the request is complete and accepted, or the level measures how incomplete requests are rather than how fast the team works.
- Hiding the pause. If waiting on the requester is not visible as its own stage, nobody can tell whose time the delay used.
- Using levels to decide priority. A level says how long accepted work takes. Which work is accepted, and what is urgent, belongs to prioritisation.
- Reporting by person. Individual targets distort the work. Report for the team and fix the stage where misses gather.
- Too many request types. A catalogue nobody can read at the moment of asking will not be used. Keep it to one page.
- Never reviewing. Levels set before a new approval step, a new tool or a change in the team stop meaning anything. Review them when the work changes.
