An agency relationship fails at both extremes. Constant intervention replaces the expertise the client hired. Uncritical delegation leaves the client accountable for decisions it cannot explain.

The right boundary depends on consequence, uncertainty and demonstrated performance. It should be designed at the start, then adjusted through evidence.

Establish what each party owns

Create a short responsibility and decision map before delivery. Cover:

  • business outcomes and success measures
  • scope, priorities and budget
  • research and user access
  • architecture and technical standards
  • security, privacy and accessibility
  • content and evidence approval
  • day-to-day delivery decisions
  • quality assurance and acceptance
  • launch and operational readiness
  • change, escalation and stop authority

Name individuals, not departments. Explain consultation requirements and response times. If several client stakeholders can issue direction, identify whose instruction is binding.

The agency should own recommendations and delivery within its competence and agreed constraints. The client retains accountability for strategy, duties, risk acceptance, funding and the service it places into operation. Independent assurance may be needed where neither party should mark its own work.

Agree outcomes with controllable measures

“Redesign the website” describes an output. Commercial outcomes such as qualified enquiries or successful portal tasks can guide choices, but the agency may not control every influence on them.

Build a balanced measure set:

  • intended client or colleague outcome
  • quality, accessibility, security and reliability
  • delivery forecast and budget
  • adoption and operating effort
  • learning about critical assumptions

Define baselines, timeframes and attribution limits. Avoid inserting unsupported uplift targets simply to make the brief feel commercial.

Acceptance criteria should cover what delivery can prove. Benefit ownership continues with the client after launch.

Give context, ask for recommendations

Clients hold knowledge an agency cannot infer: strategy, past decisions, internal politics, risk appetite, clients and operational constraints. Provide it early and update it when circumstances change.

Describe the problem and consequence before prescribing a solution. Ask the agency to explain its recommendation, alternatives, evidence and trade-offs. This preserves outside perspective while keeping decisions inspectable.

Feedback should identify the observation and why it matters. “The main audience may not recognise this term” is actionable. “Make it more premium” needs clarification. Personal preference can be stated as such and weighed against research, brand and outcome.

Agencies should challenge the client respectfully. Automatic agreement may make approval easy and the work weak.

Use a predictable governance rhythm

Set a cadence suited to the work. A delivery check may cover progress, blockers and decisions weekly. A monthly or phase review may cover outcomes, forecast, scope, risk and assurance. Discovery or a critical release may need a different rhythm.

Use one agreed channel for decisions and maintain logs for scope, risk and action. Informal messages can help a team collaborate; they should not silently alter approved work.

Protect focused delivery time. Consolidate feedback and distinguish questions from instructions. A stream of comments from several client leaders creates rework and encourages the agency to wait for approval on everything.

Reporting should state what changed, what is uncertain and what decision is needed. The client must respond constructively to early problems. If bad news reliably produces blame or another approval layer, reporting will become defensive.

Manage change without drama

Define a scope change and its route. The agency should describe the reason, outcome, effort, timing, risk and effect on existing commitments. The client decides whether to add, substitute, defer or decline.

Small changes still accumulate. Maintain a change budget or threshold where appropriate, but preserve visibility.

Distinguish discovery from delivery commitments. Some uncertainty cannot be priced responsibly at the outset. A staged commercial model with decision gates may be better than a fixed promise built on hidden assumptions.

When the client changes direction, acknowledge the consequence. When the agency missed an assumption it should reasonably have identified, address responsibility through the contract and evidence rather than reflexive blame.

Remove the blockers only the client can remove

Sponsors add most value by securing access, subject expertise, decisions, content, data, legal review and stakeholder alignment. Track client dependencies alongside agency work.

Allocate real time. A project cannot be “delegated” while waiting weeks for a partner to approve the proposition or for technology to grant access.

The sponsor also communicates purpose and change within the organisation. The agency can support this and speak directly with stakeholders where agreed; banning contact to “control the narrative” often creates distortion. The client remains accountable for transparent internal communication.

Review the work at the right altitude

Sponsors should inspect working evidence and ask how it supports outcomes. They need not attend every internal ceremony or annotate unfinished work.

Technical decisions should remain with competent specialists within approved architecture, security and operational boundaries. The client still needs adequate documentation and assurance, especially for hard-to-reverse choices. “The agency knows the platform” is not enough when the decision affects confidential data or long-term lock-in.

Design teams should have space to work through alternatives before formal review. At review, assess the rationale, research and criteria. Avoid requesting multiple variants merely to discover a personal preference.

Calibrate oversight through evidence

Trust is a prediction based on performance. Expand decision authority when forecasts, quality, reporting and judgement have been reliable. Tighten governance when evidence shows repeated failure, unexplained variance or unsafe practice.

Do this explicitly. State what changes, why and when it will be reviewed. Sudden attendance at every meeting communicates alarm without giving the team a route to restore confidence.

Monitor observable conditions:

  • forecast accuracy and early warning
  • quality and acceptance evidence
  • decision and dependency age
  • scope stability
  • continuity of the proposed team
  • unresolved risks and defects
  • client and agency effort spent on avoidable rework

Do not infer quality from charisma, chemistry or visible busyness.

Keep the client capable of leaving

Healthy partnership does not require dependence. Ensure the client has appropriate access to source, environments, content, domains, accounts, documentation and supplier contracts. Define data return, deletion, transition and intellectual-property rights.

Build knowledge transfer throughout the engagement. Confirm who will operate and improve the service after launch. An agency may continue in that role, but the choice should remain deliberate.

At closure, review outcomes, unresolved risks, operating ownership and lessons. References and case-study permissions are separate decisions and should not be assumed.

A useful sponsor posture

Own the purpose, constraints, investment and decisions. Make context and people available. Expect the agency to recommend, explain, deliver and report. Intervene when evidence requires it, at the level where you add authority or knowledge.

The one-page client governance checklist available below sets out decisions to retain, delegate and share, plus a review cadence. Adapt it to project risk and revisit it as the relationship matures. Good oversight gives capable specialists room to think while making accountability impossible to lose.