Control agency scope creep before extra work erodes the margin

An agency controls scope creep by identifying a request against the agreed scope, estimating its effect on cost and timing, and obtaining the appropriate decision before extra work starts. The useful outcome is an agreed choice: include it within existing scope, exchange it for another deliverable, approve a paid change, or decline it. A more detailed timesheet discovered after delivery is too late.

This is not an argument for charging clients for every conversation. Some requests correct agency mistakes. Others clarify a deliverable that was already promised. The commercial discipline is to distinguish those cases from additional work and make the distinction understandable to both sides.

For an agency owner, the cost of ambiguity appears in squeezed margins, delayed launches and difficult renewal discussions. An AI-assisted process can collect requests and prepare comparisons, while an accountable producer resolves unclear scope and brings a concrete decision to the authorized people. The process below is an operating design; contract terms should be established with suitable professional advice for the agency's circumstances.

Define scope in units people can recognize

A useful scope describes deliverables, quantities, review rounds, dependencies, acceptance criteria and exclusions in language the client and delivery team can both use. “Campaign support” gives little help when a client requests three additional markets. “Six approved static advertisements in two agreed sizes, with two consolidated review rounds” gives the team something to compare.

Include the client inputs that affect timing: product information, legal approval, access, consolidated feedback and a decision maker. A date without its dependencies can become a false promise when an input arrives late. The scope should explain how delays and changes are handled rather than leaving the account manager to negotiate under deadline pressure.

AIGA's standard agreement resource distinguishes the project proposal and terms and provides material for handling changes. It is a useful reference for the structure of a design engagement, not a ready-made legal answer for every agency or jurisdiction. AIGA Standard Form of Agreement for Design Services

Turn the signed agreement into a readable baseline record, but keep it linked to the original. An assistant's summary does not supersede the contract. If the baseline and signed document disagree, pause the classification and resolve the discrepancy before presenting an additional fee.

Capture requests where they actually arrive

Requests appear in email, project comments, meetings and messaging channels. Create a lightweight way to log the requested change, original wording, requester, affected deliverable and requested date. Staff should be able to capture a request without writing a commercial justification on the spot.

Do not treat every casual idea as an instruction to proceed. Mark whether the client is exploring an option, requesting an estimate or authorizing work. “Could we also do this in Arabic?” may mean any of those things. The next action is clarification, not automatic translation and an unexpected invoice.

For meeting-derived requests, link an approved note or confirmation rather than treating uncertain transcription as authority. The record should distinguish what was explicitly requested from what the agency inferred. If recording or transcription is used, establish the necessary permissions and retention arrangements first.

Keep access scoped to the people who need the material. Client budgets, product plans and personal information should not be copied into an unrestricted cross-client workspace. The ICO describes data minimisation as a core UK data-protection principle; the practical design here is to retain relevant evidence without distributing unnecessary personal data. Applicability depends on the business. ICO's guide to data protection principles

Classify the request before discussing a price

Use a small classification that explains the proposed treatment. A correction restores the agreed output. A clarification resolves an ambiguity. A substitution replaces planned work. An addition expands the work. An acceleration changes the delivery constraint. These distinctions help avoid an adversarial default that everything new is billable.

The initial classification can be assisted, but disputed or ambiguous cases need a producer or account lead. The system should surface the relevant scope passage and its reasoning. It should not confidently reject a client's request because a keyword fails to match the statement of work.

Asana's change-control guidance describes a structured process for evaluating and deciding changes. The agency-specific recommendation here is to make that evaluation small enough to use before delivery starts, with the client's actual commercial options visible. Asana's change-control process

Estimate the real effect, including the work around the work

Estimate production, project management, quality checks, localization, deployment and any external costs. A new asset rarely consumes only the time needed to draw or write it. It may require a revised brief, review coordination, accessibility checks, new tracking or changes to an existing launch plan.

Ask the delivery owner for a range where uncertainty remains. Explain the assumption that would move the estimate, such as whether the client supplies approved translations or expects the agency to arrange them. A precise number built on missing inputs is less useful than an honest range with a clear next decision.

Capacity matters separately from effort. A request may take eight hours but require a specialist unavailable until next week. Do not convert effort mechanically into a promised calendar date. Show the earliest feasible start, dependencies and effect on work already committed to other clients.

Use cost rates and billing rates deliberately. Cost estimates support margin decisions; billable values support commercial proposals. Harvest's project analysis documentation illustrates how time, fees and budgets can be tracked, but the agency must still define its own cost model and reconcile external expenses. Harvest project analysis

Present a change order that makes the choice easy

The client should understand what changes, why it changes the plan and what can happen next. A concise change order can contain the request, affected baseline, proposed deliverables, additional fee or exchange, revised milestones, assumptions, acceptance criteria and authorization method. Include a version identifier so the decision attaches to one proposal.

Offer genuine alternatives when they exist. For example, keep the launch date by exchanging a lower-priority deliverable, add the requested work after launch, or approve extra capacity if it is actually available. Avoid making a paid addition appear to be the only option when a sensible substitution would satisfy the business goal.

A useful message might say: “The additional language version adds translation coordination, layout and QA. We can deliver it after the current launch for the stated fee, or replace the two planned email variants and retain the date. Please select the option you want us to schedule.” This is an illustrative communication pattern, not a contract clause.

Do not rely on silence, an emoji or an ambiguous meeting comment unless the agreed authorization process explicitly supports that form and its applicability has been established. The system should preserve the decision, approver and approved version. If approval is required but missing, the job remains pending.

Fictional USD example. Monetary assumptions in this guide are illustrative US-dollar amounts, not quoted prices, measured results or exchange-rate conversions. The currency does not imply a US tax or legal setting.

Make approval change the delivery plan

An approved change must update the project budget, task list, timeline and acceptance checklist together. Otherwise the commercial record says the scope changed while the team continues working from the original plan. The producer owns this reconciliation and checks that the delivery team understands the revised commitment.

Concessions are a legitimate business decision. Record them so goodwill does not disappear into unexplained overrun. The owner may choose to absorb a small change to protect a relationship, but repeated concessions should be visible when discussing renewal scope and resourcing.

For urgent work, establish a narrow emergency authority before it is needed. It might allow an account lead to authorize a defined amount of corrective effort, with prompt documentation. It should not become a blanket rule that deadline pressure permits unlimited unapproved additions.

Work through a realistic margin example

Consider a synthetic fixed-fee project priced at USD 40,000. Planned direct delivery cost is USD 22,000, leaving USD 18,000 before overhead, tax and any costs excluded from that definition. The planned project contribution margin is USD 18,000 ÷ USD 40,000 = 45%.

A client requests an addition requiring 30 production hours at USD 150 loaded direct cost per hour, six management hours at USD 180 and USD 720 in external costs. The incremental cost is USD 4,500 + USD 1,080 + USD 720 = USD 6,300. If absorbed, total direct cost becomes USD 28,300 and contribution falls to USD 11,700, or 29.25% of the original fee.

Suppose the agency proposes USD 10,500 for the addition. If approved and delivered at the estimated cost, total revenue is USD 50,500, direct cost USD 28,300 and contribution USD 22,200. The resulting margin is approximately 43.96%. The change's own contribution is USD 4,200, or 40% of its fee. It does not quite preserve the original 45% project margin.

To price the change at a 45% contribution margin under these assumptions, its fee would be USD 6,300 ÷ 55% = approximately USD 11,454.55. This arithmetic informs a decision; it does not dictate market price or what the client owes. The agency may choose a different price for strategic reasons, provided the choice is visible and authorized.

Do not describe identified scope value as recovered revenue. A proposal can be declined, exchanged or waived. Track approved, invoiced and collected amounts separately, and include the cost of administering the process when evaluating its commercial benefit.

Test the awkward cases before automating client messages

Start with internal detection and draft preparation. Compare recent requests with the baseline and ask account leads whether the classification is fair. Include a correction caused by the agency, a vague original scope, a client approver who has left and a request spanning two retainers. These cases reveal whether the workflow understands commercial reality.

Run the estimate and proposal process on one active account with a cooperative delivery lead. Review turnaround time and how often the client needs clarification. If a change order creates more confusion than it resolves, simplify the explanation rather than adding more fields.

Keep automatic external sending disabled until the agency has defined which messages are authorized and how it handles sensitive disputes. A mistaken internal classification is repairable; sending an unjustified fee demand can damage the relationship. An accountable person should own disputed classifications and negotiation.

Failure handling needs practical rules. Missing scope goes to the account lead. Missing cost inputs go to the delivery owner. A stale approved version blocks scheduling. A declined request cancels pending production tasks. If the client changes their mind, create a new decision record rather than overwriting the previous one.

Measure decisions, protected capacity and client clarity

Track time from request to classification, time to a client-ready option, work started without required approval and the value of approved, waived and declined changes. Also review rework caused by agency errors. A process that bills additions efficiently while ignoring avoidable mistakes is not improving delivery quality.

Measure the capacity effect with actual hours and displaced commitments. An approved change that overloads a team can harm other accounts even when its margin appears attractive. Review whether the work was scheduled within available capacity and whether the revised milestone was met.

This is a poor first project if no usable baseline exists, the owner routinely overrides the approval process without recording it, or almost all work is exploratory and intentionally flexible. Start by agreeing a commercial model that matches the work. A rigid change-order workflow cannot repair an unsuitable engagement structure.

Sources

• AIGA: Standard Form of Agreement for Design Services
• Asana: Change-control process
• Harvest: Project analysis
• ICO: Data protection principles

Work out the economics · Use a practical playbook