Marketing Agency ROI: From Pipeline to Contribution Margin

Agency ROI, marketing ROI and ROAS answer different questions

ROAS divides attributed conversion value or revenue by advertising spend. It can be useful for channel optimisation, but it often excludes the agency retainer, creative production, technology, internal time, returns and margin. A pipeline multiple divides sourced or influenced pipeline by investment. That is a useful leading indicator in B2B, but pipeline is still a probability-weighted claim on future value. Marketing ROI moves closer to the business decision by comparing incremental economic contribution with the total investment required to create it.

Agency performance is a fourth question. An agency may execute well while an uncompetitive offer, slow sales follow-up or limited fulfilment capacity destroys the overall return. A favourable market or strong brand can also conceal weak agency work. A fair review therefore separates company economics, marketing contribution, channel efficiency and the agency's controllable delivery. Combining all four into one number creates confidence without diagnosis.

Start with the fully loaded cost of the agency relationship

The denominator must contain the costs that would change with the decision. Include media, retainers, project fees, performance fees, creative and landing-page production, tracking and MarTech, external data and relevant internal work. Spread one-off setup cost across a defensible useful period or show it separately. Do not allocate every corporate overhead line to a campaign, but do not make implementation and management work disappear when comparing an agency with an in-house or fractional leadership model.

Use contribution margin instead of a revenue illusion

For ecommerce, begin with attributed net revenue, then remove refunds, cancellations, discounts and variable product, payment, fulfilment and service costs. The remaining contribution profit can be compared with the marketing investment. For recurring revenue, a customer value may be used only when churn, gross margin, expansion and the evaluation window are explicit. An unlimited lifetime value can make almost any acquisition programme look attractive; it does not improve the underlying evidence.

In B2B, revenue may arrive months after the first marketing interaction. Use closed-won contribution profit when enough mature cohorts exist. For an earlier decision, risk-weight only sales-accepted opportunities: opportunity value multiplied by a historical stage-specific win rate and the expected contribution margin. Segment the rates by deal type when sample size permits. If the data is thin, show conservative, base and upside scenarios instead of one decimal-point estimate.

A worked B2B example: why a 6x pipeline multiple is not 6x ROI

Assume a company spends $40,000 in one quarter: $18,000 in agency fees, $15,000 in media, $4,000 in production and tools, and $3,000 in relevant internal time. The programme creates $240,000 in sales-accepted pipeline, producing a 6x pipeline-to-investment multiple. Historical win rate for this stage is 25 percent and expected contribution margin is 45 percent. The base expected contribution profit is therefore $27,000: $240,000 × 25% × 45%. Against $40,000 of investment, the base economic result is negative $13,000 before any incrementality adjustment.

This does not prove the agency failed. The cohort may still be immature, the win rate may improve, or the investment may create future demand not yet represented in pipeline. It does prove that the 6x headline cannot be treated as a 6x return. The review should identify the assumptions that would change the decision: opportunity quality, win rate, margin, lag, baseline demand and which costs will recur. The next action may be to hold, repair sales handoff or run a more controlled test rather than immediately scale or terminate.

Attribution allocates credit; incrementality tests causation

An attribution model assigns conversion value to observed interactions according to a rule or model. It helps teams compare sources and understand paths, but it does not establish what would have happened without the marketing activity. Existing demand, brand strength, seasonality, sales effort, pricing and other channels can influence the same deal. Platform reports may also use different windows, identities and modelling, so their conversion totals should not simply be added to analytics and CRM results.

Incrementality compares the observed outcome with a credible counterfactual. Depending on budget and data maturity, that may involve holdouts, geo tests, controlled experiments or a robust time-series design. A mid-market company does not need an elaborate experiment for every campaign. It does need labels that distinguish observed, attributed, modelled and experimentally validated values. This simple evidence hierarchy prevents a modelled pipeline estimate from being presented as certain profit.

Build a one-page agency ROI board

The first line of the board states the decision and evaluation window. Next come fully loaded investment, qualified economic value and the conservative-to-upside ROI range. Show the assumptions immediately below: margin, stage-specific win rate, conversion lag, data coverage, attribution model and any baseline adjustment. Channel metrics belong underneath as diagnostics. The agency then recommends scale, hold, fix or stop, with an owner, expected effect, required budget, risk and next piece of evidence.

Build the first version in 30 days

In week one, define the decision and the value chain: for example, whether to renew the agency, increase spend or repair a channel. In week two, reconcile costs, CRM stages, margins and source data. In week three, build the first scenario range and mark every missing input. In week four, run one review with the CEO, marketing, sales, finance and the agency. The first model may be incomplete as long as uncertainty is visible and the next measurement improvement has an owner.

Do not reinvent definitions every month. Version genuine changes and keep the decision logic stable. Improve offline conversion imports, lead-quality stages, contribution data and cohort maturity over time. Reserve more rigorous incrementality tests for large budgets, disputed channels and decisions that are expensive to reverse. Measurement maturity should rise with the financial importance of the claim.

1. Using only media spend in the denominator while ignoring agency and production cost.
2. Giving every lead the same value despite large differences in quality and close probability.
3. Presenting platform ROAS as the return of the entire marketing system.
4. Comparing revenue with cost instead of contribution profit with fully loaded investment.
5. Counting open pipeline at face value without stage-specific risk weighting.
6. Treating attribution as proof of causation.
7. Ignoring conversion lag and judging the newest cohort too early.
8. Blending agency delivery with internal offer, capacity or sales problems.