Measure Whether the Offer Is Working

Task

Track win rate, sales cycle, ACV, margin, and delivery variance.

Summary

Track sales results, time to close, contract value, margin, delivery differences, and time to value together.

Track the Numbers That Prove an Offer Can Be Repeated

Task ID: S2-13

A repeatable offer must work twice: it must convert qualified opportunities into worthwhile contracts, and it must deliver those contracts at a predictable cost and pace. A weekly report connecting win rate, sales cycle, contract value, margin, and delivery variance shows whether the company is building a dependable offer or merely getting better at selling exceptions.

A company can appear to have found a repeatable offer while every sale still creates a new operating problem.

The sales team reports a higher close rate. Contract values are rising. The pipeline looks healthy. Yet delivery keeps taking longer than expected, senior employees are pulled into rescue work, and projects that looked profitable when sold finish with little or no margin.

The opposite can happen as well. Delivery is efficient, customers receive the promised result, and margins are sound, but the offer rarely closes because the sales conversation, customer definition, or price is wrong.

Neither picture can be diagnosed from one metric. Win rate alone does not show whether the company is winning profitable work. Margin alone does not show whether the offer is commercially attractive. Average contract value alone does not show whether larger deals are creating longer sales cycles and more custom delivery.

The company needs a joined view of the sale and the work that follows it.

The operating principle

Track the offer as one economic system, from qualified opportunity to completed delivery.

That system has five essential questions:

1. Win rate: Of the opportunities the company chose to pursue, how many became customers?
2. Sales cycle: How long did those opportunities take to reach a decision?
3. Contract value: How much annual or project revenue did each win contribute?
4. Margin: How much of that revenue remained after the direct cost of delivery?
5. Delivery variance: How far did actual or forecast delivery differ from the scope, effort, cost, and schedule assumed when the work was sold?

These measures belong together because an improvement in one can damage another.

A discount may improve win rate while reducing margin. A larger contract may increase annual contract value while adding procurement steps, extending the sales cycle, and introducing custom requirements. A rigid qualification process may reduce the number of opportunities but raise win rate and delivery consistency. A delivery team may protect its completion dates by adding unplanned people, hiding a cost overrun inside an apparently successful schedule.

Established project-control methods therefore integrate scope, schedule, and cost rather than treating them as separate reports. NASA’s earned value guidance, for example, compares planned work, completed work, and actual cost to identify problems before a project ends. A small service company does not need the full administrative machinery used on major programs, but the underlying logic is directly useful: progress cannot be judged from spending, activity, or elapsed time alone.

The sales side requires the same discipline. A win rate is meaningful only after the company defines which opportunities enter the denominator, what counts as a loss, and which time period or cohort is being measured. HubSpot’s guidance notes that organizations differ on whether “no decision” belongs in the loss population, but consistency is essential for comparison over time.

The goal is not to produce a more elaborate dashboard. It is to create a weekly operating record that answers a practical question:

Are we selling the same promise to similar customers and delivering it with increasingly predictable economics?

Why these measures matter at this stage

This work belongs after the company has begun shaping a clear offer but before it depends on that offer for sustained growth.

At this point, the company should have at least a working definition of:

• the customer it intends to serve;
• the problem it solves;
• the result it promises;
• what is included and excluded;
• the price or pricing method;
• the major steps in delivery.

Those definitions may still change. The purpose of measurement is to show which changes are necessary.

The task can begin immediately because it does not require a mature analytics platform. In practice, however, trustworthy reporting depends on several basic decisions: a common opportunity definition, consistent sales stages, a recorded delivery estimate at the time of sale, direct-cost rules, and named owners for the underlying data. Without those elements, the report will contain numbers but not comparable evidence.

Research on business-to-business opportunity management highlights another reason to start early: sales data reflects the opportunities a company chose to pursue, not every opportunity that existed in the market. That selection effect can bias conclusions about customer fit and win probability. The data may also reflect changing sales behavior as the team learns, rather than a stable market response.

Machine-learning research has shown that historical opportunity data can improve predictions of whether a business-to-business deal will close. It also identifies the limitations that matter for an early company: business-to-business datasets are often small, noisy, and affected by changing market conditions. A company should therefore establish stable definitions and reliable records before attempting sophisticated scoring or artificial-intelligence forecasting.

The immediate objective is more modest and more valuable: create a baseline, find the largest sources of variation, and decide what to standardize next.

flowchart LR
    A[Qualified opportunity] --> B[Won, lost, or no decision]
    B --> C[Sales cycle and contract value]
    C --> D[Delivery scope, budget, and schedule]
    D --> E[Actual effort, cost, and completion]
    E --> F[Margin and delivery variance]
    F --> G[Change qualification, scope, price, or process]
    G --> A

The diagram shows the required feedback loop. A sale creates a delivery commitment. Delivery results must then change how the next opportunity is qualified, scoped, priced, or accepted.

Define the measures before building the report

Most reporting problems that look technical are definition problems. Two employees can calculate “win rate” correctly and still produce different answers because they used different populations. The same problem appears in sales-cycle dates, contract value, direct costs, and scope changes.

A metric dictionary should be approved before the first weekly report. It does not need to be long, but it should record the formula, inclusion rules, source system, owner, refresh timing, and known weaknesses.

Win rate needs a fixed denominator

For an offer-level operating report, the cleanest count-based formula is:

The company must define when an opportunity qualifies for this population. A casual inquiry should not carry the same weight as a buyer who meets the customer criteria, has a relevant problem, and has agreed to a sales process.

Two versions are often useful:

• Offer win rate includes all qualified opportunities that reached an outcome, including no-decision outcomes under a stated policy.
• Competitive win rate includes only decisions in which the buyer selected the company or an alternative supplier.

A value-weighted version should also be calculated:

Count and value win rates answer different questions. A company may win many small deals but lose most of the revenue available in larger opportunities.

The report should use either a closed cohort or a clearly labelled close-period view. A cohort view follows opportunities qualified during the same period until they reach an outcome. It takes longer to mature but avoids comparing a fast-closing quarter with a later quarter whose difficult deals remain open.

Sales cycle needs an agreed starting event

“Days to close” is not comparable if one salesperson creates an opportunity after the first email while another waits until a proposal is requested.

Choose one operational start:

• completion of qualification;
• acceptance into the active opportunity pipeline;
• completion of a paid or unpaid discovery step;
• another event that every salesperson can identify consistently.

The end should also be explicit: signed agreement, purchase order, deposit received, or closed-lost decision.

Use the median as the headline measure. A few stalled opportunities can pull the arithmetic mean upward. Also show the 75th percentile and the oldest open opportunities. Separate won, lost, referral-sourced, and outbound-sourced cycles when volume permits; each may represent a materially different buying process.

Open opportunities create a measurement problem because their final duration is not yet known. In statistical terms, they are right-censored observations: the company knows that an opportunity has remained open for at least a certain period, but not when or how it will end. Time-to-event methods such as Kaplan–Meier analysis can incorporate such observations without pretending that open deals have closed. A young company may not need that technique immediately, but it should at least report open-deal aging rather than calculating cycle length only from recent wins.

ACV is useful only when the offer contains recurring value

Annual contract value, or ACV, normally annualizes the recurring portion of a customer contract. One-time setup, onboarding, and custom implementation charges are commonly excluded so that contracts of different lengths can be compared on the same recurring basis. ACV is not a standardized accounting measure, so the company must document its own policy and apply it consistently.

That definition creates an important boundary for a productized service.

If the offer is a one-time engagement, calling its price “ACV” can obscure more than it reveals. The weekly report should then show:

• average booked project value;
• recurring or follow-on value, if any;
• one-time implementation or service value;
• total contract value for multi-period commitments.

Do not increase ACV merely by placing ordinary project work inside a twelve-month contract. The measure should reveal the economic structure of the offer, not rename it.

Margin needs three versions

The Securities and Exchange Commission explains gross profit as revenue less the costs directly associated with producing the goods or services sold, before other operating expenses are deducted. An internal delivery-margin measure can follow the same logic, but finance must decide which costs are treated as direct and reconcile the internal measure with financial reporting.

For each engagement, track:

Direct delivery cost should normally include:

• employee time at a consistent loaded cost;
• contractors and subcontractors;
• project-specific software, hosting, or data;
• travel and project expenses;
• rework, service credits, and refunds attributable to delivery.

Record margin at three points:

1. Planned margin at sale: based on the scope and cost estimate approved before signing.
2. Forecast margin: based on actual cost to date plus the current estimate of remaining cost.
3. Final margin: based on completed revenue and direct cost.

The movement between these values is often more useful than the final percentage. A project that falls from a planned 45% margin to a forecast 20% margin in its second week is already teaching the company something about its estimates, scope, or process.

Delivery variance needs an approved baseline

Variance cannot be measured against a plan that was never recorded.

At the time of sale, capture:

• promised outputs;
• exclusions;
• assumptions about customer participation;
• planned direct hours or cost;
• planned duration;
• required roles;
• external dependencies;
• acceptance criteria.

Then measure actual and forecast results against that baseline.

A simple service organization can borrow the logic, without the complexity, of earned value management: compare the work that was supposed to be completed, the work actually completed, and the cost incurred. NASA’s guidance emphasizes that budget-versus-spend alone is insufficient because it does not show how much work has been accomplished. It also recommends investigating favorable variances, since an apparent underrun may reflect delayed work or incomplete scope rather than improved efficiency.

Approved scope additions should not be hidden inside delivery failure. Report the original baseline, approved change, and current forecast separately. Otherwise, the company cannot tell whether an engagement overran because it was estimated poorly, executed poorly, or deliberately expanded.

Build one weekly sales and delivery report

The expected evidence is a weekly go-to-market and delivery metrics report. “Go-to-market” means the work used to find, qualify, sell to, and acquire customers.

The report should be short enough to review in a standing operating meeting. Its purpose is not to archive every field. Its purpose is to expose changes, exceptions, and decisions.

A useful structure has four parts.

The report should show trends and detail. A twelve-week trend reveals direction; an engagement-level exception table reveals cause.

Segment before drawing conclusions

At minimum, segment the sales measures by:

• offer version;
• customer type;
• sales source, such as referral or targeted outbound;
• new customer versus existing customer;
• price or contract-value band.

Segment delivery by:

• offer version;
• delivery lead;
• customer segment;
• standard versus approved custom work;
• project-value band.

Do not create so many segments that every category contains one deal. The point is to identify material differences, not to produce a complete taxonomy.

Offer version is especially important. If the scope or price changes, preserve the old version on existing records. Otherwise, results from materially different offers will be blended and the baseline will become difficult to interpret.

Connect the sales record to the delivery record

Every won opportunity should create a delivery record with the same identifiers and commercial assumptions. The handoff should carry:

• offer and version;
• signed price;
• recurring and one-time value;
• promised result;
• scope and exclusions;
• budgeted direct effort and cost;
• planned dates;
• assumptions made during the sale;
• approved exceptions.

The delivery team should not have to reconstruct these facts from emails. More importantly, the company must be able to compare the economics approved at sale with the economics currently forecast.

The weekly review should focus on changes since the previous report. For example:

• a proposal was discounted below the approved price;
• a referral deal closed faster than the current median;
• a project’s estimated hours increased by 20%;
• a customer delayed access to required data;
• an unpriced request was accepted;
• forecast margin fell below the company’s working threshold.

Each exception needs a reason code, a short explanation, an owner, and a next action. Narrative without ownership becomes commentary; a metric without context becomes noise.

Read the first baseline without overreacting

A working target of establishing a baseline after 30–60 days is reasonable as an operating milestone, but it is not an industry benchmark.

Whether the period is sufficient depends on the speed and volume of the business.

A company that closes several small engagements each week and delivers them within ten days may have a useful initial distribution after one or two months. A company with a 120-day sales cycle and a three-month implementation period will not. After 60 days, it may have only pipeline-aging data, early effort variance, and a few decided opportunities.

The first report should therefore distinguish three levels of evidence:

• Observed: completed sales and delivery outcomes.
• Forecast: current estimates for open opportunities and active engagements.
• Hypothesis: an interpretation that has not yet been tested by enough comparable outcomes.

Small samples produce wide uncertainty

Suppose the company wins six of twenty decided opportunities. The observed win rate is 30%. Using a Wilson interval, a method NIST recommends for binomial proportions and particularly considers suitable for smaller samples, the approximate 95% interval is 14.5% to 51.9%. The interval does not mean the metric is useless. It means the company should not treat 30% as a precise, stable property of the offer after only twenty decisions.

The weekly report should therefore show the denominator beside every rate. “Win rate: 30%” is incomplete. “Win rate: 6 of 20 decided opportunities, trailing cohort” is interpretable.

Do not wait for statistical certainty before acting. Instead, match the strength of the action to the strength of the evidence.

A single project that required twice the planned hours may justify an immediate review of what happened. It may not justify repricing every future engagement. Three similar overruns caused by the same missing customer data probably justify changing qualification, onboarding requirements, or scope.

Use internal comparisons before external benchmarks

Public win-rate and sales-cycle benchmarks often mix different opportunity definitions, customer segments, price bands, channels, and stages. Two organizations can publish different win rates while both calculate their metrics correctly because one begins the denominator at initial opportunity creation and another begins after proposal.

The initial benchmark should therefore be the company’s own comparable history:

• this offer version versus the previous version;
• referral opportunities versus targeted outbound;
• one customer segment versus another;
• standard engagements versus exception-heavy engagements;
• planned versus forecast versus final margin.

Historical project data also provides an “outside view” for future estimates. Reference-class forecasting recommends comparing a proposed project with the actual outcomes of similar completed projects rather than relying only on a fresh, inside-the-project estimate. The approach is valuable because people can underestimate cost and duration when they treat each project as unique. Research also cautions that the quality of the reference class matters and that optimism bias is not the only possible cause of overruns.

For a productized service, the practical version is simple: estimate the next engagement using the distribution of hours, duration, and change requests from genuinely comparable completed engagements.

A public example of the sales-delivery connection

Large public contractors provide a clear example of why contract value and win rate cannot be separated from delivery risk.

In its 2025 annual filing, Parsons stated that fixed-price contracts can offer higher margin opportunities because the contractor benefits from cost savings, but they also create greater financial risk because the contractor bears cost overruns. The company also explained that long-term contract profit depends on estimates of transaction price and total completion cost, which are reassessed as work progresses; revisions can produce positive or negative adjustments to reported results.

A small productized-service company is not operating at the same scale or under the same reporting rules. The operating lesson is nevertheless direct.

When the price is fixed and the promised result is defined, the supplier owns more of the estimation risk. Every sale therefore includes an implicit delivery forecast. If the company records only the contract value and not the estimated effort, schedule, and cost behind it, it cannot determine whether growth is improving or weakening the business.

The same filing notes that contract-type mix can affect profitability. That matters for a company whose “standard offer” quietly contains several commercial models—for example, some customers buying a fixed package, some receiving time-and-materials work, and others receiving open-ended support. A blended margin can conceal which model is actually repeatable.

The lesson is not that every company should adopt a public contractor’s accounting process. It is that price, scope, estimated cost, and actual execution must remain connected throughout the engagement.

Failure modes and the decision that follows

A weekly report can look complete while failing to support a decision. Several patterns deserve particular attention.

The denominator changes when the result looks bad

Opportunities are removed as “unqualified” only after they are lost, while equivalent wins remain in the dataset. No-decisions disappear from the report. Old opportunities are deleted rather than closed.

This raises the reported win rate without improving the offer.

Qualification rules should be applied before the outcome is known. Reclassification after closure should require a documented correction, not a sales judgment.

Sales cycles include only the fastest wins

Reporting average days to close for won opportunities ignores long-running losses and aging open deals. The number can improve simply because slow opportunities have not closed yet.

Report median time to decision for completed outcomes, won and lost cycles separately, and the age distribution of open opportunities.

ACV combines recurring and one-time work

A contract appears to have high annual value because implementation, customization, or project work was placed in the first year.

Separate recurring annualized value, one-time service value, and total contract value. For a one-time service, use average project value rather than forcing the work into a subscription measure.

Margin excludes the people who actually deliver the work

A project appears profitable because founder time, senior review, rework, customer-specific tooling, or presales effort is omitted.

Direct-cost rules should be consistent. Founder delivery work should be recorded at a reasonable replacement cost even when no additional cash wage is paid. Otherwise, the report measures the founder’s subsidy, not the offer’s economics.

Presales effort may be reported separately rather than charged to project margin, but it must appear somewhere in customer-acquisition economics.

Margin is measured only after completion

By the time final margin is known, the company has already absorbed the overrun.

Update the estimate to complete during delivery. The forecast margin should move as soon as the team learns that more effort, time, or outside cost will be required.

Scope changes are treated as ordinary delivery

The team says the project exceeded its budget because “the customer asked for more,” but the added work was never recorded, priced, rejected, or approved.

Record every material change. A paid change is a commercial expansion. An unpriced change is a delivery concession. An avoidable rework item is an execution problem. They should not share one category.

A blended average hides incompatible offers

A high-margin, fast-moving standard engagement is combined with a slow custom project. The average looks acceptable, but neither operating model is visible.

Segment by offer version and exception level. If custom projects require different qualification, pricing, or staffing, treat them as a separate offer rather than a variation hidden inside the standard package.

The report describes problems but changes nothing

The same variance appears for several weeks with no owner or action.

Every material exception should lead to one of a small number of decisions:

• change the target customer;
• change qualification;
• change the promise or exclusions;
• change the price;
• change the sales process;
• change onboarding requirements;
• change the delivery method;
• stop accepting a type of exception.

The work is complete when the company has more than a dashboard. It should have a governed metric dictionary, linked sales and delivery records, a weekly review routine, and enough evidence to compare what was sold with what was delivered.

The result should make the next decision clearer.

If win rate is improving while forecast margin is declining, the offer may be too easy to buy because scope or price is too generous. If margin is sound but win rate is weak, the problem may lie in customer selection, positioning, price communication, or qualification. If contract value rises while sales cycle and delivery variance rise faster, larger deals may be introducing complexity that the current offer cannot absorb. If referral opportunities outperform targeted outbound, the channels may be reaching different customers or carrying different levels of trust.

The company is ready to depend more heavily on the offer when comparable engagements show a coherent pattern: qualified customers buy it at an acceptable rate, sales cycles are understandable, contract values support the economics, forecast margins remain close to planned margins, and delivery variance is stable or narrowing.

That is the evidence that the company is no longer redesigning the work for every customer.

Sources

Primary sources

• U.S. Securities and Exchange Commission, “Beginners’ Guide to Financial Statements.”
• NASA, “Earned Value Management Tutorial.”
• NASA, Earned Value Management Reference Guide for Project-Control Account Managers.
• Parsons Corporation, 2025 Form 10-K.
• NIST, guidance on Wilson and Agresti–Coull confidence intervals for binomial proportions.
• Stripe, guidance on annual contract value and total contract value.
• HubSpot, “Sales Win Rate: How to Define, Calculate, and Improve It.”

Open research

• Shaik, Sridhar, Sriskandarajah, Mittal, and co-authors, “Opportunity Management for Business-to-Business Service Organizations: A Theory-Informed Decision Support Framework.”
• Yan, Gong, Sun, Huang, and Chu, “Sales Pipeline Win Propensity Prediction: A Regression Approach.”
• Rezazadeh, “A Generalized Flow for B2B Sales Predictive Modeling.”
• Flyvbjerg, “From Nobel Prize to Project Management: Getting Risks Right.”
• Chen, Ahiaga-Dagbui, Thaheem, and Shrestha, “Toward a Deeper Understanding of Optimism Bias and Transport Project Cost Overrun.”
• “Reference Class Forecasting: Promises, Problems, and a Research Agenda Moving Forward.”
• Rich and colleagues, “An Introduction to Survival Statistics: Kaplan–Meier Analysis.”