How to describe buyer outcomes when the baseline is unknown

Claim discipline

Name the kind of claim before choosing the number

Start with a small evidence ledger. A buyer fact comes directly from the procurement record or an authoritative clarification: for example, the stated objective is to reduce invoice exceptions, or the disclosed annual volume is 600,000 transactions. Supplier evidence describes something observed in a defined reference context. A scenario calculates what might happen if stated inputs hold. A proposed target is a future level for joint approval. A commitment states an obligation the supplier accepts. A measured result can exist only after an agreed method has observed performance. These labels are not cosmetic. They determine what the sentence may honestly say.

Unknown does not mean zero, average or whatever makes the model work. If the buyer has not provided handling time, state that it is unavailable. If a public annual report provides a related figure, cite it and explain whether its scope matches the procurement; do not silently treat it as tender data. If an incumbent case study recorded a 22 percent reduction, keep the original population, intervention, period and conditions beside that result. It can support plausibility, but it does not reveal this buyer’s starting point or predict the same change.

The wording should carry the epistemic status without forcing the evaluator to inspect a footnote. “The solution will save CHF 1.8 million” is not supported when the baseline is assumed. “Using the illustrative volumes and rates in Table 2, the scenario indicates annual capacity equivalent to CHF 1.2 million to CHF 1.8 million; the figure is not a buyer-specific saving and will be recalculated against the agreed baseline” is longer but accurate. Keep the qualification close to the number and repeat it in any executive summary that reuses the claim.

Scenario design

Show the economics as a model the buyer can challenge

Build from operational units instead of announcing a headline percentage. A simple capacity scenario might use eligible annual volume multiplied by current handling time, the plausible time reduction and expected adoption. A quality scenario might use eligible volume, current defect rate, avoidable share and cost per defect. Put every input in a table with its unit, range, source, owner and status. Use buyer-supplied numbers where available; otherwise mark supplier assumptions explicitly. Let the evaluator replace them. A model that can be challenged is more useful than a polished number whose inputs are hidden.

Use at least a conservative case and a higher case, but do not select ranges merely to frame the preferred result as moderate. Explain why the bounds are plausible and test the variables that matter. If adoption falls from 80 to 40 percent, show the effect. If only part of the time released can reduce cost, distinguish gross hours, redeployable capacity and cashable saving. Do not add all three together. Account for implementation cost, operating cost, ramp-up, benefit delay and process work the buyer must complete. A benefit that requires a new policy, clean master data or workforce change should carry that dependency.

The baseline is also a view of what would happen without the investment, not only a photograph of today. Demand may grow, an existing transformation may already reduce errors, or a contract may expire. State the business-as-usual assumption and keep it separate from the proposed intervention. The UK Green Book describes appraisal as comparison of options against business as usual and requires uncertainty to be communicated clearly. It also distinguishes appraisal before implementation from evaluation afterwards. That distinction is practical here: the tender response may appraise a plausible outcome, while post-award evidence determines what was actually realised.

• Use operational variables with visible units and ranges.
• Show who supplied each input and whether it is verified.
• Separate capacity, cost avoidance and cashable saving.
• Include implementation cost, delay and buyer-owned actions.
• Test the assumptions that move the result most.

Outcome logic

Connect what the supplier delivers to what the buyer values

A credible response explains the chain between intervention and outcome. Supplier configuration, integration and training are inputs or activities. Automated classification and an exception queue are outputs. Higher straight-through processing and faster resolution may be intermediate outcomes. Lower avoidable handling cost or improved service timeliness may be buyer outcomes. Write the chain in that order and give each link a test. This prevents a feature from being promoted directly into a financial benefit without showing the operating change in between.

Split control honestly. The supplier can commit to delivery milestones, system availability or a configured workflow when those are within its authority. It may influence adoption, cycle time and error reduction but cannot own them alone when buyer staffing, policy, demand, data or user behaviour also matters. Name a benefit owner on the buyer side and a delivery owner on the supplier side. For every material dependency, state the action, owner, due point, evidence and effect if it does not occur. “Subject to adoption” is too vague to manage.

Reference evidence at the link it supports. A benchmark may support the plausible processing range. A controlled pilot may support the effect of one workflow under defined conditions. A customer example may show that adoption was feasible in a comparable environment. None proves the complete buyer outcome by itself. Where evidence is weak, reduce the strength of the claim and propose an early test. This approach makes the value case more persuasive because the buyer can see where confidence comes from and what will be learned, instead of being asked to accept optimism as evidence.

Measurement plan

Turn the unknown baseline into an early controlled decision

Define the baseline task in the proposal rather than promising to work it out later. Specify the population, inclusion and exclusion rules, measure, unit, data source, comparison period and segmentation. State how missing records, outliers, seasonality and process changes will be handled. Identify the buyer data owner, benefit owner, supplier analyst and approval authority. Set a short discovery window and a date on which the parties will approve the baseline, revise the scenario and agree any operational target. The UK government guide to benefits management calls for baseline information and the circumstances behind it to be documented in a benefits realisation plan, with benefit ownership made explicit.

Decide how the result will be compared. Before-and-after data can describe change, but it may not show what would have happened without the solution. The Magenta Book’s guidance on impact evaluation treats the counterfactual as central to attributing impact. A tender does not always justify an experimental design, yet the response should not claim causation that its method cannot establish. Where a comparison group is impractical, disclose the limitation, track major external changes and use a proportionate combination of operational data, sampling and finance validation.

After baseline approval, preserve the original scenario and record the revision rather than overwriting history. Report forecast, agreed target and measured result in separate columns. Explain variance, including demand, adoption, data quality and buyer actions. A target becomes contractual only through authorised wording and governance, not because it appears in a benefit table. Before submission, commercial and legal reviewers should inspect any word such as “will,” “guarantee,” “minimum” or “saving,” ensuring the obligation matches actual control, pricing and remedy. The goal is a value case the delivery team can inherit without discovering that its most impressive number never existed.

• Define the cohort, period, measure and source before award.
• Name the buyer and supplier roles that will approve the baseline.
• Keep scenario, target, commitment and result in separate fields.
• Disclose limits on causal attribution.
• Revalidate the wording when the value case enters the contract.