Turn Past Outcomes into Proof

Task

Build an initial proof library from consulting outcomes.

Summary

Build a proof library from real results, evidence, and before-and-after narratives.

Turn Consulting Results Into Proof Customers Can Trust

Task ID: S1-10

Good consulting work often leaves behind scattered evidence: an approving email, a metric in a project deck, and a customer willing to speak, but no reusable proof. This article explains how to convert those outcomes into a small, verified library of case studies, testimonials, return-on-investment examples, and before-and-after narratives that supports better sales and a clearer business strategy.

The problem is not a lack of good work

A consulting engagement ends well. The customer is satisfied. The delivery team knows the work saved time, removed a costly problem, or made an important project possible. Six months later, a prospect asks for evidence.

Sales finds a positive quotation in an email, a slide showing an unexplained percentage, and a customer name that nobody is sure the company still has permission to use. The founder can tell the story from memory, but another employee cannot reconstruct it. The result is treated as proof in one sales conversation and ignored in the next.

This is not primarily a content-production problem. It is an evidence-management problem.

An initial proof library converts completed customer work into a controlled set of records that answers five questions:

1. What kind of customer had the problem?
2. What was happening before the engagement?
3. What did the company deliver?
4. What changed, and how was that change measured?
5. What has the customer approved for public or private use?

The operating principle is simple: build proof from delivery records first, then turn it into marketing material.

That order matters. When the case study is written before the facts are reconstructed, the writer tends to begin with the desired claim and search for supporting details. When the evidence record comes first, the company can distinguish what it knows from what the customer believes, what the numbers show, and what remains an assumption.

This work belongs early in a consulting-led company’s development because the company is still deciding which customer, problem, and result deserve more attention. At this stage, proof does more than reassure prospects. It helps management see whether several projects support the same business direction.

Research on customer references in business-to-business markets supports this broader role. A mixed-method study by Harri Terho and Anne Jalkala found that companies use references externally in sales and marketing, but also internally to learn from customers, identify business opportunities, and develop better practices. The study reported an association between systematic customer-reference marketing and selling performance, while also noting that the effect depends on context rather than following a universal formula.

That distinction is important. A proof library should not be built merely to make the company look successful. It should help answer whether the same type of customer repeatedly buys help with the same problem and receives a result that can be described, measured, and delivered economically.

What the library must prove

A complimentary quotation is useful, but it is not the same as proof of an outcome.

“Excellent team to work with” says something about the customer’s experience. It does not establish what problem was solved, what changed, how much the change was worth, or whether another customer should expect a similar result.

A useful proof library contains different assets because buyers ask different questions. Research on business-market references suggests that potential customers use them to judge a supplier’s competence, reduce perceived buying risk, forecast possible financial returns, discover needs they had not fully recognized, and learn from another organization’s experience. That research was exploratory and based on a single case, so its propositions should not be treated as settled causal findings, but the categories provide a practical description of what buyers seek from references.

The main proof formats serve related but distinct purposes:

The formats can reinforce one another. One engagement might produce a full case study, an approved quotation, a one-page before-and-after summary, and a sales slide. Those are four assets, but they are not four independent pieces of evidence. They are four expressions of one underlying customer record.

That difference prevents count inflation. A company with twenty derivative files based on two customer outcomes does not have twenty separate proofs that its offer works.

Relevance matters as much as positivity. Three experimental studies of supplier-selected referrals found that the best choice of referrer and message depended on the prospect’s uncertainty and on whether the supplier was already established with the buyer. In some situations, a highly credible source with a balanced account could be more useful than an entirely positive message; in others, the positive message carried more weight.

For a small proof library, this means the company should not simply select its happiest or most famous customers. It should select evidence that resembles the customers, problems, buying concerns, and results the company expects to pursue.

A relevant mid-sized customer with a clearly documented result may be more persuasive than a recognizable logo whose project was unusual, heavily customized, or unrelated to the current offer.

Build from delivery records, not marketing copy

The most reliable starting point is the completed engagement, not a blank case-study template.

The company should reconstruct each candidate outcome from records created during sales and delivery. Depending on the project, those records may include the proposal, statement of work, discovery notes, project plan, baseline reports, system data, time records, invoices, change requests, acceptance documents, customer emails, meeting notes, and post-project reviews.

The process should turn that material into one canonical proof record before producing public-facing assets.

flowchart LR
    A[Completed engagement] --> B[Identify a candidate outcome]
    B --> C[Verify the baseline, result, and data source]
    C --> D{Customer approves use?}
    D -->|Yes| E[Create the canonical proof record]
    D -->|No| F[Keep internal or anonymize]
    E --> G[Adapt into case study, quote, ROI example, or sales slide]
    G --> H[Track use, approval, freshness, and results]

The diagram shows the governing sequence: verify the outcome and secure permission before distributing derivative materials. A result that cannot be published may still be useful internally, but it should not quietly enter external presentations.

Start with candidate selection. Review completed work and look for engagements with a clear problem, a recognizable starting condition, evidence of change, a customer willing to participate, and relevance to the market the company may build around. Do not require perfect financial measurement at this stage. A well-supported operational result may be more credible than a speculative dollar figure.

Also include less obvious candidates. The most instructive project may not have produced the largest percentage improvement. It may have exposed a problem that customers repeatedly describe, a delivery method that the team can repeat, or a result that customers value enough to renew or buy additional work.

Reconstruct the fact pattern. For every candidate, record:

• the customer segment, industry, size, operating context, and relevant buyer role;
• the initial problem and why it mattered at that time;
• the baseline measure, its definition, source, and date;
• the scope of work, important exclusions, and customer responsibilities;
• the implementation period and major changes made during it;
• the resulting measure, its definition, source, and date;
• other events that could have influenced the result;
• the customer’s total cost and internal effort, where financial claims will be made;
• the named people who can verify the facts.

The baseline and result must use comparable definitions. A reduction in “resolution time” means little when the starting figure measured all support cases and the later figure excluded complex cases. A revenue comparison is weak when the earlier period covered three months and the later period covered six. A percentage improvement without the underlying numerator, denominator, and time period is difficult to audit and easy to misread.

Interview both sides. The customer interview should explore the problem in the customer’s own language, the alternatives considered, the experience of implementation, the result, the burdens or compromises involved, and what did not change. The internal delivery lead should separately explain what the team did, which parts were decisive, which parts were routine, and which parts depended on unusual customer conditions.

Separate interviews reduce the risk of creating one polished but untested account. Differences between the two versions are useful. They often reveal that the customer valued a different part of the engagement than the delivery team expected.

Verify the numbers with their owner. A marketing employee should not approve an operational metric simply because it appears in a slide. The person who owns or understands the source system should confirm the measure, period, calculation, and appropriate wording. Financial claims should also be reviewed by someone capable of checking the economic assumptions.

Secure explicit approval. Approval should cover the exact quotation, metrics, customer name, logo, speaker, channels, geographical use, and duration of permission. Approval to tell a story privately during a sales call does not necessarily authorize publication on a website. Approval to use a logo does not automatically authorize a detailed financial case study.

A maintained proof register should then make every record findable and governable. Useful fields include the proof name, customer segment, problem, result, asset type, source files, owner, approver, intended audience, approved channels, status, publication date, next review date, permission expiry, location, version, and notes. Add an evidence rating and the exact claim each asset is allowed to support.

A public example of this operational approach appears in GitLab’s handbook. Its customer-advocacy function describes an ongoing pipeline of case studies, blogs, videos, quotations, return-on-investment metrics, and approved logo uses. It also maintains centralized trackers, allows users to filter stories by factors such as industry, region, and return on investment, and states that customer references should be secured in writing.

The lesson is not that a small company needs the same organization or production volume. It is that proof should have a source of truth, selection rules, written permissions, accountable owners, and a way to retrieve the right evidence for a particular buyer.

Make financial proof honest

Financial proof is powerful because it connects delivery to a business result. It is also the easiest form of proof to overstate.

A basic return-on-investment calculation is:

The arithmetic is simple. Deciding what belongs in “verified benefit” and “total customer cost” is not.

Total cost may include consulting fees, software, implementation expenses, customer employee time, data preparation, training, process changes, travel, and continuing operating costs. Omitting customer effort can make an intensive engagement appear more economical than it was.

Benefits should be separated by type rather than collapsed into one impressive number:

Consider a claim that a new process saves 500 employee hours a year. Multiplying those hours by a loaded wage rate can estimate the value of capacity released. It does not prove that the customer reduced payroll by that amount. The more accurate statement may be that employees redirected 500 hours to other work, avoided overtime, or handled additional volume without another hire.

Revenue claims require similar restraint. “Revenue influenced” is not the same as “revenue caused.” A consulting project may have contributed to a sale while brand reputation, pricing, product changes, market conditions, and the customer’s own employees also contributed.

The same discipline applies to before-and-after stories. A change observed after an engagement is evidence that the outcome occurred. It does not, by itself, establish that the engagement caused all of it.

The United Kingdom government’s current evaluation guidance explains that attribution requires an estimate of what would have happened without the intervention—a counterfactual—and that monitoring results serves a different purpose from evaluating additional impact. Strong experimental or well-matched comparison designs provide more confidence, while a single before-and-after comparison without a suitable control offers weak evidence of causation because other factors may explain the change.

Most consulting case studies will not justify or require an experiment. The practical response is not to discard their evidence. It is to label the strength of the claim honestly.

A useful internal evidence scale is:

• Verified outcome: The before-and-after measures come from an identified customer system or record, use comparable definitions, and have been confirmed by the customer.
• Supported contribution: The outcome is verified, the engagement has a clear connection to it, and competing explanations have been considered, but causation has not been isolated.
• Customer estimate: The result is based on the customer’s informed calculation or recollection rather than a directly audited source.
• Working hypothesis: The result appears plausible but lacks enough verification for an external claim.

Only the first three normally belong in a proof library, and the wording should disclose which type is being used. Working hypotheses belong in the research backlog.

A strong financial example also exposes its assumptions. Readers should be able to see the period measured, the calculation, what costs were included, whether the result is annualized, and whether it reflects actual cash, released capacity, an avoided expense, or estimated risk.

What a usable first portfolio looks like

The working target of three to five usable assets is a practical starting point, not an established industry benchmark.

The appropriate number depends on the company’s market, sales cycle, price, range of customers, complexity of implementation, confidentiality restrictions, and quality of the available data. A company selling one narrow service to similar customers may learn a great deal from three strong records. A company serving several industries with materially different problems may need more before it can claim a repeatable pattern.

The key word is usable. An asset is usable when it is:

• relevant to a defined customer, problem, result, or buying concern;
• supported by evidence that another employee can inspect;
• approved for the intended type of use;
• specific enough to help a prospect judge fit;
• stored where sales and marketing can find it;
• current, versioned, and assigned to an owner;
• written so that its limits and assumptions are not hidden.

A sensible initial mix might include one detailed case study, one concise before-and-after narrative, one transparent financial example, and one or two short testimonials or proof cards. Another company may choose two deep cases from similar customers because the immediate question is repeatability rather than content variety.

The library should create coverage, not merely volume. Place the candidate records side by side and ask:

• Do several customers describe the same underlying problem?
• Are the customers similar enough to form a meaningful segment?
• Do they value the same result?
• Does the company produce that result through substantially similar work?
• Can the result be measured in comparable terms?
• Are the conditions required for success visible?
• Did the work produce an acceptable margin without depending on exceptional founder effort?

The last question keeps proof connected to business design. A customer may have received an excellent result from a project that was unprofitable, impossible to repeat, or dependent on one senior person working nights. That is proof of capability, but not necessarily proof of a scalable offer.

One well-structured public case illustrates the anatomy of a useful record. In an AWS case study, Canadian rehabilitation provider CBI Health is described as needing to move 70 terabytes of data and its physical data-centre infrastructure before a contract renewal. The published account identifies the intervention, reports that more than 500 servers were migrated or decommissioned, states that the migration was completed in three months, and attributes 60% infrastructure cost savings and a 75% reduction in technology footprint to the resulting change.

The example is useful because it connects a defined starting condition, a deadline, specific work, a time period, and several measured results. It should still be read for what it is: a vendor-published customer story, not an independent audit or controlled attribution study. Its claims therefore teach a content structure; they do not establish that every similar migration will produce the same result.

The initial portfolio should be measured with more than an inventory count. The primary measure remains the number of proof assets, but management should also track:

Track the underlying proof records separately from derivative formats. A case study, sales slide, social excerpt, and presentation quotation based on the same customer should remain linked to one evidence record. When a metric changes or permission expires, every derivative can then be corrected or withdrawn.

Usage data also helps refine the library. If sales repeatedly uses one case because it answers a specific objection, the company may need more proof around that concern. If a case is never used, the problem may be poor retrieval, weak relevance, unclear writing, or a market direction that does not matter to current prospects.

Failure modes and judgment calls

A proof library can look complete while remaining commercially weak.

Collecting praise instead of outcomes. A folder full of positive quotations can support trust, but it does not answer whether the company repeatedly solves a valuable problem. At least some records must connect the customer’s starting condition, the work, and an observed result.

Starting with the desired claim. Asking, “Can we say the project saved 30%?” encourages number hunting. Begin instead with the source records and ask what they establish.

Using percentages without denominators. “Twice as fast” is incomplete without the original duration, later duration, process measured, and time period. Small denominators can also produce dramatic percentages that do not represent material business value.

Treating all post-project improvement as caused by the consultant. Customers change employees, budgets, systems, policies, and priorities during projects. A credible account identifies important co-contributors and uses language such as “contributed to,” “helped achieve,” or “followed by” when stronger attribution is unavailable.

Choosing only famous logos. A recognizable customer may attract attention while providing little useful evidence for the intended buyer. Selection should balance name recognition with relevance, data quality, repeatability, and the customer’s willingness to participate.

Hiding implementation demands. A case study that omits the customer’s data cleanup, internal project team, process changes, or adoption effort creates a misleading impression of what the result requires. Conditions of success make the evidence more useful, not less persuasive.

Counting formats rather than evidence. Repackaging one quotation into ten graphics does not strengthen the underlying proof.

Leaving ownership and permission unclear. Assets become unsafe to use when nobody knows who approved them, what wording was approved, where the source data lives, or when the approval should be reviewed.

Overusing the same customer. Reference calls, conference appearances, interviews, approvals, and repeated quotation requests consume customer time. The library should track advocate activity and protect the relationship, particularly when only a few customers are available.

Confusing curation with distortion. A company may select its strongest relevant examples, but it must not fabricate experiences, remove material qualifications, or present a selected set as though it represents every customer’s result.

In the United States, the Federal Trade Commission’s endorsement guidance states that testimonials must be truthful and not misleading, that material relationships should be disclosed, and that an unrepresentative result may require information about what customers can generally expect. Its consumer reviews and testimonials rule, effective October 21, 2024, addresses fake or false testimonials and other deceptive practices; the agency also warns businesses not to supply testimonial language without a reasonable basis for believing it accurately reflects the customer’s experience.

Canadian guidance similarly states that businesses should secure written authorization before using a third party’s testimonial, preserve the meaning and qualifications of the approved statement, and avoid using a testimonial as a substitute for adequate support for a broader performance claim.

The practical requirements will vary by jurisdiction, audience, contract, and claim. The operating rule is broader: retain the source, preserve the customer’s intended meaning, disclose relevant limits, and obtain approval for the actual use.

The cost of crossing that line is not theoretical. In 2022, the Federal Trade Commission finalized a $4.2 million settlement with Fashion Nova after alleging that the company suppressed product reviews below four stars while representing the displayed reviews as the views of all purchasers who had submitted them. The case concerned consumer product reviews rather than consulting case studies, but it demonstrates the difference between selecting evidence for a particular purpose and misrepresenting a curated record as complete.

The final judgment call is whether three to five assets reveal a pattern or merely document several unrelated successes. A small library should not be forced to support a conclusion it does not contain.

The decision this work should make possible

The proof library is ready to support the next business decision when another employee—not only the founder—can retrieve several approved records and explain:

• which customer had the problem;
• why the problem mattered;
• what the company delivered;
• what changed;
• how the result was measured;
• what the customer contributed;
• how strong the attribution is;
• what the work cost to sell and deliver;
• which parts can be repeated for another customer.

When several records point to the same customer type, problem, result, and delivery pattern, the company has stronger grounds for building an offer around that pattern. It can use the evidence to sharpen sales conversations, define scope, improve pricing, identify required customer conditions, and decide what part of the work may eventually become a repeatable service or product.

When every record describes a different customer, a different problem, a different delivery method, and an incomparable result, the library has still done valuable work. It has shown that the evidence does not yet support one clear direction.

That is better than publishing confidence the business has not earned. The purpose of the first proof library is not to prove that every engagement succeeded. It is to make clear what customers have already paid for, what results can be defended, and whether a repeatable business is beginning to appear.

Sources

Open research

• Terho, Harri, and Anne Jalkala. “Customer Reference Marketing: Conceptualization, Measurement and Link to Selling Performance.” Industrial Marketing Management, 2017.
• Morgado, Andre Vilares. “The Value of Customer References to Potential Customers in Business Markets.” Journal of Creating Value, 2018.
• Hada, Mahima, Rajdeep Grewal, and Gary L. Lilien. “Supplier-Selected Referrals.” Journal of Marketing, 2014.

Primary and official guidance

• HM Treasury. Magenta Book: Central Government Guidance on Evaluation, updated May 15, 2026.
• HM Treasury and Evaluation Task Force. Quality in Policy Impact Evaluation, updated May 15, 2026.
• U.S. Federal Trade Commission. “The FTC’s Endorsement Guides: Being Up-Front With Consumers.”
• U.S. Federal Trade Commission. “The Consumer Reviews and Testimonials Rule: Questions and Answers.”
• Competition Bureau Canada. “Use of Tests or Testimonials.”
• Competition Bureau Canada. “Performance Claims Not Based on an Adequate and Proper Test.”
• Competition Bureau Canada. “False or Misleading Representations and Deceptive Marketing Practices.”

Public operating examples

• GitLab. “Customer Advocacy at GitLab.” GitLab Handbook.
• Amazon Web Services. “Increasing Migration Speed by 40% and Reducing Costs by 60% Applying AWS Experience-Based Acceleration at CBI Health.”
• U.S. Federal Trade Commission. “FTC Finalizes Order with Fashion Nova Over Allegations It Blocked Negative Reviews,” March 2022.