--- title: "Selected Evidence: The Wrong Decision Built From Correct Numbers" description: "Cherry-picking is the practice of presenting the period, segment or metric that supports a thesis while the wider set stays unpresented; because every figure is accurate, the presentation appears unchallengeable. The institutional cost surfaces in valuation and in capital allocation. The neutralising mechanism is not individual scepticism but fixing the measurement frame in writing before the result is seen, and recording the cuts that were prepared yet withheld." url: https://www.beirek.com/en/blog/cherry-picking-in-corporate-decisions canonical: https://www.beirek.com/en/blog/cherry-picking-in-corporate-decisions published: 2025-05-01 modified: 2025-05-01 category: "Organisational Psychology" category_url: https://www.beirek.com/en/blog/category/organisational-psychology language: en-US reading_time_minutes: 7 publisher: BEIREK LLC publisher_url: https://www.beirek.com license: "© BEIREK LLC — citation with attribution and link permitted" keywords: ["cherry-picking","measurement frame","investment committee decision quality","due diligence disclosure","capital allocation bias"] topics: ["Selection bias in corporate reporting","Investment committee governance","Diligence and valuation adjustment mechanics","Pilot-to-scale decision economics","Decision record and methodology discipline"] alternate_language_url: https://www.beirek.com/tr/blog/cherry-picking-in-corporate-decisions --- # Selected Evidence: The Wrong Decision Built From Correct Numbers > **In short:** Cherry-picking is the practice of presenting the period, segment or metric that supports a thesis while the wider set stays unpresented; because every figure is accurate, the presentation appears unchallengeable. The institutional cost surfaces in valuation and in capital allocation. The neutralising mechanism is not individual scepticism but fixing the measurement frame in writing before the result is seen, and recording the cuts that were prepared yet withheld. *The most expensive distortion in corporate reporting rarely originates in a false figure; it originates in how correct figures are selected. Where the choice of period, segment and metric is fixed only after the result is visible, the confirming evidence set assembles itself, and the debate becomes unwinnable precisely because the selection rule never appears on the page.* --- In a quarter-close presentation, the year at which a growth chart begins is almost never arbitrary. The left edge of the chart tends to coincide with the company's weakest trading period, since any curve originating at that point slopes upward; started one year earlier, the same dataset flattens noticeably, and started three years earlier it resolves into a horizontal band. The team preparing the deck may well have produced all three cuts, yet only one reaches the board table. No figure is wrong, no cell has been altered, and it is exactly for that reason that the ground for objection is narrow: whoever challenges the data must argue about the frame rather than the number, while the frame itself is nowhere visible in the material. The same pattern operates far beyond financial reporting, and usually more quietly. In retail, the decision whether performance is measured on a like-for-like store base or with new openings included can determine the direction of the result on its own. A customer satisfaction score computed only across respondents excludes, by construction, the segment least inclined to respond. A supplier reference list composed of three clients the supplier selected carries information not about the supplier's performance but about the supplier's best performance. In a comparable-transactions table reaching an investment committee, which deals were admitted as "comparable" is more determinative of the outcome than the multiple that emerges from them. The behaviour in question is cherry-picking — the practice of isolating the period, segment or metric that supports a thesis from the broader set already produced, and presenting only that. The mechanism operates on two levels. Cognitively, a mind testing a hypothesis is disposed to search for evidence that confirms it rather than evidence that would break it, so the supporting cut is found earlier while moving through a dataset, recalled more readily, and experienced as more persuasive. Organisationally, the person making the selection is typically the person held accountable for the result; where the team preparing the presentation is also the team being assessed within it, the choice of frame ceases to be a neutral methodological question and becomes an element of a performance defence. Recognising that this tendency is not an error is a precondition for managing it. Board attention is a scarce resource and no presentation can carry an entire dataset; every act of reporting is necessarily an act of selection, and a deck that selects nothing produces no decision. Foregrounding the favourable cut is rational in the short run for a unit competing for resources, given that a unit presenting the same information exhaustively is systematically disadvantaged against a unit presenting it selectively. The difficulty lies not in selection but in the timing of the selection rule: fixed before the result is visible, it constitutes methodology; fixed after the result is visible, it constitutes advocacy, and the two outputs are indistinguishable on paper. The most visible surface of the institutional cost is valuation. In a sale process the information package assembled by the seller consists, definitionally, of cuts the seller chose; the buyer's diligence team, in turn, begins by widening them — monthly rather than quarterly, segment-level rather than consolidated, cash collected rather than revenue contracted. Where the figures deteriorate under that widening, the buyer reads the deterioration not merely as a numerical correction but as a signal about disclosure conduct. The pattern typically observed is that the discount scales less with the magnitude of the variance than with the manner of its discovery: a weak cut volunteered by the seller enters the price, whereas the identical weak cut excavated by the buyer enters the price and additionally reopens the scope of representations and warranties, the escrow percentage and the earn-out thresholds. The second surface is capital allocation itself. Where every project arrives at an investment committee wrapped in its own most favourable frame, comparison across projects becomes technically impossible, and the committee begins ranking not the projects but the quality of the framing. On the balance sheet this shows up as capital flowing toward the team that constructs its frame most effectively rather than toward the asset promising the highest return, and the preference compounds across several budget cycles until it depresses the portfolio's return profile structurally. The effect is invisible in any single decision; placed side by side, however, the realised returns of a three- or four-year investment programme and the projections held at approval diverge in a direction that is almost always the same. The third surface is the pilot, where the cost is noticed latest. An operational transformation, a new production-line discipline or a systems implementation is naturally launched at the readiest site, under the strongest local management and against the cleanest data; to the extent this raises the probability of the pilot succeeding, it is a defensible choice. Once the scaling decision is taken on the pilot's unit economics, however, a cost and schedule assumption that does not hold at the remaining sites is carried into the entire programme. The familiar overrun pattern in rollout budgets frequently begins here: what is exceeded is not the budget but the sample on which the budget rests. The fourth surface sits on the contractual and credit side. Covenant calibration rests on the historical performance series the borrower supplies; where that series begins, and which unusual quarters have been "normalised" out of it, determines how much genuine headroom the thresholds actually carry. In the same way, an EPC contractor's schedule performance presented solely across completed projects leaves cancelled and suspended work outside the sample, so the resulting estimate of delay risk is structurally optimistic. Calibration errors of this kind are invisible at signing; they surface as a technical breach in the second or third reporting period after first draw, when the cure options are narrower and more expensive. The mechanism that neutralises the tendency is not individual scepticism or data literacy but an institutional architecture in which the choice of frame is separated from the decision, and it has four distinct components. The first is fixing the measurement frame in advance: in any investment or performance discussion, the period range, the segment definition and the metric set to be relied upon are committed to writing before the result is seen. The second is logging the cuts produced but not presented — a single paragraph appended to the material stating which alternative cuts the underlying analysis contains and why they were left out. The third is separating ownership of the frame from ownership of the result, so that the role defining the methodology and the role defending the performance do not converge in one person. The fourth is a counter-cut discipline, under which the cut showing the thesis at its weakest is required to appear on the same page as the thesis. Across the project and portfolio processes BEIREK runs, this architecture is operated as a recording discipline rather than as something left to the quality of a presentation. Before an investment decision or a capital allocation discussion opens, the measurement frame — period range, segment definition, metric set and normalisation rules — is fixed in a one-page decision record, and the decision itself is taken by reference to that record; the frame may change during the process, but any change is logged together with its rationale and therefore does not remain invisible afterwards. The same discipline applies on the diligence side through a question set aimed not at the cut supplied by the counterparty but at the selection rule that produced it: which period was excluded and on what basis, which contracts were not admitted as comparables, which site was kept out of the pilot. The operational counterpart of this intervention is a change in the review rhythm itself. Performance cuts presented in project progress meetings are held constant so as to remain comparable with those presented in the preceding period, and where the definition of a metric changes, the series computed on the old definition is carried alongside it for at least two periods. In scaling decisions, the dimensions along which the pilot site is not representative — local management capacity, data quality, supply distance, labour turnover — are written explicitly into the rationale, and the rollout budget is built not on the pilot's unit economics but on a base corrected for those deviations. This is not an approach that requires teams to be more candid; it constructs a setting in which selective framing remains visible and therefore ceases to pay, even in the short run. The decision quality of an institution can be assessed less by the accuracy of the analyses it produces than by the fate of the analyses it produces and does not present. A wrong decision assembled from correct numbers is the hardest kind to learn from later, precisely because it leaves no audit trail; and the most productive question available at a management table concerns not what the cut in front of it says, but which cuts were prepared and never made the table. ## Key Points - Selected evidence leaves no audit trail, since nothing in the data has been falsified; the distortion sits in the frame around the number rather than in the number itself. - When the starting year of a chart, the definition of a segment and the choice of metric are settled after the result is known, the deck ceases to be an analysis and becomes a defence document. - In a diligence process the largest discount is generated not by weak figures but by figures that deteriorate in the cut the counterparty did not choose to present. - Running a pilot in the most favourable site systematically shifts the unit economics underlying the scaling decision toward the optimistic side. - The mechanism that neutralises the tendency is fixing the measurement frame before the decision and logging the alternative cuts that were produced but withheld. ## Questions ### What distinguishes cherry-picking from ordinary data summarisation? The distinction lies in when the selection rule is fixed. All reporting is necessarily selective; where the rule is settled before the result is visible, the output is a methodological decision. Where the rule is settled after the result is visible, the identical output functions as a defence document. Since neither case leaves a distinguishing mark on the presentation itself, the only way to produce the distinction is to commit the frame to writing in advance. ### How is selectively presented performance data identified during diligence? By widening the cut. Requesting monthly rather than quarterly series, segment-level rather than consolidated figures, and cash collected rather than revenue contracted will usually surface selective framing at the first widening. The more efficient approach interrogates not the cut supplied but the rule that produced it: which period was excluded and on what grounds, which precedents were not admitted as comparable, which customers are absent from the sample. ### How does this tendency affect company valuation? Through two channels. The direct channel is the price effect of figures that deteriorate once the cut is widened. The indirect and generally costlier channel is the disclosure signal: weak data volunteered by the seller affects price alone, whereas the same data excavated by the buyer reopens the scope of representations and warranties, the escrow percentage and the earn-out thresholds, thereby altering the structure of the cash consideration as well. ### How is an institutional mechanism against this tendency constructed? Through four components: fixing the measurement frame in writing before the decision; logging the alternative cuts prepared but withheld; keeping the role that defines methodology separate from the role that defends performance; and requiring the cut least favourable to a thesis to appear on the same page as that thesis. Because these rest on records and role separation rather than individual resolve, they survive changes in personnel. --- Source: https://www.beirek.com/en/blog/cherry-picking-in-corporate-decisions Publisher: BEIREK LLC — https://www.beirek.com