--- title: "The Result That Stays in the Drawer: When Institutional Memory Contains Only What Worked" description: "The file-drawer problem is the failure of null or negative results to enter the record at all; in corporate terms it surfaces as the same pilot being funded twice a few years apart and as a base success rate that reads higher than it is. The neutralising mechanism is registering the start of an experiment rather than mandating disclosure of its outcome." url: https://www.beirek.com/en/blog/file-drawer-problem canonical: https://www.beirek.com/en/blog/file-drawer-problem published: 2025-05-02 modified: 2025-05-02 category: "Organisational Psychology" category_url: https://www.beirek.com/en/blog/category/organisational-psychology language: en-US reading_time_minutes: 8 publisher: BEIREK LLC publisher_url: https://www.beirek.com license: "© BEIREK LLC — citation with attribution and link permitted" keywords: ["file-drawer problem","institutional memory","decision log","base rate distortion","due diligence records"] topics: ["Organisational learning and knowledge retention","Investment committee decision calibration","Due diligence and transaction structuring"] alternate_language_url: https://www.beirek.com/tr/blog/file-drawer-problem --- # The Result That Stays in the Drawer: When Institutional Memory Contains Only What Worked > **In short:** The file-drawer problem is the failure of null or negative results to enter the record at all; in corporate terms it surfaces as the same pilot being funded twice a few years apart and as a base success rate that reads higher than it is. The neutralising mechanism is registering the start of an experiment rather than mandating disclosure of its outcome. *When an organisation's records consist exclusively of initiatives that produced a result, its internal estimate of its own success rate drifts systematically upward. A pilot that yields nothing and is never written up reflects not individual negligence but the predictable consequence of an asymmetric distribution of reporting cost.* --- The number of pilots presented at a board's year-end portfolio review is typically smaller than the number actually launched during that year, and the gap does not close through an agenda item. An initiative that produces a result becomes visible through a slide, a closing meeting and, more often than not, a scaling plan, while an initiative that produces nothing ends without any closing act at all — simply by not appearing in the next presentation. No one announces that the pilot was cancelled; no one commits the cancellation to a document; the effort dissolves quietly, either as the budget line runs out or as the responsible manager moves to another assignment. What remains is a list not of what the organisation attempted, but of what it accomplished. The same pattern appears at smaller scale inside routine operations. Where a procurement team ran a sample trial with an alternative supplier two years ago and the result was unfavourable, that result generally sits in the email archive of whoever owned it at the time; the organisation's supplier evaluation record shows neither the trial nor its outcome. A line setting tested on the production floor and reverted, a pricing model run for a quarter on the commercial side and then dropped, a recruitment channel opened and closed in human resources — each consumed cost and generated knowledge without generating any record. The organisation holds an accounting entry for the money spent and no entry at all for what the money bought in understanding. This pattern carries a name — the file-drawer problem, the tendency of a study yielding null or negative findings to remain unreported and unshelved — and its mechanism is less cognitive than it is a matter of incentive architecture. Reporting is costly regardless of the direction of the finding: data has to be assembled, a rationale written, a position defended in a meeting. What that cost purchases, however, differs sharply by direction. A favourable finding, once reported, converts of its own accord into a resource request, an expansion of mandate or a gain in visibility, whereas an unfavourable finding, once reported, first creates a duty to explain. Under that asymmetry a rational manager does not suppress the negative result; the manager merely declines to prioritise the act of reporting it, which in practice amounts to the same thing. A second layer concerns ownership. A successful pilot has an identifiable owner, and ownership there functions as an asset; the owner of an unsuccessful pilot, the moment a record exists, assumes a liability instead. As the number of experiments inside an organisation rises, that liability accumulates, and the individual cost of experimenting increases independently of the institutional return on experimenting. The typical behaviour observed under these conditions is not a reduction in the number of experiments — the count often holds steady — but the conduct of experiments outside any formal frame, since work that never officially began cannot officially fail. These tendencies are functional under identifiable conditions, and the shortcut itself is not the issue. Requiring a formal report for each of the many small trials run at an early stage pushes the cost of experimenting above its informational value and suffocates exploratory capacity; demanding a record for the dozens of minor variations a team tries and discards within a weekly rhythm equates the cost of measurement with the cost of production. The problem arises when conditions change and the shortcut persists. Once the budget attached to a trial crosses a threshold, once the same trial becomes something several units could independently run, or once its outcome feeds an investment decision, the absence of a record ceases to be an economy and becomes a direct loss of information. The first institutional expression of that loss is repetition. Where no record exists of options tested and eliminated, the same alternative supplier goes to sample a second time some years later, the same market segment is tested again, the same software is piloted again — and the budget for that second round is no smaller than the first. The trace repetition leaves in the income statement is diffuse: partly advisory expense, partly staff time, partly sample and testing cost, distributed across separate lines, so the aggregate never surfaces as a single figure in any report. The one thing that makes it visible is the sight of the same question being treated, inside the same organisation, as a new question for the second time. The second expression is a corrupted base rate. When an organisation's record of past initiatives contains only those that were completed and produced results, the internal estimate of the probability that a new initiative will succeed drifts structurally upward, and that drift passes straight into budget calibration. An investment committee looking at the available record as its comparison set observes that most comparable past initiatives delivered; what it is actually observing, however, is not the historical success rate but the selection rule governing what got recorded. This structure accounts to a considerable degree for why the optimistic scenario persistently behaves as though it were the central case, and why the same organisation reproduces a similar deviation on each successive initiative. The third expression surfaces at the transaction table. In a diligence process, the layer a buyer reaches least easily is the inventory of what the target tried and walked away from, for the straightforward reason that such an inventory frequently does not exist in physical form. The buyer cannot distinguish whether an abandoned product line stopped for technical reasons or merely because a manager changed, whether a geographic market was tested and eliminated or never approached, whether the sole-source position of the incumbent supplier reflects a preference or a constraint validated in the past. That uncertainty is rarely deducted from headline value; it is typically priced through tighter earn-out thresholds, broader representation and warranty coverage, or a higher escrow ratio. From the seller's side the outcome is identical: the record that was never kept returns as a cost embedded in the closing structure. The fourth expression is dependence on founders and key personnel. Where knowledge of what was tried and did not work never enters the record, that knowledge is the individual's asset rather than the institution's; when the individual departs, the organisation loses not merely an employee but a portion of its own experimental history. This explains why staff turnover produces operational disruption in some organisations and strategic regression in others. The same mechanism accounts for why decision speed appears high in companies that have worked with a stable team for a long period: the speed derives from unrecorded memory rather than recorded knowledge, and unrecorded memory does not transfer. The mechanism that neutralises the tendency is not a mandate to report outcomes — compulsion of that kind drives experimentation outside the formal frame altogether — but the registration of an experiment's beginning. Three separable components carry it. The first is pre-registration: at the moment a trial is approved, a single paragraph records which measurement, falling below which threshold, will stop the trial, written before any result is in hand, so that an unfavourable outcome becomes a predefined output rather than a failure to be defended after the fact. The second is a termination protocol: every trial, favourable or not, closes with the same short note, and the length of that note does not vary with the direction of the finding. The third is accessibility: those notes are held not in the archive of the responsible individual but in a single record searchable by subject and by counterparty, since a record that cannot be retrieved carries the same economic value as one never written. In managing capital-intensive projects, BEIREK embeds these three components in the ordinary rhythm of project governance rather than treating them as a separate knowledge management initiative. For every technology, supplier and site alternative assessed during feasibility and development, the reason for elimination enters the decision log, and that log travels as an annex to the investment decision file, so that an alternative resurfacing at a later stage does not consume resources as though it were being examined for the first time. The same discipline runs through the contracting and procurement line for counterparty structures explored but never executed. In practice the decisive element is that the log opens at the point of proposal rather than the point of approval, because a log kept at approval captures only what was accepted, whereas a log kept at proposal captures everything that was considered — and the difference between the two is precisely the information the organisation loses. An organisation's capacity to learn is measured not by how much it attempts but by how much of what it attempted it can still recall, and this second quantity depends, independently of the first, entirely on how the record was designed. The question an investment committee ought to put to the historical record in front of it is not what proportion of the initiatives listed there succeeded, but how many initiatives never made it onto the list at all. ## Key Points - A favourable result converts on its own into a resource request or a claim on visibility, whereas an unfavourable result, once reported, generates an obligation to explain; this asymmetry shapes what enters the record independently of what actually happened. - When institutional memory consists only of completed successes, the base rate applied to new initiatives is structurally overestimated, and that distortion passes directly into budget calibration. - Absent a record of options tested and rejected, the same supplier trial or the same market test is financed a second time several years later, at no discount to the first. - In diligence, the layer a buyer can least easily reach is the list of things the target tried and abandoned; that gap is typically priced through earn-out thresholds, warranty scope or escrow ratio rather than through headline value. - The mechanism that neutralises the tendency is writing down the stopping threshold at the point the experiment is approved, not after the outcome is known. ## Questions ### How does the file-drawer problem actually show up inside a company? It surfaces when pilots, supplier trials and market tests that end without a result generate no closing record. The initiative is not formally cancelled; it simply fails to appear in the next reporting cycle. What the organisation is left holding is a list of what it accomplished rather than what it attempted, and that list is not a representative sample of past performance. ### Why is reporting an unsuccessful pilot so difficult? The cost of reporting is independent of the direction of the finding; what the reporting buys is not. A favourable result converts into a resource request and visibility, while an unfavourable result creates an obligation to explain the moment it is written down, and it lacks a natural owner. Under that asymmetry managers do not conceal outcomes; they decline to prioritise reporting them, which produces the same effect. ### How does the absence of records on rejected options affect valuation? A buyer cannot determine whether an abandoned product line stopped for technical reasons or because a manager changed roles. That uncertainty is rarely deducted from headline value; it is typically priced through tighter earn-out thresholds, wider representation and warranty coverage, or a higher escrow ratio. The record that was never kept returns as a cost embedded in the closing structure. ### Does logging every experiment simply create bureaucracy? The threshold of scale governs the answer. Requiring formal reports for the small variations teams try and abandon within a weekly rhythm pushes the cost of measurement above the informational value of the trial. The threshold at which a record becomes necessary is crossed when the trial's budget reaches a material size, when several units could independently repeat it, or when its outcome feeds an investment decision. --- Source: https://www.beirek.com/en/blog/file-drawer-problem Publisher: BEIREK LLC — https://www.beirek.com