Drawn from the project's field studies across SaaS, finance, healthcare, manufacturing, and government. None of these is, on its own, conclusive. Three or more, occurring routinely, is.
A junior product manager points at a customer complaint and says "we should look at this." The room nods. Two months later someone unearths the 2022 spec that had already designed and de-prioritized the same fix. The most useful thing a new hire is doing, in an amnesiac organization, is being insufficiently trained to know what the institution has already given up on.
68%of new-hire-discovered "insights" in the project's field sample turned out to duplicate prior internal work the organization had already shelved.
Should the API be REST or GraphQL? Should we run incidents on Slack or PagerDuty? Should onboarding be self-serve or assisted? Healthy organizations make these decisions, write down the trade-off, and revisit them on a known schedule. Amnesiac organizations re-have the same arguments — with the same conclusions, by people who don't know they've already been had.
3.4×average number of times a "load-bearing" architectural decision was re-debated in 5-year-old-and-up software firms in the project's longitudinal sample.
The phrase is uttered with a specific resigned tone. The doc exists. The doc is wrong. Or the doc is right and unfindable. Or the doc is right and findable but contradicts a more recent, less authoritative doc. The Gartner finding that 47% of knowledge workers struggle to find information they need is not a search problem — it is a trust problem. People give up on the documents that exist before they finish failing to find them.
The Project Management Institute, summarizing decades of project research, identifies "giving lip service to lessons learned" as the single most common form of organizational amnesia in mature firms. The mechanism is mechanical: incidents produce postmortems; postmortems produce filings; filings produce no behavioral change. The next incident, when it arrives, repeats the previous one in form if not in detail.
~12%of post-incident reviews in the project's field sample were re-read by anyone within 12 months of being written.
Every organization has them — the four engineers, two operators, one CSM who collectively "know how things work." When they go on vacation, weeks slip. When they leave, quarters do. The organization is not running on its institutional memory; it is running on the working memory of a handful of overburdened individuals, who themselves are unable to take a real vacation. This is the tacit-concentration failure mode.
An engineer resigns. Their last two weeks are farewell lunches and an underweight runbook. Six weeks later, the team is reverse-engineering the failover script, the customer-tier exception logic, and the reason that one cron job runs at 04:17 specifically. The knowledge wasn't lost when they left; the knowledge wasn't there to begin with — it lived in their head and the institution had no instrument to extract it.
7.2×more codable tacit knowledge recovered when departure interviews use a structured walk-through-three-past-situations protocol vs. the conventional checklist (Project working paper WP-24-02).
The most reliable proxy for organizational amnesia is the answer to: "when you needed an internal answer last week, did you find a document or did you ask a person?" Healthy organizations land on a document. Amnesiac organizations land on a person — usually one of the load-bearing people from sign 05. The cost is double: the searcher's time plus the answerer's interruption.
102 minaverage time per knowledge worker per day spent searching for information at firms above 10,000 employees (Iterators, 2025).
The customer's contract terms are in Salesforce, in the contract PDF, in an email thread, and as told by the AE three quarters ago. The four versions disagree in subtle, material ways. Nobody has the authority to declare one canonical. The information is everywhere and so, effectively, nowhere — what the project's etiology calls dispersal.
A team is dissolved and recombined. The org chart updates by Friday. The institutional memory takes nine months. Customers report being re-introduced to the company by their new account team. Internal partners report being asked questions they had answered, in writing, to the old team. The reorg event lasts a week; the amnesia event it produces lasts the better part of a year.
~6 yrobserved periodicity of major institutional memory loss in 8–14-year-old software firms (vs. ~20 yr in century-old institutions per PMI).
Ask five long-tenured employees to write down the unwritten rules — "the things everyone here knows." A healthy organization produces five lists with substantial overlap. An amnesiac organization produces five different organizations. The differences in their lists are not folklore; they are evidence that the institution has lost the shared layer of tacit knowledge that lets a culture cohere.
The signs are easy to recognize. What is hard is admitting how long they have been recognizable.