Ask a private equity firm where its knowledge lives and you will get a sensible answer.
The deal files are in the data room. The pipeline is in the CRM. IC materials sit in a shared drive. Portfolio reporting has its own system. The partners, of course, remember the rest.
Until they do not.
That arrangement works until someone asks a question that crosses those boundaries.
Why did we pass on the company two years ago? Which customer risk changed between the first and final memo? Have we seen this pricing model before? What did we believe at entry, and when did that belief change?
The documents may exist. The answer often does not.
That is not a knowledge system. It is a scavenger hunt with expensive participants.
The distinction has a long research history. In their foundational paper on organizational memory, James Walsh and Gerardo Rivera Ungson describe memory as a system of acquisition, retention, and retrieval. Keeping information is only one part of the job. The organization must also be able to recover it when a decision calls for it.
A repository preserves files. A firm memory preserves what happened, why it happened, and what came next.
Search is only the first mile
Most knowledge projects begin with search because search is easy to demonstrate. Type a phrase, find a file, open it.
That is valuable when the user knows what to look for. Investment questions are rarely that tidy.
A question such as “Have we seen this churn pattern before?” may require the system to connect a board pack, a commercial diligence report, an IC discussion, and the eventual outcome of a different portfolio company. The same company may appear under several names. The definition of churn may have changed. The relevant insight may sit in a note, not the final memo.
This is not a keyword problem. It is a context problem.
What gets lost between the files
The most valuable knowledge in a deal is often not the polished conclusion. It is the path taken to reach it.
The original belief
What did the team think was true at the start? Which assumptions were central to the case? Which facts were still uncertain?
The disagreement
Who challenged the thesis? What evidence changed the room? Which risk was accepted rather than resolved?
The change
What moved between the first look, the final memo, and the first year of ownership? Was it new information, a new interpretation, or a change in the business?
The outcome
Which assumptions held? Which ones failed? Did the team spot the signal early, late, or not at all?
These details are difficult to recover from a folder hierarchy because they live across time and across systems.
The firm does not need more storage. It needs continuity
.
Why good firms still forget
This is not usually a discipline problem. Deal teams document a great deal. The trouble is that each system captures a different slice of reality.
The CRM records activity. The data room holds evidence. The IC memo records the approved case. Board materials show the operating story after close. Email and meeting notes carry the debate between those formal records.
No single source is wrong. None is complete.
BCG’s private equity knowledge-management research makes a similar point in industry terms. It describes an effective knowledge capability as an ecosystem of people and processes, content, and technology, with deal memory mapped across individuals, sectors, and positions.
People bridge the gaps. A partner remembers that the team disliked the quality of revenue, even though the final pass note says valuation. An associate remembers that a customer concern was resolved in a call, but the source never made it back into the memo.
That informal layer is powerful and fragile. It weakens when people change roles, leave the firm, or simply move on to the next live process.
What institutional memory should do
A useful memory layer should help the firm reconstruct a decision without pretending the record is cleaner than it was.
It needs to preserve:
the entity being discussed, even when names change;
the point in time at which a fact was believed;
the source behind the fact;
the difference between evidence, interpretation, and decision;
the people and permissions attached to the underlying material;
later outcomes that confirm or challenge the original view.
This makes a different class of question possible.
Instead of “find the memo,” a user can ask:
What did we believe about customer concentration at entry?
Which evidence caused that view to change?
Where else has the same risk appeared?
What happened in those cases?
The answer should not be a confident paragraph floating above the source material. It should show the path back to the record.
That requirement is not specific to private equity. The NIST Generative AI Profile recommends documenting data sources, preserving provenance, and testing the feedback loop between generated content and human reviewers.
Build memory through the work itself
The worst knowledge systems ask busy investors to become librarians.
Nobody joined private equity to perfect a tagging taxonomy at 11:40 p.m.
They rely on perfect tagging, separate upload routines, and a burst of enthusiasm that fades after launch. Six months later, the knowledge base is already stale.
A better system learns from normal work. A decision in the pipeline, a revised IC memo, a partner note, or a portfolio update should enrich the record without creating a second job for the deal team.
That does not mean capturing everything. It means preserving the moments that matter:
A claim is introduced.
Evidence supports or challenges it.
The team makes a decision.
New information changes the view.
The outcome becomes known.
That sequence is the real asset.
A practical first test
Choose one sector or strategy with a known history. Pick 20 questions that experienced investors regularly ask and that currently require someone to search, call a colleague, or rely on memory.
Test the system on questions such as:
Why did we pass on similar businesses?
Which risks repeated across deals?
What changed after the first IC discussion?
Which entry assumptions were most often wrong?
Then judge the answers on more than relevance.
Were the entities correct? Was the timeline accurate? Could the user see the source? Did permissions hold? Did the answer admit when the record was incomplete?
If those basics fail, a polished interface will not rescue the project. Good gradients have never fixed bad memory.
Memory is an operating advantage
Private equity firms spend years accumulating judgment, but much of that judgment remains attached to individuals and isolated files.
Turning it into institutional memory does not make every decision right. It makes past experience available when the next decision is being formed.
Recallr connects the documents, decisions, entities, and changes that make up a firm’s investment history. The aim is simple: help the team remember not only what it knew, but when it knew it and why it mattered.
That is a better starting point than another search box.