Research
Unrecorded knowledge
The information a company runs on arrives in its messages, and much of it never reaches the task list, the CRM or the knowledge base.
Last updated August 15, 2026
Tasks, contacts and knowledge all go unrecorded, and each needs a different fix.
Someone asks for something in a thread, you intend to do it, and it never becomes a tracked task. That failure is fast and visible: a deadline passes, a client chases.
Someone leaves for a new company, gets promoted into a different role, or switches to a new email address. The contact record has to be updated each time new information arrives; otherwise, it becomes increasingly wrong. You find out when the mail bounces.
Knowledge is a superset of the other two. Someone learns that a customer is opening two hundred stores, or why a pricing exception was granted, and never saves it anywhere. Nothing goes wrong at the time. It surfaces later, when an account changes hands, when nobody can say why the exception was granted, or when two people answer the same question differently.
How often it happens
Three published measurements bear directly on how much goes unwritten. None of them covers tasks: there is no published metric for how many commitments go untracked.
Contacts
2.1% a month
The rate B2B contact data decays, compounding to roughly a fifth of a database going wrong every year without anyone making a mistake. One job change invalidates the email, the phone, the title and the company at once.
Contacts
76% say less than half their CRM data is accurate
Reported by the people who use the system daily. The same survey found workers spend on average 13 hours a week hunting for basic information.
Knowledge
42% is known to one person only
Institutional knowledge unique to the person holding it: not written down anywhere, and not known by a colleague either. Nearly half of what a company knows depends on one person still being there to ask.
What it costs
Most of the published research measures the bill rather than the gap that produces it.
Tasks
26% of deadlines missed
Asana attributes this to distributed work, not to untracked commitments, so it shows what missed deadlines cost without showing how many of them started as a request nobody wrote down.
Contacts
16 sales deals lost a quarter
On average, as a result of poor-quality data, with 37% of organizations reporting revenue lost directly to it.
Knowledge
25% of the week spent searching
Leaders and teams waste a quarter of their time just searching for answers. That is the tax on knowledge nobody wrote down, paid by everyone who needs it later.
Knowledge
60% cannot get information from colleagues
Found it difficult, very difficult or nearly impossible to obtain information vital to their job from the people who have it. The knowledge exists. Reaching it is the problem.
Why it happens
Two things drive all three: how many messages arrive, and how little attention is left to process them. The same skim that misses a request misses a new phone number and a customer detail.
All three
117 emails a day
Received by the average worker, most skimmed in under 60 seconds. Skimming is a decision to decide later.
All three
153 chat messages a day
On top of the email, per weekday. Any measure that treats email as the whole inbound picture is counting less than half of it.
All three
An interruption every 2 minutes
275 times a day during core hours. Nobody reads a stream continuously, so a detail arrives in the gap between two interruptions and gets skimmed instead of recorded.
All three
87% have no capacity to coordinate
Knowledge workers say that with everyone in execution mode, they lack the time or capacity to coordinate. Writing something down where others will find it is coordination, and it is the first thing dropped when there is none to spare.
How this is sourced
Asana publishes its 2023 figures only inside a gated report. Its public page carries labels for apps used per day and percent of the workday lost to repetitive tasks, with no values attached. Checked August 13, 2026.
A whole family of tacit-knowledge percentages circulates, and none of them is sourced to a study. The Delphi Group is cited at 80% in one place and 42% in another. Gartner is credited with 70 to 80%. A 2022 book puts it at 90%, and a peer-reviewed paper in the Business Process Management Journal, published August 2024 with 344 respondents, opens on the same 90% as an established premise rather than a finding. Each is stated confidently, each is repeated widely, and none of the sources quoting them names a study, a sample or a date.
The likeliest explanation is that the 90% was never a measurement. Michael Polanyi’s 1966 book gave the field its founding line, that we can know more than we can tell. He attached no number to it. Later theorists reached for an iceberg to illustrate his idea, and Nonaka and Takeuchi described the knowledge you can put in words and numbers as the tip of it. Roughly nine tenths of an iceberg sits below the waterline. A drawing became a percentage.
The Panopto figures are from 2018. They predate remote work and every AI tool on the market, and they remain the only published measurement of how much institutional knowledge has a single copy.
We left out several widely quoted contact-decay figures. The 5-10% invalid-email rate, the 5-7% bounce rate and the 20-30% of a rep’s time all trace back to vendor blog posts with no study behind them. The 2.1% monthly rate is included but flagged, because the original MarketingSherpa study is not reachable and HubSpot’s reporting is the only version available.
What nobody has measured
Panopto took a snapshot and found 42% of what a company knows living in one person’s head. Nobody has measured the flow into that state: of all the requests, changes and decisions arriving in a company’s messages, how much never gets recorded anywhere.
That is really three questions. How many of the commitments people make in messages never become tracked work. How many changes of job, title or address never reach the address book. How many decisions and operating details never reach a document.
The 42% answers none of them. Something can be written in a Slack thread, read by four colleagues, and still never reach a task list, an address book or a document. It was written. It was seen. It is still lost.
We read the message stream and build all three records out of it, so we can count what we find: tasks identified per day, contact updates applied, knowledge updates proposed. What no software can count is what would have been recorded without it, and that is what these three questions ask for. Getting at that means asking people, and that is our next step.
Nothing slips. Everyone on the same page.
this+that reads the message stream, turns the commitments inside it into assigned tasks, and writes what the work teaches into a knowledge layer the whole team shares. Free during beta.
Free during beta · No credit card required