In the Metamorphoses, the Roman poet Ovid tells of Echo, a nymph punished by Juno, queen of the Roman gods, for talking too much. Echo kept her voice but lost the power to speak first, and she could only return the last words someone else had already said.
Echo is a proxy for a specific kind of failure: a voice that sounds like agreement and carries no new information. Everything she said back, the speaker had said first. Listening to her taught you nothing you did not already believe.
In life science tools, the most convincing echo is your own institutional knowledge. A Product Manager (PM) with ten years in western blotting knows that sensitivity drives imager upgrades. So the PM asks a customer, "Is sensitivity your biggest frustration with your current imager?" The scientist replies, "Yes, sensitivity is a big frustration." The PM writes validated. The customer never began a sentence, and the PM heard their own knowledge come back and called it evidence.
Institutional knowledge is one of the most valuable assets a life science company has. It is also the asset most likely to convince you that you are more prepared than you are.
Institutional knowledge is your moat
Institutional or tribal knowledge, the experiences and insights that reside within R&D, Applications, Product Management and really any employee, is mostly tacit knowledge. A recent California Management Review (CMR) article defines tacit knowledge as "the reasoning patterns, informal heuristics, situational awareness and nuanced interpretive skills that experts develop over years of experience." The authors argue it is a bigger differentiator than data or models, because it "determines how teams make sense of ambiguity and how they distinguish meaningful signals from noise."
In practice, it is the veteran PM who hears "my bands drift between runs" and recognizes a normalization problem worth a product line. The applications scientist knows which buffer substitution customers will try before they call technical support, and the R&D lead remembers why a chromatography column redesign failed six years ago. None of that lives in a dashboard or a queryable database.
I saw the value of capturing this knowledge early in my career. At Pharmacia Biotech and Amersham Life Sciences, our technical support knowledge bases turned what PMs and R&D knew about each product into searchable FAQs. Once that knowledge was written down, anyone could find it. It is what the CMR authors mean when they describe companies that "embed human expertise into the connective tissue of their platforms."
The urgency is growing because, as the CMR authors point out, experienced employees are retiring in large numbers and taking their tacit knowledge with them.
Your most valuable knowledge is the most likely to fool you
In eye-tracking studies by Bilalić and colleagues, expert chess players who recognized a familiar solution kept looking at it and missed a better one, even while they believed they were searching for alternatives. Psychologists call this the Einstellung effect, and in voice of the customer (VOC) work it produces interview questions that sound like discovery and function as confirmation.
Some tacit knowledge holds for decades. Some expires with the next journal policy change. Samuel Arbesman calls this the half-life of facts, and the trouble is that an undated belief gives no clue which kind it is. Take a 2016 win/loss finding: sensitivity drives imager upgrades. Since then, Nature has started asking authors to "provide original uncropped data for all of the gels used in your final figures" (Nature figure guide). A western blot buyer today may care more about quantitation that survives peer review than about another 2x in sensitivity.
The risk is highest in RUO tools, where PMs stay in one category for a decade, KOL relationships go back years, and "we tried that in 2014" still ends debates at Gate reviews. Undated knowledge walks into the Feasibility gate review looking like evidence, and the requirement it supports can commit 18 to 24 months of development.
Keep the knowledge, lose the echo
Your veterans stay at the center of this. Manage their knowledge like any other asset: know what you have and when it was last checked.
1. Write it down and date it
Before each round of customer conversations, the team builds a dated knowledge log: what each person believes is true about the customer’s application or problems to be solved, with the year and the source. For example: "Sensitivity drives imager upgrades (2016 win/loss study)." The log lives with the product line, and the core team owns it. The date shows which knowledge has been retested recently and which has not been checked since it was learned. This is the knowledge base lesson applied to beliefs: once knowledge is written down and dated, it can be checked. Then attach the log to the business case so the dates travel with it into every Gate review. I call this a Decision Log: a record earns its keep only when someone reads it at the next milestone.
2. Retest the oldest entries first
Every round of customer contact must retest at least one entry, starting with the oldest. To retest an entry, ask about the last time something happened rather than stating your belief: "Walk me through the last blot you had to rerun," not "Would better sensitivity help?" With OEM customers in IVD, ask a diagnostic manufacturer's R&D lead to walk you through the last antibody lot they rejected, not whether lot-to-lot consistency matters.
Then check the notes for one thing: who said the key word first? Count the concept rather than the exact term: "faint bands" and "sensitivity" belong to the same entry. If the PM introduced it, the customer's agreement is an echo of your institutional knowledge. Divide the PM-first mentions by the total mentions to get what I call the echo ratio. It is my own shorthand for measuring acquiescence bias, the tendency of respondents to agree with whatever is put in front of them. Here is an example across 14 interviews for a next-generation chemiluminescence imager.
| Example findings | Customers who said it | PM said it first | Customer said it first | Echo ratio |
|---|---|---|---|---|
| Sensitivity is the top unmet need | 9 of 14 | 7 | 2 | 78% |
| Quantitation must hold up to journal image review | 6 of 14 | 0 | 6 | 0% |
| Imaging time slows the lab down | 4 of 14 | 3 | 1 | 75% |
Sensitivity looks like the winner, with 9 mentions to quantitation's 6. But 7 of those 9 customers only agreed after the PM raised it, while all 6 quantitation mentions were unprompted. The 2016 entry did not survive the retest. A team that funds it anyway commits roughly 18 months of engineering to a need only 2 customers raised on their own. Treat any entry with an echo ratio above 50% as unconfirmed, and adjust that starting rule of thumb as your team learns how echo ratios track with launch results.
A low unprompted count has a second explanation, though: customers rarely mention what they take for granted. In the Kano model, must-have features are the ones customers "do not always mention" because they assume them. So a high echo ratio tells you an entry is unconfirmed, and it may still matter. Retest it by asking what would make the customer switch away from their current imager. With samples this small, treat the ratio as a flag rather than a statistic.
If a round leaves every entry untested, the team learned nothing new. Over time, the log becomes what Martin Reeves and colleagues call an evolvable script: a short, living document the team revises as outcomes come in.
Try this now: List the three beliefs your team cites most often at Gate reviews. Next to each, write when it was last tested with a customer. If any answer is "before our current product launched," you are less prepared than you think. For a sharper test, run the echo check on the interview notes behind your last launch.
Echo's lesson: date what you know
Every word Echo spoke had first been said by someone else. That is what makes undated institutional knowledge dangerous: it passes every check except the one that matters, which is whether a customer said it first, and said it recently. Institutional knowledge is your moat as long as it carries a date. Knowledge you built in 2016 and haven’t retested since is the trap.
Echo was cursed to finish other people's sentences. Your customers aren't cursed. They're just polite, and every belief you put in front of them gives them a reason to be.
FAQs
How do I get veterans to write down what they know without it feeling like an audit?
Veterans often hear "write down your assumptions" as "prove you are still useful." That is why it is called a knowledge log: the entries earned their place. Ask each veteran for the five beliefs they cite most often and the story behind each one. A veteran instrument PM who explains why liquid handler customers distrust tip-detection sensors gives you a dated, sourced entry in one conversation, plus the context that tells you how to retest it.
What happens to our institutional knowledge when the veteran retires?
If it is not written down, it leaves with them. The CMR authors recommend starting with a simpler step than documenting everything: map who holds critical know-how and where the organization is most exposed to losing it. Add an owner to each entry in the knowledge log. Entries held by one person who is close to retirement get retested and documented first. If only one field application scientist knows why a qPCR master mix underperforms on a particular cycler, that is your first entry.
Can grounded synthetic customers help, or are they just another echo?
They are an echo, and that is exactly why they are useful. A grounded synthetic customer panel is built from your team's VOC, win/loss reports, and deep research on the application or technique, so it plays that knowledge back to you on demand. Grounded means each answer the panel gives can be traced to a source in that evidence. Use it to pressure-test product concepts, messaging and sales plays. What a synthetic panel cannot do is tell you what your team does not know, so it will never produce a low-echo finding. Think of it as your knowledge log made queryable: useful for rehearsing, but only real customers can retest an entry.
