Binaly
How it works Pricing Docs Sign in Start free
Binaly
Start free

What your lab never wrote down

The most useful output of organising a research group’s knowledge is usually not what it finds. It is the list of what isn’t there.

Absence isn’t the same as evidence of absence

In a research record, a missing value and a value of zero look identical from a distance. If the substrate temperature field is empty for a deposition run, that could mean the run was performed at ambient, or that nobody recorded it, or that it was recorded in a lab book that has since been lost.

A system that answers questions smoothly over a corpus with unknown coverage isn’t merely unhelpful. It is actively hazardous, because it produces confident output with no signal about what it silently stepped around. This is the failure mode most likely to put a wrong number in a manuscript.

So every object in a Binaly graph reports its own gaps. Missing required fields, unresolved links, measurements with no recorded conditions, protocols referenced but never captured. The gaps are first-class data, queryable like anything else. You can ask what the group has not recorded, and get a list.

What groups typically find

The first gaps report is unflattering. That is the point of it. Recurring patterns across the corpora we have looked at:

Conditions recorded for some runs and not others. Usually the boundary falls on a person, not a date – one student recorded chamber pressure religiously and their successor did not. This is invisible until you count.

Protocols that exist only as a paper’s methods section. The operational version – the one with the step that isn’t in the paper because it seemed obvious – was never written anywhere. When the person who knew it leaves, the published methods section becomes the group’s own reference, and it isn’t sufficient to reproduce the work.

Negative results, nowhere at all. The three months spent on an approach that didn’t work is the single most consistently unrecorded category in every group. It is also, measured in saved student-years, probably the most valuable.

Calibration and instrument state at time of measurement. Recorded in the instrument’s own logs, on a machine that has since been reimaged, and never joined to the measurement it applies to.

The reason something was stopped. Not the fact of stopping – that is visible in the file dates – but the reason. Almost never written down, and almost always the first thing a new student needs.

The uncomfortable artefact

Handing a PI a map of what their group doesn’t know is a strange thing to sell. Most vendors would not.

But the alternative is worse. A group’s belief about its own record is systematically more optimistic than the record, because the belief is formed by the people who were present and the record is read by people who were not. The gap between those two is where reproducibility problems live, and it only ever widens.

The first honest inventory is the only way to start closing it.

Using it forward

Once gaps are explicit, three things become possible that were not:

A concrete handover task list. Instead of “please document your work,” a departing student gets a specific list of eleven runs missing conditions and four protocols never captured. Specific requests get done; general ones do not.

Reviewer-proofing before submission. Run the query before the manuscript goes out. If the analysis depends on measurements whose conditions were never recorded, you would rather find that yourself than have Reviewer 2 find it.

An honest scope for new students. “Here is what we know, here is what we think we know, and here is the boundary” is a far better first week than a shared drive and good luck.

The boundary of the answerable

When Binaly says your graph doesn’t contain something, that is a finding, not a failure.

It is also the rule we hold the system to: it will tell you a thing was never recorded rather than construct a plausible answer. For a tool whose output might end up in a paper, refusing to guess isn’t a limitation. It is the entire basis on which it can be used at all.

Find out what your lab already knows.

Thirty days free. Start with one project’s folder and see what it connects that you’d never have found yourself.

Start free

Thanks – that’s on its way. We reply within a day or two.

Not ready to sign up for anything? Ask us anything about how Binaly copes with your lab’s own files – a person answers, no discovery call.

A person answers. No discovery call.

Background
Overview Why not an ELN When a student leaves Gaps Starting with AI Security & data

What your lab never wrote down

The most useful output of organising a research group’s knowledge is usually not what it finds. It is the list of what isn’t there.

Absence isn’t the same as evidence of absence

A missing value and a value of zero look identical from a distance. If a deposition run’s substrate temperature field is empty, that could mean ambient, or that nobody recorded it, or that it was lost.

A system that answers smoothly over unknown coverage isn’t merely unhelpful – it’s actively hazardous, producing confident output with no signal about what it silently stepped around.

So every object in a Binaly graph reports its own gaps: missing fields, unresolved links, measurements with no recorded conditions. Gaps are first-class, queryable data – you can ask what the group has not recorded, and get a list.

What groups typically find

The first gaps report is unflattering. That’s the point. Recurring patterns:

Conditions recorded for some runs and not others – usually the boundary falls on a person, not a date.

Protocols that exist only as a paper’s methods section – the operational version, with the step that seemed too obvious to write down, never survives.

Negative results, nowhere at all – the single most consistently unrecorded category, and probably the most valuable.

Calibration and instrument state – stuck in the instrument’s own logs, never joined to the measurement it applies to.

The reason something was stopped – not the fact of stopping, but why. Almost never written down, and almost always the first thing a new student needs.

The uncomfortable artefact

Handing a PI a map of what their group doesn’t know is a strange thing to sell. But a group’s belief about its own record is systematically more optimistic than the record – the belief is formed by people who were present, the record is read by people who weren’t. That gap is where reproducibility problems live.

Using it forward

A concrete handover task list. A specific list of eleven runs missing conditions gets done; “please document your work” does not.

Reviewer-proofing before submission. Run the query before the manuscript goes out.

An honest scope for new students. “Here’s what we know, here’s what we think we know, and here’s the boundary” beats a shared drive and good luck.

The boundary of the answerable

When Binaly says your graph doesn’t contain something, that’s a finding, not a failure. It will tell you a thing was never recorded rather than construct a plausible answer – for a tool whose output might end up in a paper, refusing to guess is the entire basis on which it can be used at all.

Find out what your lab already knows.

Thirty days free. Start with one project’s folder and see what it connects that you’d never have found yourself.

Start free

Thanks – that’s on its way. We reply within a day or two.

Not ready to sign up for anything? Ask us anything about how Binaly copes with your lab’s own files – a person answers, no discovery call.

A person answers. No discovery call.