What should an electronic lab notebook record for each experiment?

A good experiment record lets someone who wasn't there check the result a year later without asking anyone. Most notebooks, paper or electronic, fail that test in the same place. They record what the scientist concluded, and very little of how they got there.

So this is what an ELN should record for each experiment, in the order the work happens. I'll follow one routine test all the way through: the mycoplasma screen that cleared our example lab's new HEK293 cell bank in September. Each stage has its own guide, written for the people who do that part of the work, and they're linked as we go.

What should an experiment record include?

The method and its version, the materials and lots, the equipment, who did what and when, the data, every exception, the evidence and the sign-off. Leave any one of them out and there's a question about the result that nobody can answer later.

What	Why it matters	In the September screen
Method and version	Methods change, and results change with them	Workflow "Mycoplasma PCR screen" v3
Materials and lots	"Was it the reagent?" is the first question when something drifts	Three mix components, with lot numbers
Equipment	A faulty instrument has to be traced to every run it touched	The heat block and the qPCR instrument
People and times	Who to ask, and whether the waits were real	Six steps, each started and finished by name
Data	The result, in a shape that can be compared across runs	A four-column table, sealed at 13:40
Exceptions	Overrides, deviations and repeats, each with a reason	None overridden; one sample invalid
Evidence	The raw output behind the numbers	The instrument's amplification plot
Sign-off	Someone accepting the result	Checkpoint at 13:00, record at 16:40

The screen, from SOP to signature

Here's the chain, one stage at a time:

1. The SOP changes. Version 4 adds an internal control to every PCR well, so a failed reaction can no longer pass for a clean sample. How a lab should control its SOP versions.
2. The SOP becomes a workflow. Its rules turn into things the run enforces: a 72-hour wait, a new material, a checkpoint on the controls. Turning a written protocol into a workflow the bench can run.
3. The week gets planned. The Bench Board holds what the workflow doesn't: a reagent on order and a sample to fetch from the freezer. How a lab team plans its week at the bench.
4. The plate gets set up. Sixteen wells, 264 µL of mix. The calculations a cell culture bench does every day.
5. The run gets recorded. That's the rest of this piece.
6. The record gets signed. QA checks the controls and the one invalid sample, then signs at 16:40. What a batch record should contain before you sign it.

How should the run itself be recorded?

As it happens, by the system, not from memory afterwards. Times come from when steps actually started and finished, materials from the lots that were reserved, and data goes into a table whose columns were fixed before the run.

In CCA Labs the scientist works through the experiment in the runner. Each step shows its instructions, the linked SOP, the materials with their lots, the equipment, the expected outputs, and a place for evidence and notes. On Thursday the collect step became available at 09:45, exactly 72 hours after Monday's feed was marked done, because the workflow put a wait there.

The Ct values came from the instrument's export, pasted straight into the results table. The Call column worked itself out as they landed: Positive, Inconclusive for a late signal, Negative, or Invalid when neither the target nor the internal control came up. This lab's signature policy turns sealing on, so the scientist chose Sign & seal capture and re-entered their password. The values were frozen, hashed and signed as authored, and the runner shows it as Sealed.

What about things that go wrong mid-run?

Record them where they happened, with a reason, and keep every attempt. A record that only shows the attempt that worked makes the method look more reliable than it is.

CCA Labs gives each kind of trouble its own recorded action. Pause Experiment stops the labour clock and asks for a reason. Fail Step marks a step failed, and Revert Failed Step lets you go again. Override Mandatory Wait starts a step early, with a reason and a name on it. If a step's materials are flagged, you replace them, or you Start with Deviation and say why.

In fact, the failed attempt is often the most useful thing in the record. Each attempt at a step keeps its own timing, data and attachments, so the second try doesn't overwrite the first. Compare the paper version: "repeated, OK". That tells you it happened, and nothing about what went wrong or what changed.

Structured data or free text?

Both, for different things. Anything you'll want to compare across runs goes in a structured table. Anything you noticed goes in notes.

The mycoplasma screen is the clear case for structure. Twelve months of screens in the same four columns can be read as a trend: which cultures come up invalid, whether the internal control Cts are drifting. Twelve months of "all negative, one repeat" can't. That said, structure has a price. Someone has to design the table before the first run, and a locked table will annoy whoever wants a fifth column on the day. The runner keeps Notes for everything that doesn't fit, and it should.

Where do records usually fall short?

In the gaps between systems. The method lives in one place, the stock in another, the data in a third, and the record only links them if someone copies the numbers across.

CCA Labs closes most of those gaps, and I should say where it doesn't yet. It doesn't talk to instruments, so data is pasted from an export and the export is attached as evidence. A step links an SOP document but not a particular release of it, so the record names the workflow version exactly and the SOP version only by its release date. How the ELN relates to sample tracking is a separate question, covered in ELN vs LIMS for a cell culture lab.

Recording experiments in CCA Labs

An experiment is built from a released workflow version and keeps it. Scheduling reserves its materials against named lots and books the equipment its steps name. Replicates are separate linked experiments, each with its own data and cost. When it finishes, the materials and labour it used add up to a cost on the experiment, its project and its program, and the batch record sets out the whole run for signing. The rest is on the all-in-one lab page, and you can create your lab to record your own.

The test stays the same whatever software you use. Pick last month's version of your most routine test, give its record to someone who wasn't there, and see how many questions they have to ask. Each question is something the record should have held. The story of this one starts with the SOP: how a lab should control its SOP versions.