Turn handwritten lab notes into clean spreadsheets, intelligently.
Snap photos of your handwritten data sheets — no matter how messy — and get a clean spreadsheet back in seconds. Add a few old spreadsheets you've already transcribed, and we'll match your lab's exact format and shorthand from day one.
Start with a free 14-day trial — see your own pages come back clean before you commit to anything.
We don't just scan your notebook.
Any tool can turn a legible grid into text. The chore that eats your Friday afternoon is everything around it — a continuation page with no header telling you whose data it is, a specimen that got reassigned to a new group in week three, one tech writing “NS” and another writing “0” for the same thing. That's the last mile nobody else automates.
Continuation pages, no labels
NoteMerge carries context across pages, so an unlabeled sheet still lands under the right subject and cohort.
Relocations, re-treatments, restarts
A subject moves to a new group, gets re-treated after a prior response, or restarts on a new cycle — NoteMerge keeps it one continuous record, not a duplicate.
Conflicting shorthand
A documented priority rule resolves the same measurement written two different ways — and logs which one won, and why.
Sharper with every upload
Every correction you make becomes training signal for your lab specifically — accuracy climbs with each notebook instead of resetting every time.
1 → 3
Every odd table pattern scientists actually use.
Lab notebooks are full of shorthand that trips up plain OCR — a ditto mark, a bracket spanning five rows, a value squeezed in above the line. NoteMerge reads each one the way you would, and shows its work.
Hover any example — or tap on mobile — to see exactly how we read it.
Ditto & “same as above”
A ditto mark, a repeat arrow, or a line down a column all mean “same as the cell above.” We fill the value down the column instead of leaving the cells empty.
| 22.4 |
| 22.4 |
| 22.4 |
Labels that span a block
A group or treatment name written once — or tied to its rows with a brace — is applied to every row it covers, not just the top row.
| ID | Group |
| S1 | Group 2 |
| S2 | Group 2 |
| ⋯ | Group 2 |
Stacked day headers
When one day splits into length and width sub-columns, we read the sub-headers and put each number in the right column — no guessing which is which.
| L | W |
| 4 | 3 |
Ranges, approximations & ceilings
A range, an approximation, or a “greater-than” ceiling keeps its meaning — a >2000 never quietly becomes a plain 2000.
| >2000 |
Carets & squeezed-in edits
A digit inserted above the line with a caret, or squeezed between two rows with an arrow, is read at the spot it points to — with the original kept on file.
| 23.1 |
Checkmarks, circles & ink color
A checkmark, a circled value, or a cell in a different pen carries intent — we save the mark as a note without letting it change the number.
| 2.4 |
ND, BDL, N/A & friends
“Not detected” is not zero. Tokens like ND, BDL, TNTC, and QNS are kept exactly as written, never turned into a misleading 0.
| result | ND |
Early exits & exclusions
A struck-through row or an early-exit note ends that record cleanly — and an excluded sample is dropped from the analysis, with the reason logged.
| ID | Status |
| S3 | excluded |
Plate & layout maps
A 96-well plate map or a housing/plot layout grid is a different shape entirely — captured as its own map instead of being flattened into your data table.
| A | A | B |
| A | B | B |
| · | · | · |
Three passes, one clean file.
Reading handwriting is a commodity. Resolving what it means across a whole notebook is the hard part.
Read carefully, not guessed
Every page gets read closely — dates, IDs, measurements, shorthand. If handwriting is unclear or a label is genuinely missing, it's flagged for you instead of silently filled in with a best guess.
Checked against the whole notebook
Your notes get one more pass to catch what a single page can't show on its own — a subject reassigned to a new group, an unlabeled continuation sheet, two techs writing the same result differently. Every judgment call is logged with its reasoning and a confidence score, so nothing is a black box.
A spreadsheet you can trust
You get back a correctly structured file, plus the paper trail behind every value in it. Run it again on the same notes and you get the same result — no surprises, nothing silently changed.
Wherever the notebook still lives on paper.
NoteMerge is the ingestion layer, not another system of record — it feeds whatever you already use.
Academic & discovery labs
Grant-fundable at a flat monthly rate — no procurement process, no seat count to justify. Point it at your next monthly notebook and get your afternoon back.
Industry discovery teams
Give a study team priority support and a push into whatever system already holds your data — NoteMerge reads the notebook, you keep your platform.
CROs & regulated labs
An audit trail and versioned provenance from day one, built toward the validation package regulated teams need. Tell us your requirements on the waitlist.
Priced like a lab tool, not enterprise software.
Free 14-day trial, no card required — see it work before you pay for it.
Academic
$79 /month per lab
Generous page allowance, overage billed per page. One clean line item for a grant budget.
- Reconciled Excel export
- Cell-level decisions log
- Learns your lab's shorthand over time
Industry team
from $300 /month per study team
Everything in Academic, plus priority support, more seats, and a push into your existing ELN or platform.
- Everything in Academic
- Priority support
- Push to your ELN or platform
Before you join
97.6% cell-level agreement (1,563 of 1,601) against hand-built targets across three independent multi-page studies. Every cell below our confidence threshold is flagged for your review before export — nothing ships as unreviewed truth.
Yes. A subject that's relocated, re-treated after a prior response, or restarted on a new cycle stays one continuous record instead of splitting into a duplicate. That cross-page continuity is exactly what the reconciliation engine is built for, not an edge case it happens to survive.
You see it. Low-confidence cells are flagged inline next to the original page image, with the decisions log explaining what NoteMerge chose and why. You confirm or fix it before the file is final.
No. NoteMerge is the ingestion layer that feeds whatever you already use — export to Excel or push into your existing system. It's not another system of record to maintain.
No. Your notebooks and corrections only tune your lab's own pipeline. Nothing tenant-specific ever trains another customer's model.
Photos of your notebook pages. If you have a spreadsheet those notes previously became, upload that too — it's the fastest way to calibrate to your exact format.
Not yet — that's a deliberate later phase. Every run already carries a versioned, auditable decisions log, which is the foundation regulated teams will need. Tell us your timeline when you join the waitlist.
Get early access
We're onboarding a small number of labs before opening up more broadly. Join the waitlist and we'll reach out to set up your free 14-day trial.