Guide
When should a biotech startup move from spreadsheets to a true LIMS?
Spreadsheets work for a small lab until several people depend on the same samples. This guide covers the signs you've reached that point, what to set up first, how to move your sheets across, and what audit-ready means before any regulated work.
The evidence
What goes wrong when a lab runs on spreadsheets?
Spreadsheets fail quietly. They don't record which sample came from which, who changed a value, or which copy is current, and their errors look like real data.
Gene names are the best-known case. In 2020 the HUGO Gene Nomenclature Committee renamed symbols such as MARCH1 and SEPT1, now MARCHF1 and SEPTIN1, because spreadsheets kept turning them into dates.
| Finding | Figure |
|---|---|
| Formula cells with errors, in an audit of 50 working spreadsheets Powell et al., 2009 | 0.9% to 1.8% |
| Papers in leading genomics journals whose Excel gene lists had names converted to dates or numbers Ziemann et al., 2016 | About 1 in 5 |
| Drop in the odds that a paper's data still exists, for each year after publication (516 papers) Vines et al., 2014 | 17% a year |
Check your lab
What are the signs you've outgrown spreadsheets?
Count how many of these are true today. Our rule of thumb: at three or more, start looking at a LIMS. If a partner's request is one of them, start now.
- Two plates share a name.IDs are typed by hand, so "P3-1" means different things to different people.
- Nobody is sure which sheet is current.Copies live on laptops and in email, and they disagree.
- Finding a plate takes a message.Only the person who made it knows where it is.
- Lineage lives in people's heads.Tracing an aliquot to its parent means opening ELN entries one by one, or asking the right person.
- Plate data is pasted by hand.Reader exports go into a template well by well.
- Volumes are guesses.What's left in each sample is known only after someone checks the freezer.
- Sample history leaves when people do.When someone moves on, what they knew about their boxes and plates goes with them.
- A history request takes a day.A partner, CRO or diligence team asks where a sample has been, and you rebuild the answer from several files.
Weighing a homegrown tracker instead? Read build or buy your ELN and LIMS.
Day one
What should you set up first in a LIMS?
Set up the parts that are hard to change once data is in them: how samples are named, what type each record is, where things are stored, and which sample came from which.
| Set up | Why it comes first | A good starting rule |
|---|---|---|
| Sample IDs | Every label, link and result points at the ID. | One format per record type, assigned by the system, never reused. For example SPL-20261020-0001. |
| Record types and fields | Every sample needs a type, such as plasmid, cell line or protein. | Start simple: a type and the few fields you use today. Add a field the day an assay needs it. |
| Containers and storage | A sample without a place gets lost. | Plates, racks and boxes as containers with positions, inside nested locations: freezer, shelf, rack, box. |
| Lineage | It can't be rebuilt reliably after the fact. | Every aliquot, pool and transfer records its source and the volume taken. |
| Readings on samples | A result saved in a separate file loses track of which sample it came from. | Instrument exports attach to the sample in each well. |
| Links from ELN entries | The experiment explains why the sample exists. | Each ELN entry links the samples it used and made. |
| Permissions and change history | Partners will ask who changed what. | Record the old value, the new value, who changed it and when. Make deletes reversible. |
ELN vs LIMS
Do you need an ELN, a LIMS, or both?
Most startups need both. An ELN (electronic lab notebook) records what you did and why. A LIMS (laboratory information management system) records what physically exists and what happened to it. Keep them linked, so each ELN entry points at its samples and each result traces back to its source sample.
| ELN | LIMS | |
|---|---|---|
| Records | Experiments: plans, protocols, observations, conclusions | Samples, materials, containers and plates, and the work done on them |
| Answers | What did we do, and why? | What is in this plate, where did it come from, and how much is left? |
| History | Signed, versioned entries | Every change to a field, with who and when |
| Needed from | The first experiment | The first shared or split samples |
Moving over
How do you move existing spreadsheets into a LIMS?
Clean first, import second, and pick one cutover date. After it, new work goes only into the LIMS.
- Pick a cutover date.Keep the old sheets as read-only copies. Running both systems side by side brings back the problem of two records.
- Make one row per well or container.Split combined columns such as "Box 3 / A4" into container and position, keep one unit per column, and remove merged cells.
- Look for silent conversions.Search gene and sample-name columns for values that became dates (1-Mar, 2-Sep) or scientific notation, and for IDs that lost their leading zeros.
- Separate the material from the sample.A plasmid or cell line is a material. Each well or container of it is a sample, with its own location and volume.
- Import storage, then samples, then lineage.Where a parent is unknown, mark the sample as legacy rather than guessing.
- Spot-check against the freezer.Pull 20 random records and find them in the freezer. Check counts box by box before you retire the sheets.
Audit-ready
What does audit-ready mean for an early-stage lab?
No regulator audits a research-stage lab's sample records. Here, audit-ready means records that hold up when a partner, investor or CRO asks: you can answer three questions in minutes. Where did this sample come from, who changed this record and when, and which raw data produced this result? It isn't the same as GxP validation.
FDA's data integrity guidance sums up good records as ALCOA: attributable, legible, contemporaneously recorded, original or a true copy, and accurate. It's written for drug manufacturing, but it's a useful test for any lab record, and spreadsheets struggle with the first and last of those.
The people who ask are usually outside the lab. A diligence team wants to see how key results were produced. A pharma partner wants a material's history. A CRO needs to know which samples are coming, in which plate layout. Lineage, change history and readings attached to samples answer all three.
A validated GxP system is a separate step. When GLP or GMP studies begin, records that fall under FDA predicate rules need a system validated for that use, with Part 11-style controls such as electronic signatures, secure time-stamped audit trails and limited access. Plan it when that work is on the calendar, and don't let it hold up basic sample tracking now.
Cypher AI
How does Cypher AI handle the move from spreadsheets?
Cypher AI is an ELN and a LIMS in one platform, with an AI agent that does the import and asks before it writes.
Every change to a LIMS record is logged with the old value, the new value, who made it and when, and deleted samples can be restored. Cypher AI is SOC 2 Type II. The LIMS comes with the Pro and Enterprise plans, and teams of 20 or fewer scientists that have raised under $30M can join the Startup Program.
First steps
How do you get started?
- Apply to the Startup Program.Applying is a short call to check you qualify. Or start on the free plan today.
- Bring in one sheet.Start with the freezer box or plate set you use most. The agent turns it into LIMS records you approve.
- Link your next experiment.Write the next ELN entry with its samples, then sync them to the LIMS.
Setting up for a larger R&D team? Enterprise runs in your own cloud account, with a forward-deployed engineer.
FAQ
Questions about moving from spreadsheets to a LIMS.
How much does a LIMS cost for a small biotech?
Prices vary widely, from per-user plans to annual licenses with implementation fees. Cypher AI's LIMS comes with Pro at $100 per user a month. On the Startup Program it's free for 90 days, then $25 per user a month for year one. Our best LIMS page compares other vendors' published prices.
Can we keep using Excel alongside a LIMS?
Yes, for analysis. Export readings or sample lists from the LIMS and work on them in Excel. Don't keep a separate inventory sheet, though: the LIMS should be the one record of what exists and where.
How long does it take to move from spreadsheets to a LIMS?
Most of the time goes into cleaning the sheets, not the import. Start with the boxes and plates in active use, and bring archive boxes across later. With Cypher AI, the agent helps you set it up in an afternoon, with no implementation fee.
What's the difference between a LIMS and an inventory app?
An inventory app tracks what's in the freezer. A LIMS also records what was done to each sample: which parent it came from, how much was used, which plate it went into and what was measured. That history is what partners and auditors ask for.
Should a startup set up the ELN or the LIMS first?
Most start with an ELN, because experiments come before shared samples. Adding the LIMS later is easier when both are one system, since ELN entries can link samples from day one. In Cypher AI they share the same records, so there's no integration project when you add the LIMS.
Can ChatGPT or Claude clean up our spreadsheets before we move?
They can help find duplicates, split columns and spot gene names turned into dates. Writing the result into a LIMS is a separate step that needs a connection to the system and someone to approve the changes. Our guide to using ChatGPT or Claude with ELN and LIMS data covers the options.
How did we write this?
We make Cypher AI, so we have a view on when a lab needs a LIMS. Research figures come from the public sources listed here, checked on October 20, 2026. Cypher AI details come from cypherbio.ai and the product's documentation.
Spotted something out of date? Tell us at cypherbio.ai/contact and we'll fix it.
Sources
- Formula error rates: Powell, Baker and Lawson, "Errors in Operational Spreadsheets", Journal of Organizational and End User Computing 21(3), 2009, Dartmouth
- Gene name errors, about one in five papers: Ziemann, Eren and El-Osta, Genome Biology 17:177, 2016, doi.org/10.1186/s13059-016-1044-7
- Gene symbols renamed: Bruford et al., "Guidelines for human gene nomenclature", Nature Genetics 52, 2020, doi.org/10.1038/s41588-020-0669-3
- Data availability over time: Vines et al., "The availability of research data declines rapidly with article age", Current Biology 24, 2014, doi.org/10.1016/j.cub.2013.11.014
- ALCOA: FDA, Data Integrity and Compliance With Drug CGMP: Questions and Answers, December 2018
- Part 11 controls: 21 CFR 11.10; scope: FDA, Part 11 Scope and Application, 2003
- Cypher AI plans: cypherbio.ai/pricing and cypherbio.ai/startup · Security: Trust portal
Product names are trademarks of their owners.