Fertility Clinic Software Problems: The 7 That Cost Real Money, and How to Avoid Them

Why does replacing the stim tracker become the biggest scope failure?

The project usually starts as a request to retire a shared workbook called something like Stim Tracker v4 FINAL. That framing hides the real requirement, because the workbook is not a table. It is a decision log for a process running under a hard clock. Day six oestradiol resolves at 10:15, the scan shows nine follicles over 12mm on one side and six on the other, and somebody has to decide by 11:30 whether to coast, drop the dose or trigger tonight, then reach fifty five patients before the afternoon.

What makes this specific to in vitro fertilisation (IVF) is that your protocol is genuinely yours. Your group's coasting threshold, your antagonist start rule, the different call list for a poor responder on a mini stimulation, and the fact that the practice you acquired last year does it differently. That logic lives in a senior nurse's head, and when she is away the day gets slower and the calls get inconsistent.

The fix is to make the cycle a first class object with a state machine rather than a folder of documents. One screen per monitoring day, measurements entered once by ovary and size bucket, laboratory values arriving over an interface the moment they resolve, protocol rules evaluated live, and a ranked call queue showing who needs contact and what changed. Every dose change writes an immutable audit row carrying the values that were on screen when the decision was made.

What goes wrong when you migrate fifteen years of embryo and cryo records?

This is reliably the hardest line item in the build, and it is hard for a reason teams only discover by starting. Legacy embryo and cryogenic storage data has gaps that were invisible while the data sat still: thaw events never recorded, cryo positions corrected on a laminated tank map in pencil rather than in the system, grading captured as free text in three different vocabularies depending on who was on the bench that year, and disposition consents filed as scanned documents with no link to the specimens they govern.

Migrating those records into a structured ledger surfaces every one of those gaps at once, which feels like the new system creating problems. It is not. It is the first honest inventory the practice has had.

The fix is to plan a reconciliation phase where the laboratory physically verifies tank contents against the migrated ledger before cutover, canister by canister. It is slow, it needs embryologist time rather than developer time, and it is the only way to start from a ledger you can defend. Where the source is genuinely ambiguous, record the ambiguity as a flag on the record rather than picking a value, because an inherited guess presented as fact is worse than a known gap.

Why do laboratory, records, witnessing and genetics integrations break after launch?

Each breaks differently and each has a different owner. Reference laboratory and electronic medical record interfaces over HL7 or FHIR are well understood engineering, and they break on identity: a patient with a name change, a duplicate medical record number created at a second site, or a result posted against an order the cycle no longer expects. Build an exception queue that a coordinator sees the same morning, because a laboratory value silently unmatched on a monitoring day is a clinical problem within hours, not a data problem within weeks.

Witnessing systems such as RI Witness and Matcher are vendor dependent, and the negotiation to obtain an interface sometimes takes longer than the engineering. Scope that conversation before you commit to a date.

Genetics is where post launch breakage is most predictable, because Cooper Genomics, Natera and Igenomix do not agree on a report format and any of them may change theirs without telling you. Ingestion has to be built per laboratory, and extraction should propose a match on tube identifier and accession for an embryologist to confirm rather than committing it automatically. The confirmation step is not friction. It is the control that stops a format change from becoming a transfer error.

What happens when the cryo ledger, donor eligibility and outcome capture are not covered?

Most clinics cannot quickly answer three questions: how many straws sit in tank four, which belong to patients who stopped paying storage fees years ago, and which have a signed disposition consent on file. Those three facts have never been in the same place, because the tank map is a printed grid, storage billing runs on an anniversary date set by hand, and consents live in a document tool. That gap is where the litigation risk sits and where annual storage revenue quietly leaks off the invoice run.

Donor programmes fail in a similar shape. Donor eligibility under 21 CFR Part 1271 requires infectious disease testing inside a defined window, cytomegalovirus status has to match or the recipient consents in writing, carrier screening must be cross referenced so a donor and a recipient partner carrying the same condition are never matched, and family limits have to decrement on reported live birth. When those live in a coordinator's spreadsheet, the constraint engine is a person having a good day.

Outcome reporting is the third gap. Society for Assisted Reproductive Technology reporting is an annual reckoning with data that should have been captured cleanly in real time, and no export fixes a field that was never required at the point of care. Enforce the reporting data model at capture and close the outcome loop with structured pregnancy follow up, so the submission becomes a file you generate rather than a season you survive.

Should you build custom or configure what you already own?

Keep the off the shelf product if you are a single site doing under roughly 300 retrievals a year, your donor work is referral only, you bill insurance rather than running refund programmes, and your laboratory director is genuinely comfortable with the workbook. eIVF and IDEAS are real products maintained by people who know assisted reproduction, and rebuilding demographics and scheduling simply to own them is a poor trade at that scale. EngagedMD remains sensible for consents whatever you run clinically.

The middle path is usually the right one for a growing group. Build the cycle engine and the monitoring day workflow and run it alongside eIVF or IDEAS rather than replacing them, so coordinators get relief in the first release and you avoid a rip and replace while patients are starting stimulation next Monday.

Build more broadly when three signals appear together. You are multi site or acquiring and the protocols differ enough that no configuration screen reconciles them. Your highest value operations, meaning the donor programme, the refund programme and the cryo ledger, already live outside the vendor product in spreadsheets, which means you are already maintaining custom software with no audit trail. And you asked your vendor for a change that affects margin and received a roadmap answer measured in years.

How do hidden costs get into the quote?

Through the word integration, which in this category is never one line. From Digital Heroes delivery experience a focused first release covering the cycle engine and monitoring day workflow runs $60,000 to $130,000 and ships in 12 to 16 weeks. A full platform adding embryology, the cryo ledger, donor matching, billing and outcome reporting runs $150,000 to $400,000 phased over 6 to 12 months.

The lines that get missed are each reference laboratory and records interface priced separately, the witnessing vendor conversation, per laboratory genetics ingestion because three vendors mean three formats, and the migration and physical reconciliation of legacy embryo and cryo records.

The other commonly omitted line is regulatory rigor as a deliverable. Audit trails, role based access, electronic signature handling and validation documentation carry hours, and a developer who has never sat through a College of American Pathologists inspection will not budget for them. Ask precisely what artefacts they will hand your inspector, and ask it before you agree a price rather than after.

What separates a build that works from one that fails here?

Ask a prospective developer to model an embryo on a whiteboard before discussing price. If they draw a table with a grade column, they have not built this. The right answer involves lineage from the oocyte and the sperm source, an event stream of grading, biopsy, freeze and thaw, and a cryo position expressed as a transaction history rather than a field that gets overwritten. Domain modelling decides this category and it shows within ten minutes.

Make them name integrations they have actually shipped, not HL7 in the abstract. Which reference laboratories, which records system, whether they have persuaded a witnessing vendor to return their calls, and how they handled a genetics laboratory that only sends portable document format reports. A team that has done it will have specific opinions about specific vendors being difficult, because they are.

Then ask what they would refuse to build first. Anyone who says yes to the whole platform in release one is selling a twelve month gap between payment and value, in a clinic with patients starting stimulation next week. Sequence the monitoring day first because that is where coordinators lose the most hours, take the cryo ledger and donor matching second because that is where the liability sits, and settle code ownership, repository access and the deployment path in the same conversation as the price.