Most failed EDC implementations are not caused by software. They come from a build started on a draft protocol, a paper CRF copied screen for screen, edit checks nobody tested and sites trained the day before enrolment. Here are twelve mistakes, what they look like and the fix.
Free sandbox · No credit card · 21 CFR Part 11 aligned
Study
DEMO-IMPL-04
Forms
14 drafted
Findings
5 open
The pattern behind the mistakes
Why this matters
An EDC system does not make data clean by itself. It makes data clean when the forms ask the right questions in the right order, when the rules catch errors at entry, when the right people have the right access and when the data can leave the system in a shape the analysis needs. Each of those is a decision made during implementation, usually by a handful of people under time pressure and often before the protocol is final.
The cost of a mistake rises sharply with time. A wrong unit on a form caught in design review is a two-minute edit. The same error caught after 40 participants requires a version change, a decision about already-entered data, queries to sites, retraining and an audit trail explaining all of it. Our guide to ecrf versioning for mid-study changes shows what that looks like in practice, and protocol amendments without change orders covers why who owns the build matters.
None of the mistakes below is exotic. Each one appears repeatedly in small sponsor and academic studies, and each can be prevented with an hour of planning. For the schedule view of the same project, see the EDC implementation timeline.
Design mistakes
These are made before a single form is built, or while building it.
The protocol is still changing, the visit schedule is vague and the build starts anyway. Sign: forms are renamed weekly. Fix: freeze a build baseline (schedule of assessments and primary endpoints), and treat later changes as controlled amendments. A schedule of assessments builder helps you agree the grid first.
Free-text fields, combined questions and printed-page layout carry over. Sign: coded fields hold phrases like "see notes". Fix: one data point per field, typed answers, units and options defined. See how to design an eCRF and paper CRF to eCRF conversion.
Every extra field is a query, a training point and a monitoring burden. Sign: sites say forms take longer than the visit. Fix: trace each field to an endpoint, a safety need or a regulatory requirement, and cut the rest.
Required flags on fields that may legitimately be unknown force sites to enter junk. Sign: "99" and "NA" entries. Fix: require only what the protocol needs, and give an explicit "not done" or "unknown" route.
Variable names invented ad hoc make exports hard to read. Sign: the statistician asks what "Q14b" means. Fix: use consistent names, and confirm the automatic codebook reads sensibly. See the data dictionary generator.
The first time anyone opens the dataset is at lock. Sign: reshaping work consumes weeks. Fix: export a test file in the first month and have the statistician comment on structure, labels and coding. Capture exports CSV or Excel with a data dictionary and SDTM datasets as SAS XPT with Define-XML.
Execution mistakes
These appear once the build exists and people start to use it.
Ranges set from memory fire on valid data or miss real errors. Sign: sites ignore queries because most are noise. Fix: have the medical monitor sign off ranges, test with boundary values and use priority levels. See edit checks software.
The scripts enter valid data in order. Sign: first week of enrolment surfaces skip-logic bugs. Fix: include negative tests, wrong roles and out-of-window visits. See EDC UAT and sandbox.
Everyone gets the same access because roles were never specified. Sign: site staff see coded IDs they should not, or unblinded staff see blinded views. Fix: write a role matrix before the build and test as each role. See blinded and unblinded role management and role-based access control.
Sites meet the system when the first participant is in the room. Sign: entry errors and protocol deviations from misunderstanding the forms. Fix: give sites sandbox access two weeks ahead and a one-page entry guide per form.
Amendments arrive and nobody knows who approves a form change or how old data is treated. Sign: edits made directly on live forms. Fix: define change control (request, impact assessment, test, approval) and use draft, approve and lock states on every form.
No go/no-go criteria, no first-week monitoring. Sign: issues found by sites rather than the team. Fix: use a written gate and a short hypercare period. See the EDC go-live checklist.
When the mistake is found
A qualitative view. Your own costs depend on study size and vendor terms.
| Mistake | Found in design review | Found in UAT | Found after enrolment starts |
|---|---|---|---|
| Wrong unit or range on a vital sign field | Edit the field | Edit, retest the form | New form version, query existing entries, retrain sites, explain in the audit trail |
| Required flag on a field that can be unknown | Untick it | Untick it, retest | Sites have entered placeholder values; clean up and document |
| Blinded role sees treatment arm | Fix role matrix | Fix and retest as that role | Potential unblinding event and deviation reports |
| Visit window set to the wrong day count | Edit schedule | Edit, retest | Out-of-window flags on valid visits; reassess deviations |
| Export never reviewed | Fix labels early | Fix labels | Re-map, re-export and re-validate analysis datasets |
The pattern is the point: the same fix gets slower, more visible and harder to explain on the record.
Build your forms, add edit checks, run practice participants as each role and open the export. The free sandbox has every feature, with no credit card and no time limit.
Prevention built in
Some of these mistakes grow out of the vendor model: a ticket for every change, a build you cannot see, and testing against a system you do not control. With a self-serve platform the team that knows the protocol builds and changes the forms directly.
Stage 1
DraftBuild and edit freely
Stage 2
ApprovedLocked for live use
Stage 3
In the casebookOnly approved forms
SDV requirement changed on Systolic BP
Data manager · Reason: critical safety value
Form approved: Vital Signs (VS) v2
Study owner · Locked for live use
Answer changed: Weight 68.0 to 68.5 kg
Site coordinator · Reason: transcription error
A safer path
Agree the schedule of assessments, primary and key secondary endpoints and safety reporting. Record the version the build follows.
Walk every field and ask what decision it supports. Remove, merge or justify it.
Who enters, who reviews, who sees identity, who is blinded. Test it as each role.
Enter ten deliberate errors (out-of-range, missing, wrong date order) and confirm each is caught or documented.
Give the statistician a test export and fix labels, units and coding while it is cheap.
Give sites sandbox access and a short entry guide before launch.
Use the go-live checklist, then monitor the first weeks closely.
The commercial side
Some implementation failures are commercial. A fixed build fee that excludes amendments produces change orders at the exact moment the protocol shifts. A vendor that must build every form means the sponsor cannot test a change without a ticket. A subscription that makes the data hard to export turns every later decision into a negotiation. These are choices made when buying, not when building. Hidden EDC fees lists the items to ask about, and the EDC procurement checklist covers the buying questions.
Capture lets a team build and change its own forms, with a free sandbox before go-live and paid plans from the point of going live with real participants. Specific pricing is not published here; ask for terms in writing. The aim of this page is general: whichever system you pick, spend the effort early.
Starting the build before the protocol is stable, and copying the paper CRF page by page. Both lock in a design that is expensive to change after enrolment starts.
It depends on the size of the build. What matters more is coverage: negative tests, every role and mid-study changes, not only the happy path. See the EDC UAT guide for a scope.
Usually the medical monitor or clinical lead, since ranges reflect clinical judgment. The data manager implements and tests them.
Yes, if someone owns the build baseline, role matrix and test plan, and if the system lets the team change forms directly. A free sandbox makes rehearsal cheap.
Validation is a shared responsibility. Capture provides 21 CFR Part 11-aligned controls and, for enterprise customers, documentation to support sponsor validation and UAT. The sponsor remains responsible for validated use.
Use a defined change process: request, impact assessment, test in the sandbox, approval, then a new approved form version. Capture forms move through draft and approved states with an audit trail.
Keep exploring
Free sandbox with every feature. No credit card, and you pay only when you go live with real participants.