WorkflowDesign & Development

Design Verification & Validation Protocol Authoring

Design verification and validation protocol authoring that argues each worst case, sizes the trials from the risk file and holds a protocol back from review while any choice lacks its reason.

See the demo Book a demo

Follow one protocol from copy to lab booking ↓

Watch with transcript →

The problem

A copied protocol brings the old worst case and trial count to a new design

A new verification protocol usually begins as the closest approved one, often written for the predecessor product. The copy brings its rate, its administration set, room temperature and a trial count sized for a hazard as it was rated back then. The engineer edits from memory and from the standard, and the argument for each worst case stays in that engineer’s head.

Criteria ride along too, and some trace to inputs the new design no longer has. Design quality sends the protocol back for missing rationale. The lab learns the real trial count only after approval, when a late change to the sample size moves its schedule.

Built forTest engineersDesign quality engineersSystems engineersBiostatisticiansTest lab supervisors

The Problem Today

Protocols start as copies, and the copied worst case and sample size were never checked for the new design.

12 h: to draft one protocol by hand our estimate

Argued by hand today

Worst case
Sample size
Criteria
Trace

Old numbers ride alongTrial counts sized for a hazard that has since changed
Pass lines proving nothingCriteria with no design input behind them

The Solution

Every choice re-checkedAgents read the input, the standard and the risk file for each choice
Worst case, with reasonsLowest rate, narrowest set, full back-pressure and the coldest room, each sourced
Sized by the hazardThe risk file’s severity picks the procedure row and the trial count
Approved after each reasonEach reason opened before approval, then an e-signature with its meaning

Built-in controls

Agents never approve, sign or change a record
Fewer trials need the biostatistician
A pass line with no input goes to its owner

What You Get

Protocols reach review with every reason attached, and the lab plan is sized to the hazard.

763 h: of drafting judgment saved across one sample program our estimate

Lab time known at draftingTrial count and lab days shown before approval
Variants reuse with a reasonBase protocols carried to a new model, each with its reason
Evidence on requestWho chose, why, the source and the signature, exported

One protocol, one morning

One design verification and validation protocol, re-argued choice by choice

Flow-rate accuracy at the low-rate end, for a new large-volume infusion pump. The request came in last week; this is the morning it goes from copy to booked lab time.

demoThe pump maker, its staff and each number shown here are sample data.

01Start
Seven choices copied from the predecessor pump, and not one has a reason
Test engineer
The composer opens on the protocol as it was copied. Each of its seven choices wears the same amber chip: reused, no reason yet. The rate is 25 mL/h, the set is the standard one, back-pressure is 0 mmHg and the room is at 23 °C. The meter in the header reads 7, and the protocol cannot go to review until it reads 0.
02Start
The 59 trials came from a Minor row in a risk file that has moved on
Test engineer
One click on the sample size chip opens the protocol it was copied from, approved for the predecessor pump: 25 mL/h on a standard set, 59 trials, zero failures, from the Minor row of the risk file of that time. The new risk file has split low-rate under-delivery into its own hazard, so the copied count answers an old question.
03Agent check
Serious in the risk file, so 59 trials become 299
Test engineer
Seven agents work through the choices on the rail. Rate drops to 1.0 mL/h, with a micro-bore set from the lowest-diameter lot on hand, +300 mmHg and 10 °C. In the risk file the hazard is rated Serious, so the sampling row turns 59 into 299. This worst case uses 3.7% of the 5% allowance, on the margin bar.
049:52 AM
The start-up line traces to no input, so its owner rules on it
Systems engineer
The reviewer agent flagged one carried-over criterion: flow begins within 15 minutes of start. No requirement for the new pump sets a start-up time. Dropping a pass line makes the test looser, so the author cannot remove it. The systems engineer types her own reason, keeps the start-up graph as recorded data and signs. Her ruling waits for design quality to read it.
059:52 AM
Across the program, 7 of 171 criteria trace to nothing
Systems engineer
Her own list covers the drafted protocols: 164 criteria trace to an input and 7 to none. Two were settled by adding a design input, others were ruled recorded data, and 2 are still open with her. A test engineer can only ask. Turning a pass line into data belongs to the person who owns the inputs.
069:54 AM
The meter reads 0, and the protocol goes to review signed
Test engineer
With six choices accepted and the ruling in place, the meter drops to 0 and the send button opens. She signs it as authored, with her user ID and password. It now sits with the design quality engineer, on day 0 of a five-working-day clock, and each column shows who accepted which choice.
07Review
Approval stays locked until each reason has been opened and read
Design quality engineer
The approver reads the protocol as a document, with three items to open before she can sign: the worst case, the sample size with its formula, and the criteria. The last carries the systems engineer’s ruling in her own words, marked not compared. An item counts as read only after it has been open long enough to read.
0810:00 AM
Ten of ten reviewer checks pass, then approved for execution
Design quality engineer
The reviewer agent’s ten checks are computed from what the protocol says now: three builds, the lowest-diameter lot, the coldest temperature, balances calibrated past the test window. All pass. She signs approved for execution, and that signature is what puts the start-up ruling into effect. The screen refuses an approval from the protocol’s own author.
0910:02 AM
299 trials booked as 747.5 station-hours, about four lab days
Test lab supervisor
The lab sees a protocol only once it is approved. Each trial holds a station for 2.5 hours, so 299 trials come to 747.5 station-hours: 62.3 hours across 12 gravimetric stations, about 4 lab days on two shifts. The supervisor books the environmental chamber and the stations. He can book, and he cannot edit or sign anything.

The risk file’s severity picks the row, and the row fixes the trial count

Attribute plans with no failures allowed, from the sampling procedure’s table. The protocol was copied from the Minor row; the hazard it verifies is now rated Serious.

Severity row	Confidence and reliability	ln(1 − C) ÷ ln(R)	Trials
Negligible	95% and 90%	28.43	29
Minor	95% and 95%	58.40	59
Serious	95% and 99%	298.07	299
Critical	99% and 99%	458.21	459

demoSample data, from the working solution. The 299 is split as 23 pumps from three builds, each pump run on 13 new sets from one lot.

299 trials cost about four lab days, where the copied 59 cost one

Sets and water are conditioned at 10 °C in the chamber, in batches, so conditioning does not hold a station.

Trials	299
Station time per trial20 min to prime, 2 h to run, 10 min to clear, our estimate	2.5 h
Station-hours	747.5
Across 12 gravimetric stations	62.3 h
On two 8-hour shifts	about 4 lab days
The copied 59 trials, by the same sum147.5 station-hours, 12.3 h	about 1 day

demoSample data, from the working solution.

Around the protocol

Where the program’s verification protocols stand, and who holds each

The program around this one protocol, as the other roles see it.

Protocols list for one device program: 212 protocols filtered by planned, with author, in review and approved, with the drafting estimate — **The whole program**All 212 protocols by status, type and owner. The banner holds our estimate: about 3.6 hours of drafting judgment saved per protocol, about 763 across the program.

Variant reuse screen: 24 base protocols, 16 reused with a reason, 5 delta protocols and 3 new, with a sampled sign-off — **Reuse on a variant**For the dual-channel model, 16 of 24 base protocols reuse with a reason, 5 need a delta and 3 are new. Signing waits until a sample is opened.

Design quality review queue: 18 protocols in review, 2 due within a day, 1 over the clock, with reviewer flags per protocol — **Review queue**Protocols waiting for design quality, each on a five-working-day clock from arrival. One is over and two are due within a day, with the reviewer agent’s flags beside them.

Worst-case configuration coverage matrix for flow-rate accuracy: rate, set, back-pressure, temperature, head height and set age — **Worst-case coverage**Each factor the input claims, from worst end to other end, against the protocol that tests it. The blood set, −100 mmHg and 40 °C show as planned gaps.

Design verification sample size table by severity, with confidence, reliability, trials and drafted protocols per row, and a calculator — **Sampling table**The procedure’s rows with the drafted protocols in each, and a calculator that shows the formula. A plan that departs from the table goes to the biostatistician.

Audit trail for one protocol: accepts, the ruling, sending to review, the approval and the lab booking, each with session and meaning — **Audit trail**Who did what, in which session, why, and the meaning of each signature. Agent lines are marked, and no agent line carries a signature.

demoSample data, from the working solution.

Five people decide, and the author signs only as author

Each looser move goes to the person who owns it. The agents propose and cite; they hold no signature.

Test engineer

Sees: Each choice with its reason, sources and confidence
Signs: Accepts or edits each choice, then signs the protocol as authored

Systems engineer

Sees: Pass lines that trace to no design input
Signs: Rules a line recorded data, or adds an input, with a typed reason

Biostatistician

Sees: Plans that depart from the procedure table
Signs: Agrees or declines each departure, with a reason

Design quality engineer

Sees: The worst case, the sample size and the criteria, each opened first
Signs: Approved for execution, and never on their own protocol

Test lab supervisor

Sees: Trial counts and lab time for approved protocols
Signs: No signature: books chamber and station time only

The agents

Eight agents draft and check, and none of them approves a protocol

Each reads one kind of record and hands its work to a named person or to the next agent.

Input reader

Reads the design input, its tolerance, range and conditions, with the passage.

Never edits an input

Standards reader

Finds the method and clause in the company’s licensed standards and guidance.

Never reproduces licensed text
Never says a device conforms

Risk file reader

Reads the hazard chain and its severity from the risk management file.

Never changes a severity
Never reads the post-production section

Worst-case planner

Proposes each worst-case condition with its reason, source and confidence, and lists what it considered.

Never proposes a condition the labeling does not claim

Sample size agent

Applies the procedure row for the severity, shows the formula and splits trials across builds.

Never picks a looser row

Criteria writer

Writes acceptance criteria from the input’s numbers, with the arithmetic shown.

Never invents a limit

Protocol reviewer

Checks sources, trace, test articles and calibration, and flags what fails.

Never approves, fixes or removes a criterion

Reuse checker

Compares a variant with the base protocols and proposes reuse, a delta or a new protocol.

Never reuses without a reason

Fewer trials, a dropped pass line and a release question each go to a person

A move that tightens the test applies at once. A move that loosens it waits for its owner.

Fewer trialsThe test engineer cannot cut the sample below the procedure row. A smaller plan, such as a variables plan, waits for the biostatistician’s signature.

Removing a pass lineOnly the systems engineer can turn a criterion into recorded data, and the ruling takes effect when design quality has read it and approves.

Turning off a reviewer checkAn administrator can propose it. A design quality engineer signs, and never on a proposal of their own.

Clinical and release questionsQuestions about use on a patient, off-label use or releasing product get no answer, and name the person who decides.

Post-production dataThe risk file reader reads the hazard chain and severity only. The post-production section of the risk file stays unread.

What test engineers ask about design verification and validation

What is design verification and validation protocol authoring?

Design verification and validation protocol authoring is writing the test plans that prove a device meets its design inputs. Here agents draft each protocol from the design input, licensed standards, the risk file and the sampling procedure, and attach a reason and a source to each worst case, sample size and criterion.

How is the design verification sample size chosen?

From the severity of the hazard the protocol verifies. The sample size agent reads that severity in the risk management file, applies the matching row of the company’s sampling procedure, and shows the formula and the split across builds. It cannot pick a looser row; a departure goes to the biostatistician.

How does it argue a worst case configuration?

The worst-case planner proposes each condition that pushes the result toward failure, from the input’s ranges, the tolerance stack, the labeled accessories and the standard method, each with sources and a confidence. It also lists what it set aside and which protocol covers that instead.

Can the test engineer change what the agents propose?

Yes, with a reason in their own words. Filler is refused, the approver sees the edit as the author wrote it, and an edit that breaks a reviewer check becomes an open flag that blocks approval.

Does it use our own standards and sampling procedure?

Standards come from the company’s licensed copies, cited by clause, not reproduced. The sampling procedure is read only here and changes through document control.

Can a variant reuse the base model’s protocols?

The reuse checker compares parts, firmware and housing and proposes reuse, a delta protocol or a new one, each with a reason. A design quality engineer signs the reuse only after opening the higher-severity rows and a sample of the rest.

What if the requirements system cannot be read?

The sources screen says so at the top. Agents work from the last copy read, and say that later changes will not show until a read succeeds.

Design & Development

Bring one protocol you copied
and the input it now has to prove.

We’ll check each choice against your risk file and sampling procedure and show you which ones have no reason.

Book a demo Browse solutions

Design Verification & Validation Protocol Authoring

A copied protocol brings the old worst case and trial count to a new design

The Problem Today

The Solution

What You Get

One design verification and validation protocol, re-argued choice by choice

Seven choices copied from the predecessor pump, and not one has a reason

The 59 trials came from a Minor row in a risk file that has moved on

Serious in the risk file, so 59 trials become 299

The start-up line traces to no input, so its owner rules on it

Across the program, 7 of 171 criteria trace to nothing

The meter reads 0, and the protocol goes to review signed

Approval stays locked until each reason has been opened and read

Ten of ten reviewer checks pass, then approved for execution

299 trials booked as 747.5 station-hours, about four lab days