Block3 FMEA help and guide

Build a Process FMEA one part at a time: generate the failure modes and causes for each process, rate the risk, and export a clean Excel workbook.

Getting started

1. Create your account at block3.dropfeed.io. You start on the Free plan — no credit card required.
2. Build a process. Go to Process library → New process and name the operation (e.g. "CNC Drilling") and what it does. Click Draft requirements & failure modes and Block3 FMEA writes its requirements, potential failure modes, effects, causes and current controls — or add a requirement and write it yourself. Nothing is generated you can't edit.
3. Rate the risk. Fill in Severity, Occurrence and Detection (1–10). SEV×OCC, RPN and Action Priority calculate as you type. Rate on the FMEA, where the numbers belong to that part. A rating set in the library comes across as the starting point.
4. Repeat for every operation the part uses. Each one is analyzed once and lives in the library from then on.
5. Assemble the FMEA. Click New FMEA, enter the Part number, Revision, Part description and Standard, then tick the processes this part actually uses. The document is built from them.
6. Set the line numbers. On the document, give each step its Line # so the form prints in your process order — Op 105 before Op 120, whichever you added first. If an operation runs twice, add the process twice and give each placement its own number and title.
7. Answer the action items. Any cause your ratings pushed to High Action Priority, or past your RPN threshold, appears under Action items with a suggested Recommended Action. Accept it, edit it, put it off, or delete it.
8. Export. A clean PDF with no watermark on any plan, including Free. The editable Excel .xlsx with merged cells needs a paid plan or a single-document unlock.

Why processes come first

Most FMEA tools make you re-describe the same operation on every part that uses it. Ten parts through the same drill means ten copies of the same analysis, and when a control changes you have ten documents to remember. Block3 FMEA inverts that: the process library holds the analysis, and an FMEA is the list of processes a part uses. Fix a control once and every document that uses that process is correct.

Process steps are named by operation with no made-up operation numbers, so you number and align them to your own process flow and control plan. Set the Line # on each step inside an FMEA and the steps sort by it everywhere, on screen and in both exports, so Op 105 prints before Op 120 no matter which you added first. Whole numbers only, and numbering by tens leaves room to insert an operation later. Steps you leave unnumbered keep the order you added them in and follow the numbered ones.

What you change where

The analysis is read-only on an FMEA. Requirements, failure modes, effects, causes and controls come from the library and are edited there.

What belongs to the part, you set on the document: the Line #, the title the step carries on this form, its Severity, Occurrence and Detection, and which processes are in it. You can also remove a step.

The split is deliberate. A process is shared by every FMEA that uses it, so editing the analysis on a document would either quietly diverge from the library, which is the drift the library exists to stop, or push a change into every other document from a screen that gives no hint that is what it is doing. Ratings are the other way round: the same operation carries different risk on a new part than on a mature one, so they belong to the part.

Using the same process more than once

A routing often runs an operation at more than one stage. You can add the same library process to an FMEA as many times as you need: tick it in the picker again and it comes in as a second step.

Each placement is its own step on the form. It gets its own Line #, its own ratings, and its own title on this FMEA so the two are tellable apart in the export. The title is blank by default, which means the library name is used. Both placements stay tied to the one analysis in your library, so a fix made there still reaches both.

The library is the master copy

When you edit a process in the library and it saves, that change is pushed into every FMEA already using it. You do not remove the process and add it back, and you do not open those documents; they are updated where they sit, and the editor tells you how many changed. Treat the library as your controlled master: a correction there is a correction everywhere, including on documents you finished months ago.

What stays with the document rather than the process, because it belongs to the part and not to the operation: the recommended action you committed to on a line, the line number and title that operation has on this part, and the S, O and D ratings you set. A library update keeps your ratings wherever the row they belong to is still there. Where a requirement, failure mode or cause has been rewritten, that row comes back unrated rather than carrying a score that was assigned to something else. Everything else (requirements, failure modes, effects, causes and controls) comes from the library.

The process editor saves on its own as you work, so nothing is lost to a closed laptop or a dropped connection. It also means there is no moment where an edit is finished but not yet shared. If a process is used across parts that are already released, review it before you change it: the correction reaches all of them.

Organizing parts: families and sub-assemblies

Two different relationships, and they do different jobs.

• Part family groups parts that belong to one deliverable set. Members sit level with each other, share an FMEA type and a quality standard (both inherited, so they can't drift), and a new member starts with the family's processes pre-ticked — a starting point you can change.
• Sub-assembly of hangs a document underneath a higher assembly and indents it there. It inherits the type and standard from the assembly above, but chooses its own processes — a sub-assembly can share nothing with the part above it.

About the ratings (S, O, D)

Ratings belong to the FMEA. The same operation does not carry the same risk everywhere: a new part and a mature one differ on Occurrence, and an inspection early in the routing differs from the same inspection late in it. So Severity, Occurrence and Detection are set on the document, and each placement of a process carries its own.

You can still rate a process in the library, and those numbers come across as the starting point when it is added to an FMEA. Once you change them on a document they are yours: a later library update keeps them, as long as the row they belong to is still there.

Block3 FMEA deliberately leaves Severity, Occurrence, and Detection blank: those ratings depend on your team's scales and real data, and shouldn't be guessed by AI. Once you fill them in, the tool computes them for you:

• SEV × OCC: a criticality figure.
• RPN: Severity × Occurrence × Detection.
• Action Priority (AP): High / Medium / Low for each cause.

When an action is required

A row is flagged as owing a Recommended Action when either of two things is true. The flag appears in the editor and prints in both exports as “Action required” with the reason, so an empty Recommended Action cell always means answered rather than overlooked.

Flagged causes are collected under Action items, with a suggested action written for each one. That page is the only place an action is written or changed — there is no Recommended Action box in the editor, because an action is a decision about risk and a text field in the middle of a rating screen invites it to be answered in passing, or skipped and never returned to. Accept an action and it is written into the document; put it off; or delete it if the cause genuinely needs none. A banner follows you around the app while any are unanswered.

Lower a score later and the action stops printing on your exports without being deleted: the item moves to No longer flagged and comes back exactly as it was, wording included, if the ratings go back up. Only you can delete one for good.

• Action Priority is High. This one is always on and cannot be switched off — it is the standard’s own trigger: act, or record why you did not.
• RPN reaches your threshold. This is a setting, found in Settings → Flag an action when RPN reaches. It is on by default at 150. Enter 0 to turn it off and leave Action Priority as the only trigger, or enter whatever number your customer’s supplier manual mandates.

Why 150 rather than 100: Action Priority High already flags a large share of rows on its own. A threshold of 100 roughly doubles what gets flagged, and most of what it adds is already rated AP Medium — the same information twice, which trains people to ignore flags. 150 still catches the case Action Priority is blind to, a low-severity failure that happens often and is hard to detect, without burying it.

The two triggers answer different questions and neither replaces the other. Action Priority asks must this be acted on, weighted so severity dominates. The RPN threshold asks is this noisy. A safety failure caught reliably can be AP High at an RPN in the thirties; a cosmetic defect that happens constantly can pass RPN 100 and still be AP Low.

What the AP column means

Action Priority ranks each cause using all three ratings together, rather than by multiplying them. Multiplying ranks badly. A rare, well-detected safety failure scores 9×2×2 = 36. A frequent cosmetic one scores 4×3×3 = 36. Identical RPN, nothing like the same risk — and a team working top-down through an RPN column would give them equal attention.

AP weights Severity first, then Occurrence, then Detection. A high-severity item therefore stays high priority even when you detect it well, because detecting a hazard reliably is not the same as removing it.

• High — act on it, or record why you decided not to.
• Medium — action should be taken.
• Low — action is optional.

AP stays blank until Severity, Occurrence and Detection are all entered, and recalculates the moment any of them changes. RPN is still exported alongside it, because plenty of customer requirements continue to ask for it.

Each rating attaches to a different part of the row, which is why the exported table merges the way it does: Severity comes from the effect, so it spans the whole failure mode. Occurrence and Detection belong to the individual cause and its controls, so they sit on the cause row — and so does the AP computed from them. Two causes of the same failure mode can, and often should, end up with different priorities.

Choosing your standard

Pick the quality standard you work to: ISO 9001, AS9100, IATF 16949, and more, or choose Other and type any customer-specific standard. Each built-in standard carries its own analysis focus, so the draft reaches for what that standard expects: foreign object damage, lot traceability and special-process control under AS9100; special characteristics, control plan alignment and error-proofing under IATF 16949; operator exposure and guarding under ISO 45001. A standard you type yourself is used by name only.

Medical device work is not offered as a built-in standard. A device FMEA has to stay consistent with the risk management file it belongs to, and a draft that looks finished but does not trace into that file helps nobody. You can still type any standard into Other, where it is used by name only.

Plans & usage

Only one thing is metered by the month for cost reasons: AI drafts. Writing requirements by hand is never limited, and assembling an FMEA from processes you already have costs no draft at all. Everything a Free account produces is usable — there is no watermark and nothing is crippled; the paid line is the editable spreadsheet and the volume.

• Free — 25 AI process drafts per month, 20 saved processes, and 3 FMEAs in total. The PDF export carries no watermark, so the documents you produce on Free are real documents you can file, send and use. The three are a one-time allowance rather than a monthly one: enough to build a library and finish real work before deciding anything.
• Single document — $39 once — bought from a finished FMEA, it unlocks the editable Excel for that document, adds one more FMEA to your account, and lifts the library cap entirely for 30 days — save as many processes as the work needs. AI drafts stay metered but rise to 50 extra, since drafts are the only thing that costs anything to serve. After the 30 days the document is yours to keep and re-read, and any processes you built stay where they are — the cap only stops you adding new ones until you are back under it.
• Pro — $79/month, or $790/year (two months free) — 200 AI drafts per month, unlimited saved processes and FMEAs, and the editable Excel export.

Need more help?

Email support@dropfeed.io and we'll get back to you.