A food recall plan is a written plan that assigns responsibility and spells out procedures for pulling unsafe product from the market: notifying customers who received it, notifying the public when needed, checking the recall worked, and disposing of what comes back. Under FSMA, FDA requires one in writing for any food with a hazard requiring a preventive control (21 CFR 117.139).

Recalls are decided in hours by people who are tired and scared. The plan exists so those hours run on decisions you made calmly, in advance. This guide covers the FDA recall classes, exactly what 21 CFR 117.139 requires, the recall team roles, the decision tree, the communication fan-out, and the timeline your plan should be built to hit, plus the practice regime that keeps it real.

What are FDA recall Classes I, II, and III?

FDA assigns every recall a classification based on the health hazard the product presents (defined in 21 CFR 7.3 and used in FDA's enforcement reports):

ClassDefinitionFood examples
Class IReasonable probability that use or exposure will cause serious adverse health consequences or deathPathogen contamination (Listeria, Salmonella), undeclared major allergens, hazardous foreign material
Class IIUse or exposure may cause temporary or medically reversible adverse health consequences; remote probability of serious harmSome quality-related contamination, certain labeling violations with limited health impact
Class IIIUse or exposure not likely to cause adverse health consequences, but the product violates regulationsMinor labeling defects, quality standard violations

Classification matters because it drives depth (how far down the chain you recall, wholesale, retail, or consumer level), publicity (Class I usually means a public press release), and effectiveness-check intensity. Note the related but different actions: a market withdrawal removes product for a minor issue that isn't a violation, and a stock recovery pulls product that never left your control. Your plan should define all three, because calling an event the wrong thing has regulatory consequences.

Meat, poultry, and egg products run through USDA FSIS rather than FDA, with an equivalent class system and its own recall process dual-jurisdiction companies need both paths in the plan.

What does FSMA actually require? 21 CFR 117.139

For any food whose hazard analysis identifies a hazard requiring a preventive control, 21 CFR 117.139 requires a written recall plan that includes procedures, with assigned responsibility for each, to:

Two more legal gears sit next to that rule. The Reportable Food Registry requires registered facilities to report to FDA within 24 hours when there is a reasonable probability that a food will cause serious adverse health consequences. And FSMA gave FDA mandatory recall authority, if a company refuses to recall voluntarily, FDA can order it. In practice nearly all recalls are voluntary, run by the company, with FDA reviewing strategy and monitoring effectiveness. FDA's draft guidance chapter on recall plans is the best free blueprint for writing yours.

Who is on the recall team? Seven roles

Names and backups, not departments. Put phone numbers in the plan and keep a printed copy, the recall that starts at 6 p.m. Friday does not wait for the org chart.

  1. Recall coordinator. Owns the event end to end: convenes the team, makes the recall/withdrawal call with QA and legal input, talks to FDA. Usually the quality or food safety leader. Needs a named deputy.
  2. Traceability lead. Produces the affected-lot picture: what was made, from which inputs, shipped where, how much remains. The clock on the whole event runs on this person's records, see traceability in manufacturing.
  3. Operations lead. Holds remaining inventory, stops further production and shipment, quarantines returns, manages reconditioning or destruction.
  4. Customer notification lead. Owns the direct-consignee calls and written notices: who received affected lots, confirmation of receipt, return/disposal instructions, response tracking.
  5. Communications lead. Public notice, press release, website statement, consumer hotline, coordinated with FDA when public notification is warranted.
  6. Regulatory/legal lead. RFR submission, FDA/FSIS liaison, records preservation, insurer notification.
  7. Scribe. Underrated and essential: timestamps every decision, call, and quantity into one event log. That log is your defense in the after-action review and any regulatory follow-up.

How is the recall decision actually made?

Every trigger, a customer complaint, a positive environmental or product test, a supplier notification, an internal discovery at a changeover, should route into the same evaluation. The tree keeps 2 a.m. decisions consistent:

The recall decision tree trigger: complaint · test · supplier notice · internal find is the product violative or unsafe? (health hazard evaluation) no violation, minor issue → market withdrawal has product left your control? check shipping records still in-house → stock recovery + hold RECALL, assess hazard severity, notify FDA, assign class w/ agency Class I · serious harm likely consumer level + publicity Class II · temporary or reversible effects Class III · violative, harm unlikely 24-hour RFR report runs in parallel when serious harm is reasonably probable
One tree for every trigger. The left exits (withdrawal, stock recovery) are why definitions belong in the plan, calling a recall a withdrawal, or vice versa, creates its own problem.

Who gets told, in what order?

A recall is mostly a communication event. The plan pre-writes the fan-out: templates, contact lists, and owners for every arrow.

Recall communication fan-out Every arrow has a template, an owner, and a deadline Recall team coordinator + log FDA / FSIS district + RFR within 24h direct consignees written notice + confirm internal: floor, sales, leadership, switchboard public: press release, website, hotline suppliers · insurer · legal counsel rust = regulatory-clock arrows: fastest deadlines first
The fan-out from the recall room. Consignee notification and regulator contact carry the hard deadlines; internal notification is the one plants forget until the hotline rings the front desk.

What does the first week look like?

Build the plan against a clock. Hours are targets to design for, not regulatory promises, but if your plan cannot hit them on paper, it will not hit them under stress:

Recall timeline: hour 0 to day 7 The first week, planned in advance hour 0 team convened, product held hours 0–4 trace complete: lots, quantities, ship-to within 24h RFR report if serious; FDA contacted day 1–2 consignees notified, public notice if warranted day 2–7 effectiveness checks, returns & disposition then: root cause, corrective action, after-action review, plan revision the hours 0–4 box is the one you rehearse, everything downstream waits on the trace
The traceability window sets the pace of everything after it. If assembling the lot picture takes two days, every regulatory and customer clock starts two days late.

Notice what gates the timeline: records. The 4-hour trace assumes you can join production lots, ingredient lots, and shipments quickly. That is precisely the muscle mock recalls exercise, the capability FSMA 204 will demand within 24 hours for listed foods, and the reason plants pull production and shipping records out of paper and into live systems, plant-wide, searchable operational data is the core of what Harmony builds (see the platform).

The numbers that justify the binder

How do you keep the plan alive?

A recall plan written once and filed is a compliance artifact, not a capability. Keep it real with a standing regime: run a mock recall at least annually (GFSI-benchmarked schemes expect a documented traceability/recall test at that frequency, and most plants that take it seriously run two, one paper trace, one full simulation with calls); update the contact lists quarterly, because phone numbers rot faster than procedures; re-verify the plan after every change to products, customers, co-packers, or systems; and grade every exercise against the clock, trace time, notification time, mass-balance percentage, so the next revision has targets. Your HACCP system decides how unlikely the recall is; the plan decides how bad it gets when unlikely happens anyway.