River
Y CombinatorBacked by Y Combinator

Free template

Crisis Communication Plan Template

A crisis communication plan in ten files: each scenario's deadlines, whether you can meet them, holding statements, a contact tree and a drill log.

Free download · No account needed

What you get

Pack of 10 files: 5 documents, 5 spreadsheets

Crisis Communication Plan Template
  • Clock MethodDocument
  • Scenario RegisterSpreadsheet
  • Notification RequirementsSpreadsheet
  • Escalation LagSpreadsheet
  • Holding StatementsDocument
  • Escalation ProcedureDocument
  • Spokesperson GuidanceDocument
  • Contact TreeSpreadsheet
  • Drill LogSpreadsheet
  • Post-Incident ReviewDocument
RiverCrisis Communication Plan Template

Scenario Register

One row per obligation, and the window is what is left

drafting window = deadline − hours gone before comms was told − hours the facts take

ScenarioClockDueLagFactsWindowApprovals
Misconduct allegation against an officerMedia convention, the outlet’s stated deadlineMedia4h0.52.01.5h5.5h
Death of a field engineer on a customer siteOSHA, 8h from the employer learning of itStatutory8h1.55.21.3h1.6h
Access unit unlocks on power loss, two injuriesCPSC, within 24h of obtaining the informationStatutory24h31.012.0−19.0h4.0h
Personal-data breach reaching EU data subjectsArticle 33, 72h from becoming awareStatutory72h19.058.0−5.0h3.5h
Same breach, largest customer’s data agreementContractual notice, 24h from becoming awareContract24h19.04.01.0h2.2h
Cyber incident determined material to investorsForm 8-K Item 1.05, four business daysStatutory96h0.024.072.0h6.0h

What the subtraction found

Nine obligations. Five cannot produce an approved statement inside the deadline, because the window is smaller than the approval path. Two are already gone: the lag and the facts use up the whole deadline, and on one the clock ran out before comms was told the incident existed. The 30-minute holding-statement convention binds in none of them.

The row with a zero lag is the only one where comms is in the room when the clock starts. That is the whole difference, and it is a seating arrangement rather than a process.

Page 1 of 2738 words

How to write a crisis communication plan

A crisis communication plan decides in advance who says what, to whom and how fast when something goes wrong. A good one lists the incidents you could face and every deadline each one triggers. It names the spokesperson, the approvers and a deputy for each, keeps pre-approved holding statements ready, and maps how bad news reaches those people. It is tested in drills, not just filed.

Build it around the earliest deadline, not the first call from a reporter. Many legal clocks start when someone in the company learns of a problem, which can be hours before the communications team hears.

A woman pins a printed organization chart to a whiteboard covered in handwritten notes, including talking points, while colleagues work at computers behind her
Agree on roles and talking points before an incident, not during one. Photo: CDC on Unsplash

List your scenarios and their deadlines

Start with the incidents you could realistically face: a serious injury at work, a data breach, a faulty product, an allegation against a leader, a key supplier failing. Each can trigger three kinds of deadline: legal ones, notice clauses in customer contracts, and media ones, such as a reporter's request for comment.

Give each deadline its own row. One breach can owe notices to a regulator, a customer and the people affected, each on a different clock. Put them in a single row and the tightest deadline hides behind the loosest.

Common deadlines and when their clocks start

IncidentDeadlineThe clock starts
Work-related death (US, OSHA)8 hours1At the death, or when it's reported to you or your agents
In-patient hospitalization, amputation or loss of an eye (US, OSHA)24 hoursWhen it happens, or when it's reported to you or your agents
Product defect that could create a substantial risk of injury (US, CPSC)24 hours2When the company has information reasonably supporting that conclusion
Personal data breach likely to put people's rights at risk (EU, GDPR)72 hours3When the company becomes aware of the breach
Material cybersecurity incident at a US public company (SEC)4 business days4When the company determines it is material
A reporter's request for commentThe outlet's deadlineWhen the request arrives

Rules as of October 2026. Check the current wording, and which ones apply to you.

Read your largest customer contracts for notice clauses, which can be shorter than the law's. The European Commission notes that under the GDPR, a company processing personal data on a customer's behalf must report every breach to that customer3. In the US, the FTC points out that every state has its own breach notification law5.

Have a lawyer confirm the list

Notification rules vary by country, state, industry and contract, and they change. Ask a lawyer which ones apply to you and what each notice must say. This guide and the template are not legal advice.

Find when each clock starts

Write down the event that starts each clock, in the rule's own words, and whose knowledge counts.

The CPSC, for example, treats product safety information received by any employee able to appreciate its significance2 as known to the company. A defect report sitting in an engineering queue is already known, even if nobody in charge has read it. In ordinary circumstances, the rule allows at most 5 days for it to reach whoever handles reporting.

Other clocks start at a decision. The SEC ties the Form 8-K deadline not to discovery but to the determination4 that an incident is material, and that call must come without unreasonable delay. If your communications lead sits in that meeting, no time is lost passing on the news.

Then measure the delay: the hours between a clock starting and your team being told. Take your last real incident of any size, find those two timestamps and subtract. If nobody kept a timeline, rebuild one from alerts and chat logs. With no incident to measure, leave the cell blank and plan a drill. A flattering guess is worse than a blank, because nobody questions it.

Check that you can meet each deadline

Now do the arithmetic. From each deadline, subtract the hours lost before your team was told, then the hours needed to confirm the facts the notice can't leave out. What's left is your drafting window. Compare it with how long sign-off really takes at 2 a.m.

ExampleTwo rows from a scenario registerAn illustrative example written for this guide. The company and its timings are made up.

Field engineer dies on a customer site. OSHA report due in 8 hours. Communications told after 1.5 hours, facts take 5.2, so 1.3 hours left. Sign-off takes 1.6. Short by 0.3.

Customer data breached. Contract requires notice within 24 hours of awareness. Told after 19 hours, facts take 4, so 1 hour left. Sign-off takes 2.2. Short by 1.2.

  • Two causes, two fixes. Row one closes if the safety lead may release a pre-approved statement. Row two needs IT to call communications as soon as a breach is confirmed.

If sign-off is too slow, pre-approve the first statement, let someone lower down release it, and name a standing deputy for anyone hard to reach at night. If the window is negative, the delay and fact-finding have used up the deadline, and no approval change helps. Shorten them instead, starting with the people who discover incidents: they need to know their discovery can start a clock.

Write holding statements in advance

A holding statement is the short first response you issue while facts are still coming in. Crisis researcher W. Timothy Coombs, writing for the Institute for Public Relations, advises a first response within the hour6 that is accurate and consistent. It should tell people anything they need to stay safe, say what happened and express concern for anyone hurt.

Keep each one under 75 words, in four parts: what you know, what you're doing, who is affected and what they should do, and when you'll say more. Leave these out:

  • A cause, until it is confirmed
  • A count of people affected that hasn't been checked twice
  • An update time nobody has agreed to keep
  • Stock lines such as "we take this extremely seriously"

Hold back nothing people need, though. In its data breach guide, the FTC warns against misleading statements, or withholding details that help people protect themselves5.

Draft first for the shortest drafting windows, not the likeliest scenarios. Coombs recommends having your legal team review and pre-approve each draft. Write any statement that goes out with a regulatory filing in the same sitting as the filing, so the two never contradict each other.

Name your audiences, spokespeople and approvers

Work out who needs to hear from you. Ready.gov, the Department of Homeland Security's preparedness site, lists audiences that will want information7: employees, customers, suppliers, management, and government officials and regulators. It also advises deciding in advance who will speak to the media and preparing them with talking points. Pick a spokesperson and a named deputy for each scenario.

Then build a contact tree: each role, a deputy, a personal mobile and a second route that works with company systems down. Time how long each approver takes to answer at night, and check it fits the drafting window.

Tell employees before the public, or at the same time. Coombs lists including employees in the initial response6 among his best practices. For planned news, such as a reorganization, use an internal announcement template instead.

Test your crisis communication plan

Coombs recommends updating the plan and testing it at least once a year. A 2-hour tabletop drill is a good start, as long as people act rather than describe what they would do.

  1. Pick the weakest row

    Choose the scenario with the least time to spare, or a delay nobody has measured.

  2. Start at the real zero

    Begin when the person who would really find the incident does, without warning communications.

  3. Call the real numbers

    Run it out of hours and phone each approver on the number in the plan.

  4. Change a number

    Update the register with what you found. A drill that changes nothing hasn't finished.

Review real incidents the same way, false alarms included, using timestamps rather than memory.

Before you rely on the plan

  • Each deadline an incident triggers has its own row
  • The delay before communications hears is measured, not guessed
  • Every drafting window beats its sign-off time, or has a fix and an owner
  • Every spokesperson and approver has a deputy and a tested night number
  • The plan was drilled in the last 12 months

The template on this page builds this arithmetic into its scenario register, and Edit with AI has River check a plan you already have. Pair it with a security incident response plan for the technical side of a breach, and a business continuity plan to keep work going.

How to use this template

Work from the five Word documents and five CSV sheets on your own, or open them in River. The sheets come filled in for a made-up company, so you can follow the arithmetic before replacing its rows with yours.

With Edit with AI, River asks what you make or sell, where your customers and their data are, and who you employ. It lists the notification deadlines those answers point to, and checks your current plan against them if you send it. Then it asks how long your last incident took to reach your communications team, and how long sign-off takes at 2 a.m. It tells you plainly whether your tightest deadline can be met.

  1. Download it or open it in River

    Download the five Word documents and five CSV sheets free, or open them in River to work through them with AI.

  2. List the deadlines you're on

    Record each legal and contract deadline and the event that starts it in Notification Requirements, or let River draft the list.

  3. Check each deadline can be met

    In the Scenario Register, subtract the measured delay and the time the facts take, then compare what's left with sign-off.

  4. Write statements, then drill

    Draft holding statements for the shortest windows first, then run a 2-hour tabletop on the scenario with the least time to spare.

What's in the template

The sheets that treat the deadline as arithmeticNine obligations with the subtraction run on each, where the measured lag came from, and what six hours of tabletop changed.
Crisis Communication Plan Template
RiverScenario Register

Scenario Register

Nately: US-listed, connected building-access hardware plus a SaaS platform, EU customers, field engineers on customer sites. Nine obligations, sorted by deadline.

ScenarioClockDueLagFactsWindowApprovals
Misconduct allegation against an officerMedia4h0.52.01.5h5.5hshort 4.0h
Death of a field engineer on a customer siteStatutory8h1.55.21.3h1.6hshort 0.3h
Field engineer hospitalised as an in-patientStatutory24h6.04.513.5h1.6hfits
Access unit unlocks on power loss, two injuriesStatutory24h31.012.0−19.0h4.0hmissed
Same breach, largest customer’s data agreementContract24h19.04.01.0h2.2hshort 1.2h
Key component supplier enters administrationMedia48h8.06.034.0h1.0hfits
Personal-data breach reaching EU data subjectsStatutory72h19.058.0−5.0h3.5hmissed
Cyber incident determined material to investorsStatutory96h0.024.072.0h6.0hfits
Same breach, residents of a 30-day stateStatutory720h19.058.0643.0h3.5hfits
All nine obligations7 statutory or contractual11.6 mean5 short of approvals2 already gone

One row per obligation, not one per event. The same breach appears three times here, on a 72-hour supervisory-authority clock, a 24-hour contractual one and a 30-day state statute, from one awareness timestamp. Collapse them into a single row and the failure hides behind the 643 hours of slack.

Rank by deadline, then by feasibility, never by likelihood. The once-a-decade product hazard with a 24-hour clock and a 31-hour lag is more urgent work than the annual supplier scare with 34 hours to spare.

Window plus lag plus facts must equal the deadline, per row. If it does not, somebody has adjusted an estimate to make the row survivable, which is the one edit this sheet exists to prevent.

RiverEscalation Lag

Escalation Lag

The hours between a clock starting and comms being told. Measured, from real incidents and from drills where the escalation was walked rather than described.

Incident or drillFound byEvidenceLagClockLeft
2024-03 access-unit firmware faultFault ticket opened with two injury reports attachedEngineeringReal incident31.0h24h−7.0h
2024-11 vendor credential compromiseVendor credential confirmed live in our tenantIT securityReal incident19.0h24h5.0h
2025-01 supplier distressAdministration notice appeared on the insolvency registerProcurementReal incident8.0h48h40.0h
2025-06 depot fall, in-patientSite manager told the engineer had been admittedSite managerReal incident6.0h24h18.0h
2025-05 tabletop, field fatalityDrill zero, the supervisor reaching our engineerSite supervisorTabletop1.5h8h6.5h
2025-09 tabletop, officer misconductDrill zero, the enquiry email arrivingPress mailboxTabletop0.5h4h3.5h
Disclosure committee, 8-K determinationThe materiality determination itselfThe committeeStructural0.0h96h96.0h
Distribution across nine scenariosComms discovered none of them11.6 mean, 8.0 medianmax lag 31.0h

Read the Found by column, not the average. Engineering, IT security, procurement and a site manager. In the worst case an engineering team held a firmware fault as a ticket for 31 hours with no idea a 24-hour reporting clock had started when they opened it.

Do not interpolate one scenario’s lag from another. The lag is a property of who discovers the thing. The site manager who knew there was a reporting duty produced 6.0 hours; the ticket queue that did not produced 31.0.

An empty cell is a finding and it argues for a tabletop. A flattering estimate is worse, because it survives review and the register gets built on top of it.

RiverDrill Log

Drill Log

Three tabletops, two hours each, each started at the clock’s real zero and run out of hours. A drill that changed no number on any sheet did not count.

DrillWhat it foundNumber it changedRow
2025-02EU data breachThe plan started the 72-hour clock at the authority’s acknowledgement, not at our own awarenessClock zero corrected on three rows. Article 33 window moved to −5.0hStill broken
2025-05Field fatalityApprovals ran 1.6h against a 1.3h window, because counsel was the only approverDelegated the first statement to the safety lead. Approvals 1.6h to 0.4hClosed
2025-09Officer misconductThe board chair was unreachable for 5.5h. Nobody had tried the out-of-hours numberNamed a standing deputy. Approvals 5.5h to 1.2hClosed
Six hours of tabletopThree findings, all of them arithmeticClosed 2 of the 5 infeasible rows2 need the lag fixed

Split the failures by cause, because the fixes are unrelated. A row short on approvals is closed by a delegation decision somebody makes in a meeting. A row with a negative window is not, and no approval change reaches it.

The two with a negative window are not a comms problem. They need an engineering team and an IT team to know a statutory clock starts when they open a ticket, which is training and tooling and belongs to whoever owns those teams.

Every reach time in the contact tree dates from the 2025-09 drill. Before somebody actually dialled the numbers at 21:15, all six were estimates and the chair’s was wrong by five hours.

  • Clock MethodWhich deadlines you're on, the event that starts each one, and what counts as evidence of delay.
  • Scenario Register sheetOne row per deadline, with the hours left to draft and whether sign-off fits inside them.
  • Notification Requirements sheetEach legal or contract deadline as written, the event that starts it, the recipient and a citation.
  • Escalation Lag sheetMeasured hours between each clock starting and communications being told, from real incidents and drills.
  • Holding StatementsSpace for one pre-approved statement per scenario, under 75 words, with what to leave out and which to write first.
  • Escalation ProcedureHow news travels from whoever finds an incident, alert tiers set by deadline, and ways to shorten sign-off.
  • Spokesperson GuidanceWho speaks for each scenario, their deputy, three lines any spokesperson can use and what never to say.
  • Contact Tree sheetEvery role's out-of-hours route, standing deputy, measured time to reach and what they may approve.
  • Drill Log sheetWhat each tabletop drill found and which number it changed on which sheet.
  • Post-Incident ReviewThe real timeline against the register's predictions, and whether sign-off or late escalation caused any miss.

Common questions

What should a crisis communication plan include?

A crisis communication plan should list the scenarios you could face, the legal, contract and media deadlines each one triggers, and when each clock starts. It also needs the audiences you must reach, pre-approved holding statements, a spokesperson and deputy for each scenario, and a contact tree with out-of-hours numbers. Add an escalation procedure and a log of drills and incident reviews.

What is a holding statement?

A holding statement is a short first public statement issued while the facts are still being confirmed. It confirms what has happened and what you are doing about it, tells affected people what to do, and says when you will update. It leaves out causes and unconfirmed numbers. Write and pre-approve one for each scenario before you need it.

How quickly should a company respond to a crisis?

Aim for a first public statement within an hour, a common rule in crisis communication, but plan around the legal deadlines too. In the US, OSHA requires a work-related death to be reported within 8 hours. Several of those clocks start when the company learns of a problem, not when its communications team does.

Who should be on a crisis communication team?

Keep the core small: a communications lead who owns the plan, a lawyer and an executive who can approve statements. Add the heads of the teams most likely to discover an incident, such as safety, IT security and customer support. Give each person a named deputy and an out-of-hours number, so one unreachable approver can't hold up a statement.

How often should you test a crisis communication plan?

At least once a year, and after any real incident or major change to your people, products or obligations. Start with a 2-hour tabletop drill on the scenario with the least time to spare. Hold it outside office hours, start from the moment the incident would really be found, and call approvers on the numbers your plan lists.

Is this crisis communication plan template free?

Yes. The five Word documents and five CSV sheets download free, with no account needed. The sheets hold a worked example for a made-up company, which you can study and then overwrite with your own scenarios. Edit with AI opens the same template in River instead.

What does Edit with AI do?

It opens the template in a free River workspace. River starts with three questions: what you make or sell, where your customers and their data are, and who you employ. From the answers it lists the deadlines you are likely on. Send your plan and it shows which deadlines the plan covers, starts at the wrong moment or misses, then tests your tightest one.

Sources

  1. eCFR. “29 CFR 1904.39 — Reporting fatalities, hospitalizations, amputations, and losses of an eye as a result of work-related incidents to OSHA.” ecfr.gov (opens in a new tab)

  2. eCFR. “16 CFR 1115.14 — Time computations.” ecfr.gov (opens in a new tab)

  3. European Commission. “Obligations.” commission.europa.eu (opens in a new tab)

  4. SEC. “Cybersecurity Risk Management, Strategy, Governance, and Incident Disclosure: A Small Entity Compliance Guide.” sec.gov (opens in a new tab)

  5. FTC. “Data Breach Response: A Guide for Business.” ftc.gov (opens in a new tab)

  6. Institute for Public Relations. “Crisis Management and Communications.” instituteforpr.org (opens in a new tab)

  7. Ready.gov. “Crisis Communications Plans.” ready.gov (opens in a new tab)

Written by

Chandler Supple

Co-founder & CEO, River

Chandler has spent the last six years working in AI, building tools for millions of users and Fortune 100 companies. In 2025, he co-founded River to build a new kind of operating system for AI agents.

Reviewed by

Brian Joseff, Co-founder, River