An outbound policy is a short document. It says who you will contact, what you will say, how often, and what happens when someone answers. The person who approves it reads it once, signs, and from then on every send is measured against it.
This guide gives you the sections, a worked example for each, and a note on what the approver should confirm before signing. The examples use bracketed placeholders. Fill them with your own decisions.
Who this is for
This guide is for the person who signs off on outreach: a founder, a head of sales, a revops lead. It is also for whoever drafts the policy for them.
It suits you if any of these is true:
- You are about to start outbound and have no written rules.
- Outreach already runs, but the rules live in someone's head or in a chat thread.
- You use software that sends on your behalf and you want the software held to something you wrote.
On the last point, Outlevo works from a written policy. Nothing is sent outside a policy that a person approved. [[1]](#sources) Policies set who can be contacted, on which channels, how often and with what budget. [[1]](#sources) Every external action is checked against the approved policy before it runs, and recorded in a ledger. [[1]](#sources) The template below is useful with any tool, but it maps closely to those four settings.
One limit on scope. This guide does not tell you which countries to sell into. That is your decision, and the template leaves it to you.
Steps
Work through the sections in order. Each one has a worked example and a line on what the approver confirms. Keep each section short enough to read in a minute.
1. Target accounts and ICP
Write down which companies are in scope and why you believe they fit. State the evidence behind the belief, not only the belief.
Include these lines:
- Company type, size band and industry.
- The countries you sell into. Outlevo asks which countries a customer sells into and never proposes markets, so this line is your own decision. [[2]](#sources)
- The evidence behind the segment: won deals, lost deals, or a stated hypothesis.
- Status of each segment: approved or experimental.
Worked example:
Segment A: [company type], [size band], [industry]
Countries: [your list, written by you]
Evidence: [number] won companies from our CRM export
Status: approved by [name] on [date]
Segment B: [company type], [size band], [industry]
Evidence: hypothesis, no won deals yet
Status: experimental, small test only
The status line matters. AI-proposed ICP segments are labelled experimental until a person approves them. [[2]](#sources) Your policy should keep that distinction after approval: an approved segment and an experimental one should not get the same volume.
If you build the segment from past deals, check the sample size. Win/loss analysis counts companies, not deals, and under 10 won companies is labelled very low confidence. [[2]](#sources) You can import past deals from a CSV or your CRM to get that analysis. [[3]](#sources)
What the approver confirms:
- The countries listed are the ones your company has chosen to sell into.
- Each segment says whether it rests on deals or on a hypothesis.
- Segments with a small deal sample are marked as such, not presented as proven.
2. Roles to contact
Name the job titles that are in scope and the ones that are out. Say how many people per company you will contact.
Seniority changes reply rates, so state why you chose your titles. One study of 2025 campaigns by Belkins gives reply rates by buyer seniority: founders and owners 0.57%, C-level 0.42%, VPs 0.32%, against an average of 0.45% over all emails sent. [[4]](#sources) That is one dataset, not a rule. It is a reason to write down your choice instead of defaulting to the most senior title. If you sell to founders, the cold email to founders guide covers that buyer in more detail: Cold email to founders.
Worked example:
In scope: [title family 1], [title family 2]
Out of scope: [titles you will not contact]
Contacts per company: [number], never two at once
If no in-scope contact is found: skip the company, do not guess
Say where contact data may come from. If you import contacts from HubSpot, that import is optional and only happens after an admin turns it on with a stated purpose. [[5]](#sources) Put that stated purpose in the policy so the two match.
What the approver confirms:
- Titles in scope match the buyer you described in section 1.
- The "no contact found" rule is skip, not guess.
- The source of contact data is named, and so is its purpose.
3. Exclusion rules
Skipping this section is a common way to end up emailing someone you should have left alone. List who must never be contacted, regardless of fit.
Typical categories:
- Current customers.
- Companies with an open deal.
- Anyone who opted out or unsubscribed, on any channel.
- Competitors and partners.
- Named individuals or companies your team asks you to leave alone.
- Domains that are your own.
Worked example:
Never contact:
- companies at lifecycle stage [customer stage name]
- companies with an open deal in [pipeline name]
- any address on the suppression list
- anyone who replied "stop", "unsubscribe" or similar, on any channel
- [named accounts]
Suppression list owner: [name]
Review: [frequency]
If your CRM is HubSpot, Outlevo reads company name, domain, owner and lifecycle stage, plus deal pipeline and stage. [[5]](#sources) Use those field names in the rule. The public API has "suppressions" and "policies" resources, so exclusion lists and policy settings can also be managed in code. [[6]](#sources)
Add a recipient check for one-off human emails. The Gmail extension has a recipient guard that warns before you email a contact Outlevo should not reach, and it never sends from Gmail. [[7]](#sources) Your policy can require people to use it.
Add a short legal note, written by your own counsel or compliance lead, not by this template. Three public statements are worth knowing:
- The FTC says of CAN-SPAM: "The law makes no exception for business-to-business email." [[8]](#sources) A B2B label does not exempt you from the rules.
- The UK ICO says: "You can email or text any corporate body (a company, Scottish partnership, limited liability partnership or government body)." [[9]](#sources) That is about corporate bodies, not every individual at a business.
- GDPR Recital 47 says direct marketing processing "may be regarded as carried out for a legitimate interest." [[10]](#sources) "May be" is not "is". Your reasoning should be written down and approved.
Which rules apply depends on the markets you chose in section 1. Have your counsel confirm that, and record who confirmed it and when.
What the approver confirms:
- Every exclusion category has an owner and a place where the list lives.
- Opt-outs apply across channels, not per campaign.
- A named person has reviewed the legal note for the markets in section 1.
4. Message claims allowed
Decide what an email is allowed to say about your company and about the recipient. In our view, a claim in an email is a promise, so this section deserves care.
Split claims into three groups:
- Claims about you. Allowed only if you can show proof on request.
- Claims about the recipient's company. Allowed only if there is a source and a date, the same standard Outlevo applies to the observations on a Readiness Card. [[11]](#sources)
- Claims you will not make. Write these out.
Worked example:
Allowed about us: [list of approved statements, each with its proof location]
Allowed about the prospect: only facts with a source URL and a date
Phrase unverified ideas as questions, not statements
Not allowed: invented customer names, results without proof,
comparisons with named competitors, urgency we cannot back up
Quoted numbers must come from our own approved facts file
You can hold a tool to the same standard. Outlevo's Readiness Card needs at least three observations with a source and a date. [[11]](#sources) Guesses are labelled as hypotheses, never presented as findings. [[11]](#sources) Copy that pattern: if an email states something about the prospect, a person can find where it came from.
What the approver confirms:
- Every approved claim about you has proof the approver has seen.
- The "not allowed" list is specific enough to check a draft against.
- Prospect facts need a source and a date, and guesses read as guesses.
5. Tone
Tone is easy to leave vague. Pin it down with a few rules and one short sample the approver can accept or reject.
Worked example:
Voice: plain, direct, first person, no exclamation marks
Length: [maximum words] for a first email
One ask per message
No fake familiarity ("hope you're well"), no fake urgency
Do not pretend a message is personal if it is templated
Sign off with the sender's real name and role
Add the channel limit where one exists. A LinkedIn connection invitation note is at most 300 characters. [[12]](#sources) Your tone rules have to fit inside that.
Include one approved sample first email and one approved sample follow-up. Without samples, "plain and direct" means different things to different readers.
What the approver confirms:
- The tone rules can be checked on a draft, not just agreed with.
- A sample message is attached and the approver has read it.
- Channel limits are respected.
6. Sending limits
This section sets how much goes out, from where, and with what budget. Write it as numbers your team chooses. Do not copy a number from a blog post.
Worked example:
Channels allowed: [email / LinkedIn / none other]
Emails per mailbox per day: [number]
Follow-ups per contact: [number], spaced [days] apart
LinkedIn: [your rule for invitations and follow-ups]
New mailboxes: send nothing until warm
Monthly credit ceiling: [number]
Stop condition: pause the campaign if [bounce or complaint threshold you set]
Preconditions. With Smartlead, every mailbox must be connected and warmed up before a campaign sends, and its domain must have SPF, DKIM and DMARC checked in the last 7 days. [[13]](#sources) With managed mailboxes, a new mailbox sends nothing until it is warm: at least 14 days, usually about 3 weeks. [[14]](#sources) Plan your start date around that.
Outside rules. Google says "email senders who send more than 5,000 messages per day to Gmail accounts must meet the requirements in this section. Set up SPF and DKIM email authentication for your domain." [[15]](#sources) Your policy may sit well below that volume, but the authentication step is worth writing down either way. Whether to send from a separate domain is a decision for the approver. See Secondary vs primary domain for cold email.
Budget. Metered API calls quote their credit cost first and run only when confirmed. [[16]](#sources) Sending, follow-ups and the mailboxes your plan includes use no credits. [[13]](#sources) Research steps do cost credits: contact discovery costs 4 credits, email verification costs 1 credit, qualification costs 3 credits and deep research costs 30 credits. [[17]](#sources) Write the monthly ceiling the approver is willing to spend, not the plan maximum.
Benchmarks. Do not write a target reply rate unless you can say where it comes from. Smartlead's State of Cold Email 2026 puts the median cold email reply rate at 0.74% per contact, the top quarter at 1.43% and the top tenth at 2.63%. [[18]](#sources) The bases differ from the Belkins figures in section 2, which are over all emails sent. [[4]](#sources) Do not set the two sets of numbers side by side. Either is a reference point for judging results after the fact, not a promise to put in the policy.
To compare tool prices before you fill in this section, see email verification tools and email deliverability tools.
What the approver confirms:
- Every limit is a number someone chose, with an owner.
- Mailboxes are warmed and domain authentication is checked before the start date.
- There is a stop condition that someone can act on.
- The credit ceiling is written down.
7. Reply handling
Decide in advance what each kind of reply triggers, and who acts. This is where policy turns into customer experience.
Worked example:
Interested: route to [owner]. Answer within [hours].
Not now: pause the lead for [days], then stop.
Not interested: stop all channels, add to suppression list.
Unsubscribe or "remove me": stop all channels, suppress, no further message.
Bounce: stop, mark address invalid, do not retry.
Out of office: pause until return date.
Unclear: send to a person, do not auto-reply.
Autopilot for interested replies: [yes / no]
The autopilot line deserves a deliberate choice. For interested replies with a Calendly link, on autopilot the answer goes at once. [[19]](#sources) Otherwise it waits in Replies for a person to send. [[19]](#sources) An approver who wants to read the first answers before anything goes out should write "no autopilot" and revisit it later.
What the approver confirms:
- Every reply type has an action, an owner and a time limit.
- Unsubscribes and opt-outs stop every channel.
- The autopilot choice is explicit.
- Unclear replies go to a person.
8. Escalation
Escalation answers two questions: who hears about it, and who can stop it.
Worked example:
Hot lead: alert [owner of the account]; they act within [hours]
Problem that stops sending: alert [name] and [backup]
Credit alert: alert [name]
Complaint or legal request: stop the campaign, then [name]
Who can pause a campaign: [names]
Who can pause a channel: [names]
Who can pause the whole workspace: [names]
Use named people, not "the team". In HubSpot, hot leads go to the company owner, so each account has a named person. [[5]](#sources) Telegram alerts cover hot leads, approvals, problems that need you and credit alerts. [[20]](#sources)
Then name who holds the stop button. Kill switches pause a campaign, a channel or the whole workspace at once. [[1]](#sources) The policy should say who may press each one, and that using it needs no further approval.
What the approver confirms:
- Every alert type has a named person and a backup.
- At least two people can pause the workspace.
- Pausing does not need a meeting.
9. Sign-off block
End the document with the part that makes it a policy and not a draft.
Policy version: [number]
Approved by: [name, role]
Date approved: [date]
Next review: [date]
Changes since last version: [list]
Review it on a date, and also whenever you add a channel, a segment or a country. Re-approve after each change. Whatever tool sends, an outdated policy is the same as no policy.
What to check
Use this as the approver's last pass before signing.
- Scope: countries are ones your company chose; segments are marked approved or experimental; small deal samples are labelled.
- People and data: titles in and out of scope are listed; contact data sources and purpose are named; exclusions have owners; counsel has confirmed the legal note.
- Message: each claim about you has proof; prospect facts have a source and a date; sample emails are attached.
- Sending: limits are numbers someone chose; mailboxes are warm and authentication is checked; a stop condition and a credit ceiling are written down.
- Replies and escalation: every reply type has an action, owner and time limit; autopilot is a deliberate yes or no; alerts reach named people and at least two can pause everything.
- Record: version, approver, date and review date are filled in.
If any box stays empty, do not approve. A blank line in the policy becomes a decision someone makes alone, later, under pressure.
Next step
Start with section 1, because the others depend on it. If you have no deal history to draw on, run the free ICP assessment. It reads up to 5 public pages of your website and returns up to 5 experimental segment hypotheses. [[21]](#sources) Treat those as drafts for you to approve or reject. Then fill in sections 2 to 8 in one sitting, and give the draft to the person who will sign.
To see how a written policy is enforced once it is approved, read How Outlevo works. To manage policies and suppressions in code, start at the developer hub.
Sources
- How Outlevo works
- ICP from evidence
- Release 2026-10-19
- Belkins, cold email response rates (all 2025 campaigns; page updated 26 June 2026)
- HubSpot integration
- Outlevo API reference
- Gmail extension integration
- FTC: CAN-SPAM Act compliance guide for business
- ICO: Electronic mail marketing (PECR)
- GDPR Recital 47: legitimate interests of the controller
- Readiness Cards
- LinkedIn integration
- Smartlead integration
- Managed mailboxes integration
- Google: Email sender guidelines
- Developer hub
- Outlevo pricing
- Smartlead, The State of Cold Email 2026: reply rate per contact (top quartile; January to June 2026, more than 850 million emails)
- Calendly integration
- Telegram integration
- Free ICP assessment