B2B cold email with real data: writing a first contact with context

A B2B cold email does not start with the subject line. It starts with the decision to contact that company, at that moment, for that reason.

Without that decision, the team tries to compensate with copy: adding adjectives, inventing personalisation and sending more messages. With context, the work changes. The list gets shorter. The opening line stops being decorative. The request becomes specific. And there are clear criteria for proceeding or stopping.

This guide proposes a process for working with 10 to 25 accounts at a time, with minimal data, human review and metrics that help improve selection — not just the volume sent.

Editorial note: This guide is operational and does not replace legal validation.

Editorial image: Ada, the Makers virtual persona, in a photorealistic composition created for this edition.

Initial go/no-go check: deciding whether the approach can proceed

Before writing, confirm whether the team can proceed at all with that recipient, that purpose and that channel.

In Portugal, Law no. 41/2004 distinguishes subscribers who are natural persons from subscribers who are legal persons under the rules for unsolicited direct marketing communications. For natural persons, Article 13-A provides, as a general rule, for prior express consent; there is a limited exception for existing customers and similar products or services where the other legal conditions are met, including a clear, free and easy way to refuse. Legal persons are subject to a different regime, but this is not unlimited authorisation: objections must be respected, the sender’s identity must not be concealed and there must be a valid means of ending the communications. The law also provides for consultation of the national list of legal persons that have objected.

An address on a business domain does not settle the classification on its own. [email protected] and [email protected] may involve different recipients, data and expectations. A sole trader should not automatically be treated as a commercial company either. If there is uncertainty about who the subscriber is, the nature of the contact, the source of the data or the applicable rule, stop and request legal review.

Legitimate interest is not a free pass to send email either. The European Data Protection Board identifies three cumulative conditions for using this basis to process personal data: a real and lawful legitimate interest, the necessity of the processing, and a balancing assessment that is not overridden by the person’s rights and freedoms. This analysis should be documented and does not replace the specific rules for electronic communications.

Before including an account in the batch, validate:

1. Recipient: is this a legal person, a natural person or a situation still to be confirmed?
2. Source: where was the contact obtained, on what date and for what purpose was it published?
3. Purpose: is there a concrete, proportionate commercial hypothesis, or just a desire to increase volume?
4. Basis and channel rules: has the team documented the applicable analysis and safeguards?
5. Objection: does the address appear on an internal suppression list or an external list the company is required to consult?
6. Identity: does the message clearly identify the sender, company and a valid means of contact?
7. Exit: can the recipient ask for contact to stop simply and free of charge?
8. Owner: is there a team member who reviews replies, removes contacts and decides when to stop?

If any of these answers is missing, the account is not ready for contact.

What this process excludes

This method is not intended to turn a purchased database into a campaign that looks personalised.

It excludes:

• purchased lists without verifiable provenance;
• mass scraping of emails, directories or personal profiles;
• collecting private contacts that were not published for a business purpose;
• indiscriminate enrichment to discover people’s email addresses;
• addresses whose source and date the team cannot demonstrate;
• identical messages sent to hundreds of companies;
• persistence after an objection, removal request or clear sign of disinterest;
• misleading subject lines, such as Re: or Fwd:, where there is no previous conversation.

More data does not make the approach more legitimate. It only increases the risk when the source, purpose or contact is wrong.

Minimum data for a good first contact

There is no need to build a complete profile of every company. You need to keep enough to justify the selection and write a verifiable message.

For each account, record:

• the company’s legal or trading name;
• website and domain;
• sector and location relevant to the campaign;
• public source of the contact and date consulted;
• address type: generic business, named professional or to be confirmed;
• a factual signal observed on the website or another legitimate source;
• the commercial hypothesis linked to that signal;
• the reason for contacting now;
• legal and operational validation status;
• objection or suppression status;
• internal owner and next step.

The signal should be factual. “Lists three locations on its website” is observable. “Must be struggling to manage enquiries” is an inference. The team can test the second as a hypothesis, but should not present it as fact.

Avoid collecting personal data unnecessarily. Date of birth, private telephone number, family profiles, habits outside the professional context or sensitive information do not help you write a responsible first B2B email.

A short workflow for 10 to 25 accounts

A small batch lets you review context, follow replies and correct the hypothesis before scaling the mistake.

1. Define a commercial hypothesis

Start with one sentence:

Companies in [sector] in [territory] with [verifiable signal] may benefit from [concrete outcome], because [connection between signal and offer].

If the sentence is not specific, the list will not be either.

2. Select 10 to 25 accounts

Use the same criteria for every company in the batch:

• the right sector;
• the right territory;
• a compatible size or level of complexity;
• at least one verifiable signal;
• a suitable contact channel;
• no known objection.

The aim is not to find “the best companies in Portugal”. It is to create a unit of work that one person can review and follow through.

3. Classify the strength of the context

Use three simple states:

• A — strong context: there is a current signal, a clear connection to the offer and an appropriate recipient.
• B — partial context: the company appears to be a fit, but the signal, responsible person or timing still needs confirming.
• C — insufficient context: there is only a sector, a location or an isolated contact detail.

Write to A accounts first. Research the B accounts. Do not send to the C accounts.

4. Create a base message for each hypothesis

The base message should reflect the shared problem in that batch. It should not pretend to have been written from scratch for each company.

Then personalise only what is verifiable:

• the observed signal;
• how that signal relates to the hypothesis;
• the final question.

Changing only the company name is not personalisation. It is mail merge.

5. Have a person review before sending

For each message, a person should confirm:

• recipient and address;
• source and nature of the contact;
• the fact used in the opening;
• consistency between fact, hypothesis and offer;
• sender identity;
• a clear way to object;
• the absence of invented or overly personal information.

6. Send in small batches and observe

Do not send all 25 messages at once just because the tool allows it. Start with a small batch and follow replies, bounces, removal requests and signs of discomfort. If the hypothesis is wrong, correct it before the next batch.

7. Record the outcome and decide

Each account needs a status: positive reply, neutral reply, no fit, wrong recipient, objection, bounce, no response or agreed next step.

If contact continues by phone or WhatsApp, use a sales cadence with a clear role and limits for each channel. Do not repeat the same message across three channels.

For more detailed writing examples, see Subject line, opening line and CTA: the anatomy of a B2B cold email.

Anatomy of a cold email with context

A first email can have five parts. This is the summary; the subject line, opening line and CTA guide explores sector-specific examples and anti-patterns in more depth.

1. Factual subject line

The subject should help the recipient understand the topic without tricks.

Examples:

• Sales operations across [Company]'s three locations
• Training providers in the Porto region
• A question about enquiries through your website

Avoid artificial urgency, promises of results, capitals and a false continuation of a conversation.

2. Identification

Say who is writing and which company they represent. Do not conceal the commercial nature of the message.

3. Verifiable observation

Use a detail the recipient can recognise. A location, service line, public page or recent change can provide context. Do not turn that detail into an unsolicited diagnosis.

4. Short hypothesis

Explain the possible connection between the context and what your company does. Use conditional language: “may make sense”, “I wanted to understand whether”, “we are testing this hypothesis”.

5. One request and an exit

Ask a question that can be answered without a 45-minute meeting. Include a simple way to say no and treat that response as an operational instruction, not an objection to overcome.

Annotated example

Imagine a company that helps groups of training centres centralise sales enquiries. The team has selected 18 companies with two or more locations published on their websites and is only contacting business addresses after the initial go/no-go review.

Subject: Sales enquiries across [Company]‘s locations

Hello.

I am Leonor, from Example Company. I saw on [Company]‘s website that you operate in Lisbon and Porto.

We help training teams bring enquiries from several locations into one workflow, with a visible owner and next step. I wanted to understand whether this fragmentation also exists in your team or whether it is already resolved.

Would it make sense to send a one-page summary for you to consider?

If this topic is not relevant, let me know and I will remove this contact from this outreach.

Leonor Silva

Example Company · website · contact

The email works as a hypothesis test:

• it identifies the sender and company;
• it uses a public fact without pretending familiarity;
• it explains the connection between the fact and the offer;
• it asks a short question;
• it allows an explicit objection.

It does not claim the company has a problem. It does not promise savings without evidence. It does not hide the commercial intent.

Deliverability is also an operational control

A responsible message can fail if the infrastructure is poorly prepared.

Gmail’s email sender guidelines require SPF or DKIM authentication for senders to personal Gmail accounts and add requirements for volumes above 5,000 messages a day, including SPF, DKIM, DMARC and one-click unsubscribe for marketing or subscription messages. The same guidelines recommend accurate identities and headers, non-misleading subject lines, controlled spam rates and an easy way to opt out.

Even for a batch of 10 to 25 accounts, check:

• SPF and DKIM are configured on the sending domain;
• DMARC is monitored;
• the From address is recognisable and aligned with the domain;
• a person monitors the reply inbox;
• bounces are handled;
• removal requests are applied before any further sending;
• the internal suppression list is used in every campaign.

Deliverability is not about finding a way into the inbox against the recipient’s wishes. It is about sending with enough identity, security and reputation for a legitimate message to be received — and stopping when the person objects.

Metrics for learning without rewarding volume

Measure the quality of the process first.

Selection quality

• accounts reviewed;
• A, B and C accounts;
• percentage with a verifiable source and signal;
• wrong recipients detected before sending;
• accounts removed by the initial check.

Delivery quality

• messages delivered;
• permanent bounces;
• authentication errors;
• spam complaints;
• objections received.

Conversation quality

• positive replies;
• neutral replies or requests for information;
• “not my area” replies with a referral;
• replies indicating no fit;
• accepted meetings or next steps.

Commercial learning

• signals that led to conversations;
• sectors where the hypothesis was confirmed or rejected;
• recurring objections;
• offers or requests that were unclear;
• changes needed in the list, message or proposal.

Open rate can help with technical diagnosis, but it does not prove interest and can be distorted by privacy mechanisms. Replies, objections, referrals and next steps say more about the quality of the batch.

An objection is not a stage in the cadence

The EDPB explains that an objection to processing for direct marketing does not require a justification and should lead to an immediate end to that processing for that purpose. The CNPD notes that the recipient must be able to refuse communications clearly, free of charge and easily.

In practice:

1. mark the contact as objected as soon as you receive the request;
2. add it to the suppression list used by all tools and teams;
3. do not move it to another sequence with the same purpose;
4. confirm internally that removal has been applied;
5. retain only the minimum necessary to respect the objection and demonstrate the control, according to the organisation’s validated policy.

“Maybe later” is not permission to persist. “Remove me from the list” is not a sales objection to overcome.

Pre-send checklist

• The recipient and nature of the contact have been confirmed.
• The source, date and purpose of the data are recorded.
• The legal analysis and channel rules have been reviewed.
• Objection and suppression lists have been consulted.
• The account belongs to the defined segment and territory.
• There is a factual signal and a proportionate hypothesis.
• The subject line describes the topic without tricks.
• The message identifies the sender and company.
• There is only one clear request.
• The exit is simple, free and will be respected.
• SPF, DKIM, the reply inbox and bounces have been checked.
• A person has reviewed the message and will follow the replies.
• The team has defined when to pause the batch.
• The CRM or operational record can store the outcome and objection.

If the team cannot complete the checklist for 10 accounts, increasing to 100 does not solve the problem.

Next step

The best cold email is not the most persuasive one. It starts from a defensible selection, uses enough data to create context and asks a question the recipient can answer without pressure.

Start small. Review the initial go/no-go check. Choose 10 to 25 accounts. Write one hypothesis per batch. Measure replies and objections. Stop when the context no longer holds.

To adapt the structure to different sectors without copying messages blindly, download the cold email pack by sector.

Official sources

• Diário da República — Law no. 41/2004, consolidated version
• CNPD — Unsolicited promotional communications and cookies
• EDPB — Guidance on legitimate interest
• EDPB — Right to object
• Google — Email sender guidelines