Email Deliverability: The Complete Guide to Getting Cold Email Out of Spam (Domains, Warm-Up, Monitoring)

Key takeaways

Email deliverability is the share of your emails that actually reach the inbox, rather than landing in spam or getting blocked along the way. In outbound prospecting, once you go past a few hundred emails a month, it stops being a setting and becomes infrastructure you have to size and operate. To improve it for the long run, start from your pipeline target to calculate sending volume, check that your clean addressable market can absorb it, look at which mail providers your prospects use (Gmail, Outlook, security gateways), then work out the number of inboxes and domains you need, with a 20 to 30% buffer. Next comes a three-week warm-up before the first cold email, a ramp-up that keeps total volume constant, regular deliverability tests, and monitoring at three speeds: bounces daily, reply rates weekly, placement tests monthly. A domain that starts to weaken is rested and replaced by an inbox that is already warmed up, before the drop reaches your pipeline.

This guide is for teams that have outgrown their first sending domain: sales leaders, growth and RevOps teams who watch campaigns perform for one quarter and then collapse the next for no obvious reason. If you are sending your first 500 emails, start with the technical basics of deliverability: SPF, DKIM, DMARC and a first warm-up. This guide covers what happens when volume grows tenfold.

Why deliverability breaks as volume grows

The story is almost always the same. A team starts with two domains and six inboxes, sends a few hundred emails a month and gets solid reply rates. The board sets a more ambitious pipeline target: inboxes get added as needed, a second tool gets plugged in, a new region gets launched. Six months later, the team is sending several thousand emails a month across fifteen or so domains bought at different times, configured by different people, and nobody really knows which ones are working.

At that point, mistakes no longer stay contained. A sequence flagged as spam doesn't just hurt one inbox: it damages the domain's reputation, and therefore the five inboxes using it, and therefore every campaign running through them. And the symptom shows up late. Emails keep going out, the tool says "sent," but a growing share lands in spam. The first visible signal is a drop in replies, which people readily blame on the message or the targeting. By the time the diagnosis comes in, the quarter's pipeline has already taken a hit.

The teams that clear this hurdle without damage are not the ones that found the right tool. They are the ones that treat sending like an operations system: they size before they build, measure continuously and document every change. The loop fits in five words: test, measure, adjust, document, repeat. Without measurement, you react to symptoms. Without documentation, you lose the context that explains why a domain dropped off in March.

The six levers mailbox providers evaluate

Gmail, Outlook and security gateways don't judge an email on a single criterion. They cross-check six families of signals, and any one of them can cancel out the work done on the others.

1. Technical setup. SPF, DKIM and DMARC authentication, consistent DNS records, a clean tracking domain. This is the eligibility requirement: without it, nothing else matters.
2. Sender reputation. Domain and IP history, domain age, past sending behavior.
3. Data quality. Bounce rate, invalid addresses, "catch-all" addresses that accept everything and reject later, outdated contacts.
4. Content. Number of links, images, attachments, aggressive sales language, identical structure repeated across thousands of sends.
5. Volume and cadence. Consistent sending, gradual ramp-up, no sudden spikes.
6. Recipient behavior. Replies, deletions without reading, spam reports, messages moved out of spam.

A flawless technical setup won't rescue a two-year-old contact database. A great message won't get through on a domain whose reputation is already damaged. Every decision in this guide, from choosing domains to rotating them, ties back to one or more of these six levers.

Step 1: start from pipeline, not from the number of domains

Most sending infrastructures are sized by gut feel: "let's get ten domains, three inboxes per domain, and see how it goes." The result is either oversized, with licenses paid for nothing, or undersized, with too much volume concentrated on too few inboxes and deliverability that collapses. Either way, the link between the sales goal, the market and sending capacity is missing.

The right starting point is a capacity question: how many emails do you need to send to hit the pipeline target? You work the chain backward: closed deals, then meetings, then emails sent, then contacts to reach.

Take a French SaaS vendor aiming for €2.4M in new annual revenue from outbound prospecting.

On a monthly basis, that comes to roughly 20,800 emails, or close to 1,000 sends per business day. That is the demand side. The rates used here are assumptions: replace them with your own, measured over the last six months. An email-to-meeting rate of 0.4% is common for a well-defined mid-market target; it can fall below 0.2% for enterprise accounts or exceed 1% in a tightly targeted niche.

Step 2: check that volume against your clean addressable market

Second question: can the market handle this volume without burning through contacts or damaging your reputation? For planning purposes, the addressable market is not the theoretical list of every company that could buy. It is the subset you can reach with clean, verified data that matches your ideal customer profile.

You get there with four successive filters:

1. Ideal customer profile. Size, industry, geography, tech stack, buying signals: you go from "every company" to "the companies that fit."
2. Enrichment. You look for decision-makers and their emails. Not every company has findable contacts: plan for 60 to 80% coverage depending on the market, often lower for French SMBs than for tech scale-ups.
3. Verification. A validation tool (ZeroBounce, NeverBounce, MillionVerifier, etc.) filters out invalid and disposable addresses and unqualified catch-alls. This step often removes 10 to 20% of the list.
4. Exclusions. You remove unsubscribes, addresses that have already bounced, customers and open opportunities in the CRM.

You need 62,500 contacts a year, and the clean market holds 32,100. So you will have to reach each contact about twice a year, with two sequences spaced at least five to six months apart and using different angles. That is sustainable. If the math had come out at five or six touches a year, the problem would not have been technical: the target would have exceeded what the market can absorb, and you would have needed to widen the target, add other channels or revisit the goal.

Three mistakes come up often at this stage:

• Sizing for current volume. If the target doubles next year, volume follows. Plan six months out, not six weeks: a domain takes several weeks to become operational.
• Forgetting seasonality. If sending peaks in September-October and in January, size for the peak, not the average.
• Thinking in sending capacity instead of pipeline. The question isn't "how much can we send?" but "how much do we need to send, and can the market take it?"

Step 3: map your prospects' mail providers

The question that always comes up: should you send from Google Workspace, Microsoft 365 or a mix of both? The answer doesn't depend on the tool your team knows best, but on where your prospects receive their email. Each provider filters differently, and the same email can reach the inbox of a Gmail user and be blocked by a corporate security gateway.

Read your target's MX records

A domain's MX records tell you which server receives its email. By querying them for every domain in your addressable market, you get the real breakdown of your prospects. Most verification tools export this data; otherwise, a simple DNS lookup does the job:

dig +short MX example-prospect.com
10 example-prospect-com.mail.protection.outlook.com.

A security gateway (often called a SEG, for Secure Email Gateway) sits in front of the inbox: it analyzes links, headers, IP addresses and reputation to catch phishing, and can reject an email before Google or Microsoft even evaluates it. Gateways are common at enterprise accounts and in finance, healthcare and the public sector, including when the mailboxes are hosted by Google or Microsoft.

For a SaaS vendor targeting the mid-market, a typical breakdown looks like this, but yours may be very different. Traditional French targets (manufacturing, mid-size companies, local government) are often mostly on Microsoft 365, unlike tech scale-ups.

Gmail and Outlook don't filter the same way

Gmail relies heavily on machine learning and recipient behavior: replies, reads, deletions without opening, spam reports. Google asks senders to keep their spam complaint rate below 0.1% and never reach 0.3%, the threshold beyond which deliverability degrades sharply. Reputation builds quickly when engagement is good, and is lost just as quickly.

Outlook and Microsoft 365 rely on Exchange Online Protection, which focuses more on historical reputation and sending patterns. Every message gets a spam confidence level (SCL): from 5 to 6, it goes to the junk folder; from 7 to 9, it is treated as spam with high confidence. Microsoft rewards consistency: domain age, flawless authentication, a steady sending pace.

For a target that is mostly on Gmail, the effort goes into engagement: short, personalized messages that make people want to reply. For a target that is mostly on Outlook, it goes into consistency: domains bought early, volume spread evenly, perfect authentication.

Domain age carries a lot of weight at Microsoft

Placement tests show it again and again: domains less than six months old struggle to reach the Outlook inbox, even after a proper warm-up and with clean sending behavior. Past that point, with a consistent history, placement becomes reliable. If your target is mostly on Microsoft, buy your domains several months before you need them and let them age at low volume.

Launch on Google first, open up Outlook later

When you can't wait for domains to age, a phased approach works well: after two to three weeks of warm-up, launch campaigns only to prospects on Google Workspace, which respond faster to good engagement signals. Pipeline starts flowing, domain reputation builds, and contacts on Microsoft 365 or behind a gateway are brought in gradually once the domains are established. You gain several weeks without exposing your domains to the strictest filters on day one.

Step 4: size your inboxes and domains

The goal isn't to mirror your target's mail provider breakdown exactly, but to build infrastructure that performs well within it.

How many emails per inbox per day

There is no universal number. Some practitioners never go above 20 sends per inbox per day, others send 40 without trouble. The right cap depends on domain age, data quality, budget and risk tolerance.

For our 1,000 emails per business day, the pace you choose completely changes the size of the infrastructure:

Our recommendation: if volume is going to grow, start conservative or moderate. The spare capacity becomes a testing ground for pushing some domains further without putting the core of the setup at risk. Keep the aggressive pace for isolated tests, never for the campaigns that carry your pipeline. From here on, we use the moderate pace of 20 emails per inbox per day.

The math for our example

• Daily volume: 20,800 emails ÷ 21 business days ≈ 1,000 emails per day.
• Active inboxes: 1,000 ÷ 20 = 50 inboxes.
• Active domains: 50 ÷ 5 = 10 domains. Vary between 3 and 5 inboxes per domain for a less mechanical footprint.

Splitting between Google and Microsoft

Say your MX mapping shows 60% Google, 28% Microsoft and 12% other providers or gateways. You don't copy that split: you start with roughly 80% Google inboxes and 20% Microsoft inboxes. Google inboxes provide the stability you need to reach most of the target; the Microsoft share, slower to build and more demanding, starts out as a test of that channel. If it performs, you scale it up. If not, you haven't overinvested. This starting point isn't set in stone: revisit it quarter after quarter based on actual results.

Another useful rule: don't put all your domains in a single Google Workspace or Microsoft 365 account. Spread them in groups of 3 to 5 domains per account. If an account gets suspended or flagged, which is rare but does happen, you lose three domains instead of ten. This separation also lets you organize the infrastructure by region, use case or performance tier.

Add a 20 to 30% buffer

Infrastructure sized too tightly breaks at the first warning sign. The buffer serves four purposes:

• Rotation: rest a weakening domain without reducing volume.
• Testing: try a new offer or a new sequence without touching the domains that carry your pipeline.
• Backup: keep inboxes already warmed up, ready to take over immediately.
• Growth: absorb the volume increase planned for the next six months.

Put a number on the budget

Attaching a cost to the setup turns scaling decisions into an ROI calculation. Rough figures, based on current list prices (check at the time of purchase, excluding taxes):

• Google Workspace licenses (entry-level plan, no commitment): about €8 per inbox per month, or about €400 a month for 50 inboxes.
• Microsoft 365 licenses (Business Basic plan): about €6 per inbox per month, or about €90 a month for 15 inboxes.
• Domains: about €10 to €15 per domain per year, or less than €200 a year for 13 domains.

Total: around €6,000 a year, not counting the sending tool, the warm-up tool, data verification and people's time. Against the €2.4M revenue target, it's a rounding error. What's expensive is a quarter of campaigns landing in spam.

Step 5: build the infrastructure

The plan is set: 13 domains, 65 inboxes, roughly 80% Google and 20% Microsoft. Now you need to decide how to build it. There are three routes, each with different trade-offs between speed, control and required expertise.

Route 1: automated setup tools

Some platforms buy the domains, configure DNS, create the inboxes and start the warm-up from a simple form. Within a few hours, everything is ready, no technical skills required.

The downside: less visibility into how things are built, vendor lock-in that makes any migration harder, and shared risk. If the vendor suffers an incident or gets blacklisted, hundreds of its customers' domains can be hit overnight, and that has happened before.

• Good for: teams validating their outbound channel or testing a new market before investing.
• Avoid when: outbound is your main source of pipeline, or you need to understand and control every component.

The right way to use them: as an accelerator, not the foundation. Keep part of your infrastructure independent so a vendor incident doesn't wipe you out.

Route 2: building in-house

You buy the domains yourself, open the Google Workspace and Microsoft 365 accounts, configure SPF, DKIM and DMARC, and connect each inbox to the sending tool. You own everything, you can audit every setting, fix a problem directly and switch tools without rebuilding anything.

The cost is time: setting up 65 inboxes properly takes several days, and operations (warm-up, rotation, documentation) require ongoing effort. It's the most robust route for a team with in-house expertise that treats outbound as a strategic channel. On one condition: document every component (DNS records, inbox assignments, warm-up schedule, rotation rules). That documentation becomes the blueprint for your next domains and your safety net when someone leaves the team.

Route 3: have a specialist build it

The middle ground: a specialized provider buys and configures everything, creates the inboxes, starts the warm-up, then hands the infrastructure over to you. You keep ownership of the domains and accounts. The benefit is avoiding rookie mistakes: a good provider has built this kind of setup dozens of times and puts naming conventions, allocation logic, buffer, monitoring and documentation in place from the start.

The downside: higher upfront cost, quality that varies widely from one provider to the next, and the need for a precise brief. The right provider is the one who leaves you with a structure your team can run on its own.

Many teams end up with a hybrid model: a core built in-house or by a provider, supplemented by automated tools for testing, and several vendors so they don't depend on any single one. Each client or business unit gets its own isolated infrastructure: an overly aggressive campaign on one side doesn't contaminate the others.

Domains: naming and redirects

Never send cold email from your primary domain: it's the one that carries your invoices, your support and your transactional emails. Sending domains are recognizable variants (getbrand.com, brand-team.com, trybrand.com, etc.), avoiding multiple hyphens and exotic extensions associated with spam. Each sending domain redirects to the main website, has a working MX record and, ideally, a custom tracking domain if you track clicks.

Shared IPs or dedicated IPs

Google Workspace and Microsoft 365 send from IPs shared across many senders: your reputation is partly pooled, but those IPs are maintained by companies with every incentive to keep them clean. SMTP servers with a dedicated IP give you full control and isolate your reputation, at the cost of higher spend and real technical expertise; a new IP also has to be warmed up. Start on shared IPs, and only consider dedicated IPs at very high volume or for a specific isolation need.

Authentication: the eligibility requirement

SPF, DKIM and DMARC don't directly improve deliverability: they make the sender eligible. Without them, the major mail providers reject you or send you straight to spam. The detailed settings are covered in our guide to the technical basics; here are the points that cause trouble at scale.

• SPF declares the servers authorized to send on behalf of the domain. Example for a domain on Google: v=spf1 include:_spf.google.com ~all. Watch out for the 10 DNS lookup limit set by the standard. Every tool you add with another include: uses some up; past the limit, SPF fails silently.
• DKIM signs every email to prove it really comes from you and hasn't been altered. Each sending service generates its own key; use 2048-bit keys.
• DMARC tells receivers what to do with an email that fails SPF and DKIM: p=none (monitor), p=quarantine (spam folder) or p=reject (reject). Example: v=DMARC1; p=quarantine; rua=mailto:[email protected]. Start at none with reporting, confirm that all your legitimate mail passes, then tighten. A reject policy set too early will block your own emails if a service is misconfigured.
• Alignment: the domain visible in the From address must match the domain validated by SPF or DKIM. That's what DMARC checks, and it's the most common mistake when a tool sends "on behalf of" your domain.

The rules have gotten stricter. Since February 2024, Google and Yahoo require senders of more than 5,000 messages a day to their consumer accounts to have SPF and DKIM authentication, a DMARC record, an aligned domain, one-click unsubscribe for marketing emails and a complaint rate below 0.3%. Microsoft has enforced comparable requirements since May 2025 for high-volume senders to Outlook.com, Hotmail and Live. These rules primarily target bulk sending to consumer mailboxes, but corporate filters take their cue from them: cold email infrastructure should comply regardless of volume.

Step 6: warm-up, before the first cold email

A new domain has no history. Gmail and Outlook don't know whether you're a legitimate business or a spammer, and they watch your first sends closely. The first few hundred emails shape their first impression: steady sending, positive engagement and few bounces are reassuring. Complaints, bounces and volume spikes are a red flag.

Warm-up means building those positive signals before you send a single cold email: the inbox sends and receives emails that get opened, read and replied to, and pulled out of spam if they land there. Mail providers look at consistency, engagement, complaints and bounces. Warm-up works on all four at once.

Peer network or seed list

Two families of tools exist side by side:

• Peer-network warm-up. Your inbox joins a network of other users: it sends them emails, they send you emails, and every email is automatically opened, replied to and marked "not spam." It's cheap and often built into the sending tool. The limits: network quality varies (other cold senders, abandoned inboxes that generate bounces), the activity can look artificial to mail providers, and these tools struggle to rescue a domain that has already been penalized.
• Seed-list warm-up. Your emails go to a set of real inboxes, spread across mail providers and maintained by the vendor, where they get opened, replied to and pulled out of spam. The network is cleaner, with fewer bounces, and the approach is far more effective at recovering a damaged domain. It is, however, much more expensive.

The rule of thumb: a peer network to build the reputation of a new, healthy domain, a seed list to repair a domain whose reputation has slipped.

The ramp-up schedule

Plan on two to three weeks of warm-up alone, gradually ramping up to 30 to 40 emails per inbox per day. Once engagement is stable, introduce a few cold emails, five per inbox at most, then increase the cold share each week while reducing warm-up by the same amount. The core principle: total daily volume stays steady, only its mix changes. Even at cruising speed, keep 10 to 15 warm-up emails per inbox per day to sustain engagement signals.

The reserve stays warm at all times

Reserve inboxes are useless if they're cold. In our example, the 15 reserve inboxes stay in continuous warm-up. When an active domain drops off, you shift its volume to a reserve inbox the same day, instead of waiting three weeks for a new domain to be ready. You're paying licenses for capacity that isn't sending at full speed: it's insurance against a pipeline outage.

Monitor the warm-up itself

Warm-up isn't something you launch and forget. Three things to check:

• Bounces. A high or rising bounce rate during warm-up often points to inactive inboxes in the peer network.
• Disconnections. Warm-up tools regularly lose their connection to an inbox without any alert. Warm-up then stops silently. Across hundreds of inboxes, that justifies an automatic daily alert, or even automated reconnection: what takes half an hour to check by hand gets fixed in seconds.
• Placement rate. The warm-up tool shows the share of its emails that reached the inbox. A sudden drop, or falling below your threshold, means pausing the inbox and finding the cause.

How to test your email deliverability

Testing deliverability means checking three separate things: that the technical setup is correct, that domain reputation is healthy, and that emails actually reach the inbox at your prospects' mail providers. No single tool covers all three.

A good mail-tester score doesn't prove your emails reach the inbox: it only proves the email is technically clean. A new, perfectly configured domain can score 10/10 and still land in spam at Outlook for lack of history. Conversely, a single placement test has no statistical value: what matters is the month-over-month trend, on the same seed inboxes. The only metric that captures everything, setup, reputation, content and targeting, is still the reply rate.

Step 7: monitor like an operations team

Healthy infrastructure doesn't stay healthy on its own. The goal of monitoring isn't to pile up dashboards, but to spot small signals early (replies down, bounces up, placement slipping) and tell normal variation apart from the real start of a problem.

Organize the infrastructure so it's readable

Visibility depends first on structure. Two setups work well beyond a few hundred inboxes.

• By pod. The infrastructure is split into pods of about fifty inboxes and ten or so domains, each pod assigned to specific campaigns. When performance drops, you know immediately which campaigns and which ten domains are affected, then drill down to the domain level if needed. Testing a new offer, isolating a risky experiment: you dedicate a pod to it. You can rest one domain within a pod, or rotate an entire pod.
• By country or region. Each market has its own domains and inboxes. A bounce spike in France doesn't force you to check your German or US domains. Each market can be optimized (send times, language, pace) without affecting the others, and opening a new country doesn't put at risk the infrastructure already producing pipeline elsewhere.

Define what normal looks like

Before you can spot an anomaly, you need a baseline. Your first two months of real sending are what establish it.

• Bounce rate: consistently below 2%. Occasional spikes to 2 or 3% happen (a batch of less clean data), but a level that settles above 2% signals a data or infrastructure problem.
• Reply rate: varies widely by market and offer, often 2 to 4% for a mid-market SaaS target, 0.5 to 1.5% for enterprise accounts. What matters is stability: going from 3% to 2% is a one-third drop, and worth investigating.
• Inbox placement: the vast majority of emails should land in the inbox. A visible drop means you're losing visibility.
• Sending volume: steady from one day to the next. No sudden tripling, no unexplained drop by half.

Noise or signal

The difference comes down to duration and scope. One bad day on one domain is noise: you note it and don't react. A lasting trend across several domains or several campaigns is a signal: you investigate.

Monitoring cadence

When bounces rise, separate hard bounces (invalid address) from soft bounces (full mailbox, server unavailable), then look for the common thread: one domain, one campaign, one data source? At 2.5%, first check that catch-alls were actually filtered out. Above 5% on a sustained basis, or at the first spike to 10%, pause the domain while you figure it out.

When the warm-up score drops, stop cold sending on the affected domains, let warm-up run on its own, and look for what changed recently: content, a volume spike, a DNS change. Once the trend stabilizes, reintroduce cold sending gradually.

Monthly placement tests (GlockApps, for example) shouldn't be read like a grade. Design each test around the question you're asking: if replies are dropping on enterprise accounts, test Outlook and security gateways specifically; if Gmail is the concern, focus on it. The test is there to validate a hypothesis, not to award a badge.

Automate alerts

Beyond a few hundred inboxes, manual monitoring no longer holds up. Teams managing thousands of domains centralize all their metrics (bounces, warm-up scores, replies, per-domain performance) in a data warehouse, with automatic alerts as soon as a threshold is crossed. Problems get caught the same day, before pipeline moves. The team spends its time investigating and acting instead of collecting numbers, and has the history to answer with facts when sales leadership gets worried.

An investigation example: the gateway that blocks everything

A common case: daily monitoring reveals a bounce spike concentrated in a single pod. The data shows the what, not the why. Digging deeper, you find that the rejections all come from the same security gateway, used by several accounts in the segment, which blocks emails before they reach the inbox.

In the short term, contacts behind that gateway are moved to LinkedIn and phone: you stop damaging your reputation without losing touch with those accounts. In the medium term, you run small tests to understand what gets through: plain text, no tracking pixel, different domain profiles. You scale what works, drop the rest, and document it for future campaigns. Metrics detect, investigation explains, multichannel outreach buys time, and tests find the fix.

Step 8: rest, recover or retire a domain

Even with perfect monitoring, some domains eventually decline. Rotation and recovery are the system's safety valves.

When to take a closer look at a domain

• Reply rate 20% below average for several weeks.
• Bounce rate above 5%.
• Placement declining from one test to the next.

The rest and rotation procedure

1. Immediately stop cold sending on the affected domain.
2. Find the cause: data, content, volume, DNS setup.
3. Activate an already warmed-up reserve domain to pick up the capacity.
4. Move the domain to light warm-up, 10 to 15 emails per day, for two to four weeks, using a seed list if the reputation is badly damaged.
5. Measure placement every week.
6. Decide: if placement has recovered after four weeks, reintroduce cold sending at low volume; if not, retire the domain.

When to retire a domain for good

Not every domain can be recovered. You retire a domain when it appears on major blacklists despite delisting requests, when several recovery attempts (seed lists, extended rest) have gone nowhere, or when fixing it costs more than setting up a new one. At that point, you stop all sending, disconnect it from every tool, document the cause so the mistake isn't repeated, and never use it for email again: it's burned.

The habits that protect your infrastructure

Good infrastructure doesn't make up for bad practices. Well-configured domains, flawless DNS, a clean warm-up: none of it matters if the way you handle data, campaigns and content sends your emails to spam.

Data hygiene

• Verify every list before import. No exceptions. Catch-alls, which accept every address and reject later, should be excluded unless your verification tool scores them. If you contact them anyway, do it from reserve domains to contain the damage.
• Refresh your data. Contacts go stale: job changes, deactivated mailboxes, companies that shut down. A database that has sat idle for a year needs to be re-enriched and re-verified before any send.
• Maintain three suppression lists. Unsubscribes, bounces, clicks on the unsubscribe link. Every import is checked against all three. Emailing someone who has already said no is the surest way to trigger a complaint.

Campaign management

• The unsubscribe link cuts both ways. At moderate volume, it protects you: a prospect who can unsubscribe in one click doesn't report you as spam. At very high volume, it can get your emails classified as bulk commercial mail, which is filtered more aggressively. Test with and without it, but always offer a simple way out ("just reply 'stop' and I won't email you again") and process every unsubscribe within 24 hours.
• Open and click tracking costs you deliverability. Open tracking relies on an invisible image; mail providers read the loading of external images as a bulk-sending signal. And the measurement is unreliable: privacy protections like Apple Mail's trigger automatic opens. A sharp swing (opens down 30%) is still worth digging into, but ask yourself whether that signal is worth what it costs. Steer by replies instead.

Content and copywriting

• One link at most. Links are the top phishing vector, so they're the first thing filters target. If a link is essential (calendar, resource), use just one, at the end of the message, and never in the first email if you can avoid it.
• Vary your content. Thousands of identical emails give away the automation: same phrasing, same structure, same subject line. Spintax and AI let you produce variants that keep the intent without repeating the same text.
• No images. No logo, no infographic, no embedded media. A text signature works better than an image signature.
• Plain text over HTML. Rich HTML triggers promotional filters. A prospecting email should look like an email written by a person.
• Rotate signatures. Three or four variants, so the same block isn't repeated on every send.
• Go easy on sales vocabulary. "Free," "guaranteed," "urgent," "limited offer," "click here": one word in context gets through, a pile of them trips the filters.

For the writing itself, our B2B cold email templates follow these constraints.

The legal framework in France

In B2B, email prospecting without prior consent is allowed in France under three conditions set out by the CNIL (the French data protection authority): the message must relate to the person's professional role (pitching payroll software to an HR director, not to a developer), the person must have been informed that their address would be used, and they must be able to opt out easily and free of charge with every send. Generic addresses (contact@, info@) fall under the same rules. On top of that, GDPR requires a documented legal basis (usually legitimate interest), disclosure of the data source and a limited retention period: the CNIL recommends not keeping an inactive prospect for more than three years.

These rules line up with deliverability best practices: targeting precisely, offering a simple way out, processing unsubscribes quickly and purging inactive contacts are also what reduce complaints and protect your domains.

The summary checklist

• ☐ Volume calculated from the pipeline target (deals → meetings → emails → contacts).
• ☐ Clean addressable market measured, number of touches per contact per year sustainable (two at most, ideally).
• ☐ Target's mail provider breakdown measured through MX records, security gateways identified.
• ☐ Sending pace chosen (conservative or moderate for the core setup), inboxes and domains calculated.
• ☐ Google / Microsoft split decided, domains spread across several accounts (3 to 5 per account).
• ☐ 20 to 30% buffer, in continuous warm-up.
• ☐ Sending domains separate from the primary domain, redirected, SPF (fewer than 10 lookups), 2048-bit DKIM, DMARC with reporting, alignment verified.
• ☐ Three weeks of warm-up before the first cold email, total volume held constant during ramp-up.
• ☐ Deliverability tests: mail-tester for every new domain or template, placement test every month.
• ☐ Bounces and volume monitored daily, replies and warm-up weekly, placement monthly.
• ☐ Written procedure for resting, rotating and retiring domains.
• ☐ Lists verified before every import, three suppression lists up to date, unsubscribes processed within 24 hours.
• ☐ Plain-text emails, one link at most, no images, varied content and signatures.
• ☐ Documentation up to date: domains, inboxes, accounts, DNS, schedules, incidents.

Cold email infrastructure is never finished. What works at 10,000 emails a month breaks at 50,000; what performs in Q1 degrades by Q3. That's not failure, it's what scaling looks like. The teams that get through it without damage are the ones that start from their reality (market size, pipeline target, prospects' mail providers), derive the infrastructure from it, then run it like an engine: monitor, test, document, optimize.

At Uclic, we build and run this kind of infrastructure for B2B scale-ups, from mapping the target to day-to-day deliverability monitoring, as part of our cold email and multichannel prospecting engagements.

Frequently asked questions