The question always arrives in the same shape: how many emails a month? Two, four, one a week? I have never had a good answer, because the question is aimed at the wrong thing.
An email costs nothing to send. It costs when it serves no purpose. And that bill is not settled at the moment of sending: it is settled later, on the sends that follow, and not always with the person who received the one too many.
Email pressure is not measured in sends per month. It is measured by what a pointless send destroys behind it.
1. Two exits, two very different bills
A recipient who is done with you has two doors. They unsubscribe, or they mark you as spam. To them the two gestures are neighbours, one click apart. To you they are not the same invoice at all.
An unsubscribe costs you one contact. That is expensive, because an acquired contact cannot be bought back at the price it was won for. But the loss stops there.
A complaint is paid by the whole list. Gmail tracks a spam complaint rate, recalculated every day, and Google sets the instruction in plain words in its email sender guidelines: keep that rate below 0.1 % and never let it reach 0.3 %. On a list of 20,000 addresses, 0.3 % is 60 people, and those 60 are enough to degrade delivery for the other 19,940.
Above that threshold, Google's sender guidelines FAQ is unambiguous: no mitigation is available on delivery problems, and the sender must stay below 0.3 % for seven consecutive days to become eligible again. It is the one number I really watch when I take over an email programme.
2. Bulk sender status never expires
Plenty of owners assume the rules are not aimed at them because they send little. The threshold is low: roughly 5,000 messages in twenty-four hours to personal Gmail accounts. It is counted at the primary domain, subdomains included. So 2,500 messages from your domain and 2,500 from a sending subdomain add up to 5,000.
The line nobody reads sits on the same page: senders who meet the criteria at least once are permanently treated as bulk senders. Google adds that the status has no expiry date and that changing your sending practices will not undo it.
In practice, one seasonal campaign sets the regime for the twelve months that follow. And since November 2025 Gmail has enforced that regime with temporary and permanent rejections, not only with a trip to the spam folder.
3. You do not get to decide that a message is promotional
Google states it: recipients, not Google, determine the nature of the messages they receive. A purchase receipt stays a receipt for exactly as long as nobody reads it as something else.
Which brings me to the practice I meet most often, and the one that costs most: slipping an offer into an order confirmation. Google lists it among the practices to avoid, in so many words: do not add promotions to purchase confirmation messages. The logic holds. A receipt always gets read, and that is precisely what makes it a poor advertising slot. The day the reader marks the whole thing as spam, the receipt goes with the promotion, and the complaint rate does not tell the two apart.
Google's email subscription guidelines for senders draw the technical consequence: one From address for subscription messages, another for receipts and password resets, and a human-readable List-id header per list. This is not housekeeping. A reputation damaged by the newsletter stops contaminating the emails your customer is actually waiting for.
| What the reader does | What it costs you | What usually triggered it |
|---|---|---|
| Ignores the message | Nothing today, a complaint later | Content that has nothing to do with them |
| Unsubscribes | One contact, for good | No intermediate exit on offer |
| Marks as spam | Delivery, across the whole list | An unsubscribe link missing or ignored |
| Marks a transactional email as spam | Receipts and password resets land in spam | A promotion slipped into a confirmation |
4. A segmented unsubscribe keeps the reader a single exit would lose
Since June 2024, one-click unsubscribe has been required on every marketing or promotional message above the sending threshold. It relies on the List-Unsubscribe headers defined by RFC 8058, and two traps come up again and again: a mailto link does not meet the requirement, and neither does a button that lands on a preferences page.
The same documentation hands over the lever almost nobody uses. One-click unsubscribe only removes the recipient from the list attached to the message they received. In other words, you set the granularity, by segmenting your lists.
A reader saturated by your weekly offers can leave those and stay on the monthly letter, on one condition: that the two are two distinct lists. If there is only one, the only available exit is the total one. That is the most common setup, and it is a configuration choice rather than a fact of life.
The rest is a matter of timing. Google asks senders to honour unsubscribe requests within forty-eight hours and counts a breach of that deadline among the failures that remove access to support on delivery problems. The upside is written by Google itself: making it easy to unsubscribe can improve open rates, click rates and the effectiveness of your sending.
5. Remove the inactive reader before they complain
This is the most counter-intuitive line in the documentation, and it comes from Google: consider unsubscribing recipients who do not open or read your messages. No sending platform will ever suggest it, since billing runs on contact volume.
An inactive address weighs twice. It inflates the denominator of your sends without ever feeding the numerator of clicks, and it is the first to hit the spam button the day a subject line irritates.
Watch the instrument, though. Google states on the same page that it does not monitor open rates, and that a low open rate is not a reliable indicator of a delivery problem. Opens therefore no longer diagnose deliverability, for the reasons I set out in the piece on behavioural segmentation. They stay useful for one decision: who you stop writing to.
Further upstream, the same documentation asks for double opt-in, meaning confirmation of the address after it has been entered on your site or in your app. It is the last moment at which a fake address, or one entered by somebody else, can still be kept out of the list. An address that never confirmed will eventually complain, and on a shared pool it damages the reputation of the other senders too, as I described in the piece on deliverability without a dedicated IP.
6. What European law regulates, and what it leaves open
If you send to recipients in France or elsewhere in the European Union, this part applies to you whatever your own jurisdiction, and it is the part foreign senders discover late.
No European text sets a number of emails per month. The ceiling is set by the inbox, not by the regulator. What the law regulates is entry to the list and exit from it.
The French data protection authority, the CNIL, restates the principle for consumers in its guidance on commercial prospecting by electronic mail: prior consent, freely given, specific, informed and unambiguous, collected through a dedicated box that is not pre-ticked. Accepting terms and conditions does not amount to consent.
The exception everybody leans on is the existing customer contacted about similar products or services. The CNIL has tightened how it reads that exception, and this is the point most senders get wrong: it cannot be relied on where no sale or service has actually taken place, including where the customer has merely created an online account. An account opened without an order does not make a customer.
For business recipients, prospecting can rest on legitimate interest, provided the approach relates to the role the person actually performs, and provided they were informed and can object. Generic addresses such as contact@ or info@ belong to the company rather than to an individual and fall outside that regime.
Two points are not negotiable. Harvesting addresses from public spaces on the web, directories and forums included, is prohibited, as the CNIL's golden rules for email prospecting set out. And the right to object is absolute: it needs no justification, and the organisation must answer requests to exercise data rights within one month at most.
In the crucible: frequency is not the subject
The lead is the sending calendar decided in a meeting, a round number laid over twelve months, and the list burned in small doses to keep to the cadence announced.
The gold takes more work to set up and lasts far longer: lists segmented finely enough that a reader can turn the volume down instead of walking out, inactive addresses removed before they complain, one From address per type of message, and the complaint rate watched weekly the way you watch a cash position.
With those four pieces in place, the question of how many sends dissolves on its own. You write when you have something to say, and the list holds.
Is your list wearing out faster than it fills up?
I take the sending machinery apart end to end: list segmentation, separate From addresses, removal of inactive contacts, complaint rate monitoring. See the email marketing and automation service, the analytics and tracking service or the CRO and conversion rate service.