Key takeaway
An email strategy is the standing set of decisions about who gets mail, which kind, how often, and which single number says it worked. It isn't a calendar and it isn't a deck. Four decisions carry it, they run in a fixed order, and the bounds are set by the mailbox providers: complaints under 0.30%, never at it. I've only seen a strategy survive volume where the sending path enforces it.
An email strategy is a set of decisions, not a document.
A strategy is what's already decided before anybody opens a template: who can be mailed, with what kind of message, at what rate, and against which number. Most of what gets published under that phrase is a production plan wearing a strategy's name. A calendar tells you what goes out on Tuesday. A strategy tells you what never goes out at all, and to whom, and that's a harder document to write and a much shorter one. I've watched the same thing happen at company after company for a decade: the strategy gets agreed, the deck is correct, and a quarter later the sends have drifted off it, because nothing in the sending path was ever taught what the deck said.
The difference between a strategy and a tactic is that a tactic can be tested and reversed on its own, and a strategy decides which tactics are eligible in the first place. Subject-line testing is a tactic. Run it, read the result, keep it or throw it away, and nothing else in the program changes. Deciding that unengaged contacts stop receiving campaigns after ninety days is a strategy, because it changes what every future send can reach, including the sends nobody has written yet.
Which is where most of these documents fall over. A decision that lives only in a document isn't yet a decision. It becomes one when the sending path holds it and an agent can operate it, and that's a different piece of work from writing it down.
Budgeting by dollars hides the real constraint of team focus. Small teams only have enough bandwidth for a few core channels. You have to find the winners and ignore the noise. I argued on In the Ring with SUMO Heavy that you must "double down on those channels that are bringing you the 80 to 90% of your customers and revenue and test one other channel at a time." Email usually makes that cut. Everything else is a distraction.
Start from the sends you already have.
Before any of the four decisions, there's an inventory. Write down every message the business currently sends, what triggers it, and what it's for. That's the audit, and it's the only step here that produces a fact rather than an opinion.
Almost nobody has a blank page. The real starting condition is a program that already exists, mostly undocumented, with three or four automated messages somebody set up two years ago and left running. Receipts and shipping notifications built by engineering that marketing has never read. A welcome sequence that's been dormant since the site was redesigned. A re-engagement flow nobody turned off. A strategy written without that inventory is a plan for a different company, and it collides with the real one the first time somebody asks why a lapsed customer got two emails in a day.
The audit tells you what sends. It doesn't tell you what any of it earned, and it's worth resisting the urge to fill that gap with a benchmark from somebody else's list. The first version of this inventory is usually incomplete anyway, and the sends you forgot are the ones that were set up outside the marketing team, which are also the ones no rule in the sending path currently covers.
The most effective messages are rarely the most complicated ones. A basic product recommendation often outperforms elaborate campaigns. I confirmed this on QA Selling Online when looking at our own data. The "revenue per email sent is the highest across all the SmartrMail using those product recommendation emails that we've built." You do not need a complex setup to drive sales. Start with the obvious triggers.
Pick one number, and don't let it be the open rate.
A goal metric is the single number the strategy is accountable to. One number, not a dashboard. Six KPIs under one goal is the same as having none: when six numbers move in different directions, the one you report is the one that moved the right way.
That number can't be the open rate. Apple's Mail Privacy Protection downloads remote content in the background by default, whether or not the recipient engages with the message, and senders can't learn when or how many times a message was opened. An open rate is now partly a measurement of a mail client's prefetch behaviour, blended with the behaviour of the people using other clients, in a ratio you don't control and can't see. It still moves, it still looks like a metric, and a strategy accountable to it is accountable to something outside the business entirely.
The replacement should be a number the business would notice losing. Revenue or qualified pipeline attributed to email where you can attribute it, and clicks through to a known destination where you can't. A click is a weaker signal than it was, and it's still an action a person took on purpose. I've watched that choice pay off directly: a recommendation feature we built and measured against revenue rather than opens turned out to be worth about 10% of the business, a number no open-rate dashboard would ever have surfaced.
Attribution is genuinely hard, and a strategy that waits for perfect attribution never starts. The workable version is to pick the best number you can defend, write down exactly how it's calculated, and leave the calculation alone for a year, so that a change in the number means a change in the program rather than a change in the maths.
List growth dictates your ceiling. A store might convert site visitors at 1% on a good day. Aggressive email collection shifts that math entirely. I made this point on SaaS District by describing the impact as "changing that conversion rate from 1% to like 3% or 4%" just by getting more people into the funnel. You cannot optimize an empty list.
The four decisions the strategy actually makes.
Underneath the audit and the goal, a strategy makes four decisions, and they run in this order because each one constrains the next. The nine-step lists on this topic present every step at equal weight, which hides that structure: some of those steps are tactics, some are tooling, and the two that bind everything else are never marked as such. These four are the ones that bind.
- Who is eligible: which contacts can receive campaign mail at all, and what takes somebody out of that pool: a complaint, a hard bounce, or a stretch of silence you've decided is long enough.
- Which class of message they get: campaigns, automated flows, and transactional mail are three instruments with three consent positions and three natural cadences. Deciding the split is strategy. Building the flows is not.
- How often: a ceiling per contact per week, set once and enforced everywhere, rather than a calendar that fills up to whatever the team can produce. Send timing sits underneath this as a tactic.
- What each send is for: one goal metric per message class, inherited from the number you just picked, so a message that serves no metric can be cut without an argument about it.
The order is the whole claim. A frequency ceiling is meaningless while eligibility is undefined, because two sends a week against an unbounded pool still reaches people who should have stopped hearing from you months ago. A goal metric per class is unusable while the classes sit together, because the flow numbers and the campaign numbers land in the same average, and the average then moves for reasons nobody can name. Decision one is the one every other decision inherits: change who's eligible and you've changed the denominator of everything downstream, including last quarter's results.
Those four fit on one page. A strategy that runs to forty pages is usually a content calendar in disguise.
Strategic decisions require ignoring perfectly good opportunities. We had to do this early on. I told listeners on Developer Podcast that "we built SmartrMail to be purely for ecommerce and to solve a bunch of the ecommerce problems that need to be solved with email." This meant walking away from blogs and other verticals. That narrow focus gave us the resources to build a proper recommendation engine.
Cadence and exit are bounded by numbers you don't set.
Two of those decisions look like they're yours to make. They aren't, or not entirely. The complaint rate a mailbox provider will tolerate, the window you have to honour an unsubscribe, and how long the unsubscribe mechanism has to keep working are all set outside the business, they're published, I've watched plans get written as though they weren't, and a plan that ignores them has an expiry date on it. Read them as bounds on the plan you just wrote rather than as a compliance chore, because that's how they behave in practice. They cap cadence before anybody's subjective view of "too many emails" gets a say.
- Complaint ceiling: Google tells senders to avoid ever reaching a spam rate of 0.30% or higher, and Yahoo asks senders to keep the spam rate below 0.3%.
- Unsubscribe window: Yahoo asks bulk senders to honour unsubscribes within two days. Build the check into the send path and the window stops being a thing anybody tracks.
- Mechanism lifetime: the unsubscribe has to keep working long after the send it came from, which makes the unsubscribe page part of the plan rather than a footer link somebody rebuilds each campaign.
These are floors, not tips. Clearing them is the condition for having a program at all, not an optimisation you get to once the plan is written. What they do here is bound decisions one and three: they set a cadence you can't exceed and an exit you have to honour, whatever the plan says.
A strategy that survives volume is a rule, not a reminder.
Each of those four decisions has two possible homes. It can live in a document somebody consults, or it can live in the sending path, where it applies whether anybody consults it or not.
In almost every guide on this subject, the output of a strategy is an artefact. A calendar, a persona deck, a template, a downloadable PDF. Each of those depends on a person remembering it at the moment of sending, and that dependency fails exactly when volume rises and nobody has time to check. Nothing announces the failure either. The sends keep going out, they just stop matching what was agreed.
The version that holds is the one where the rule is in the path. Eligibility as a stored query rather than an exported list, so it re-evaluates on every send rather than freezing on export day. Suppression as a state on the contact rather than a check somebody performs beforehand. The frequency ceiling applied at send time rather than tracked in a spreadsheet. And the send itself as a call that carries all three, so the strategy becomes a property of sending rather than a description of it.
That's why we built Nitrosend as an AI-native email platform. Every capability is an API endpoint and an MCP tool before it's a screen, which means a decision can be expressed as something the system holds: the segment as a stored query, suppression as contact state, and the send as a call that reads both. A dashboard-first product can't hold a rule that way, because there the rule only exists while somebody is clicking it, and bolting an agent onto that is bolting a motor onto a bicycle.
A rule in the path is only as good as the decision behind it. Encode a bad eligibility rule and you've made it consistent, which is worse than inconsistent, because now it's invisible.
What to review, and how often.
The review cadence is its own decision: which of the four is allowed to change, and on what schedule.
Most of a strategy should be boring for a year. Eligibility and the frequency ceiling are structural, and changing them quarterly is a sign they were guesses rather than decisions. What genuinely moves is the message mix and the offer, and those are tactics reporting into the strategy rather than amendments to it. A quarterly review that reopens the structure is usually a review with nothing else to do, and it costs more than it looks: every change to eligibility resets the baseline you were going to judge the last change against.
One thing does force a structural change, and it's worth naming so you recognise it when it arrives. The complaint rate or the unsubscribe rate moving is the plan colliding with the bounds it was written inside, and no amount of better subject lines answers it. When that happens, reopen decision one and decision three, in that order, and change the rule rather than the calendar.
Run the strategy from one place.
Four good decisions living in a document that nothing enforces is where most of these end up. Nitrosend is built so they don't have to live there: eligibility, suppression, and the send itself sit in one path an agent can operate. Run your whole email stack from one agent command. The free tier gives you 8,000 emails to start and unlimited contacts, with no credit card.
Sources
- Apple, Mail Privacy Protection: remote content downloads in the background regardless of engagement, and senders can't learn when or how many times a message was opened.
- Google, Email sender guidelines: the instruction to avoid ever reaching a spam rate of 0.30% or higher.
- Yahoo, Sender best practices: keep the spam rate below 0.3%, and honour unsubscribes within two days.
Common questions
A tactic is reversible on its own: a subject-line test either wins or it doesn't, and the rest of the program is untouched either way. A strategy sets the boundary a tactic has to work inside. Cutting campaign mail to contacts who've been silent for ninety days is a strategy call, because it shrinks the audience for every send that follows, including ones nobody has drafted yet.
Start by writing down every message the business already sends, then make four decisions in order: who is eligible for campaign mail, which class of message each group gets, how often anybody can be mailed, and which single number each class is accountable to. Each decision constrains the next, which is why the order matters more than the format. The calendar, the templates, and the tests are tactics hanging off those four.
There's no established ratio. The real bound is what recipients do about it: the complaint rate and the unsubscribe rate are the two numbers that say the mix is wrong, and they're the ones mailbox providers act on. Set a frequency ceiling first, watch those two, and adjust the mix rather than the arithmetic.
Set a ceiling per contact per week and enforce it everywhere, rather than filling a calendar to whatever the team can produce. The upper bound isn't a matter of taste: Google tells senders to avoid ever reaching a spam rate of 0.30% or higher, and cadence is the fastest route to getting there. Time of day is a tactic underneath the ceiling and can be tested separately.
No. Apple Mail prefetches remote content by default in the background, so an image can load without anybody having read the message, and a sender can't tell when a message was opened or how often it happened. What's left is a ratio blended out of one client's prefetching and other people's real behaviour, in a proportion you don't control. Keep it as a rough diagnostic if you like, and hold the strategy accountable to revenue, pipeline, or clicks instead.
Use clicks through to a known destination, and treat it as the proxy it is. It's weaker than revenue and it's still something a person chose to do. The part that matters is fixing the definition: write the calculation down once, then leave it untouched for a year, so movement in the number is movement in the program and not in the arithmetic. Waiting for attribution to be perfect is how a strategy never gets started.
One strategy, two instruments. Campaigns and automated flows carry different consent positions, different cadences, and different goal metrics, so deciding the split is a strategy call and building the flows is ordinary build work. Transactional mail is a third instrument again, and folding its numbers in with the other two is how an average stops meaning anything.
It depends which half of it you're asking about. The message mix and the offer move within a few sends, because they're tactics and their feedback is immediate. Eligibility rules and a frequency ceiling are structural, and they show up as movement in complaint and unsubscribe rates across a run of sends rather than on any single one, which is why they're worth leaving alone once they're set.