Vega

A retail brand he worked with launched a major holiday campaign. The marketing team was thrilled. Open rates were 31% on the first send, against a 22% baseline. The team celebrated. The director sent the metric screenshot to the CEO. The CEO replied with applause. There was talk of replicating the campaign approach for the rest of the season.

Vega asked for the revenue numbers.

Revenue per send: $0.87. The baseline (the boring weekly newsletter) was running $1.04 per send. The "successful" campaign was producing 19% less revenue per send than the routine cadence.

He brought the numbers to the marketing director. The director didn't want to hear it. The CEO had already congratulated the team. The narrative was set. The data was inconvenient. Vega persisted. He showed the cohort breakdowns. The campaign had attracted lower-quality opens (subscribers who opened out of curiosity at the subject line but who were less likely to convert). The opens were inflated. The conversion rates were lower. The math was clear.

The CEO eventually saw the revenue analysis. The campaign approach was killed before being scaled. The retention strategy that quarter went back to the boring weekly newsletter, which produced more revenue than three of the planned holiday-style campaigns combined.

The marketing director was annoyed. Vega didn't apologize. The lesson he teaches now: opens are vanity. Revenue per send is reality. The senders who learn the difference do better than the ones who don't. The senders who don't, eventually meet someone like Vega in a meeting where the numbers can't be ignored anymore.

---

A SaaS sender had stable engagement metrics. Open rate flat at 28% for three years. Click rate flat at 4%. The dashboards looked healthy. The team was confident.

Vega ran a cohort analysis. He grouped subscribers by the month they signed up, then tracked engagement separately for each cohort over time. The result was unsettling.

Subscribers who joined in 2022 were now opening at 8%. Subscribers who joined in 2023 were at 12%. Subscribers who joined in 2024 were at 22%. New subscribers from the past 90 days were at 41%.

The flat overall open rate was hiding a steep decline within each cohort, masked by a continuous flow of new subscribers replacing the engagement that the older cohorts were losing. The list was a leaky bucket. The senior sales team had been making decisions for three years based on a stable headline number that was actually disguising a structural problem. Long-term engagement was collapsing. They just couldn't see it because the new subscribers kept refilling the average.

He showed the team the cohort chart. They were quiet for a long time. The fix was a serious investment in re-engagement and content quality, neither of which had been priorities while the headline number looked fine. The team rebuilt their measurement framework around cohort tracking instead of overall averages.

Eighteen months later, the cohort decay had been arrested. Not reversed (cohort decay is hard to reverse), but stabilized. The newer cohorts were holding their engagement longer than the older cohorts had.

The lesson he teaches now: stable averages lie. Cohort analysis tells the truth. If you've never grouped your subscribers by signup period and tracked engagement separately for each, you don't actually know what's happening in your list. You know what the average says, which is rarely the same thing.

Vega's intro:

Email analytics is the practice of measuring an email program honestly. The "honestly" is the hard part. Most senders have dashboards. Most senders look at them. Few senders read them in a way that survives second-guessing. Here is what email analytics actually is, what to track, and why most of what gets celebrated in industry reports is a distraction.

---

A measurement is a fact. Stars are at certain altitudes at certain times. Numbers in your dashboard are the values they are. Neither lies on its own. What lies is the interpretation, which can be optimistic, defensive, or oriented around whatever the person looking at the dashboard wants to be true.

Analytics is the discipline of reading the numbers without lying to yourself.

This is harder than it sounds.

The senders I work with who do this well share three habits.

One: they distinguish between metrics that move money and metrics that look good. A 50% open rate looks good. It might or might not move money. A 5% click-to-open rate is less impressive in screenshots. It correlates with revenue much more reliably. The senders who do well over time learn to favor the second kind even when the first kind is what gets reported.

Two: they review the same metrics on the same schedule. Once a week, same day, same time. The discipline of consistent review surfaces patterns that ad-hoc looking misses entirely. A subscriber who looks at the dashboard once a month and notices the open rate is fine has not done analytics. They've done a vibe check.

Three: they track trends, not snapshots. A single week's metrics are noise. Three weeks of declining metrics are signal. The senders who do this well react to signal, not to noise.

What to actually track

There are four metrics that matter for almost every sender. Most other dashboard items are decorative.

Deliverability rate. The percentage of sends that reach the inbox. Different from "delivery rate" (which includes spam-folder placement). Inbox placement specifically. This is the foundation. If 30% of your emails go to spam, every other metric on the dashboard is misleading because the addressable audience is smaller than the dashboard suggests.

You can't measure inbox placement perfectly without seed-list testing tools. You can approximate it by combining bounce rate, complaint rate, and engagement decay. If bounces are climbing, complaints are climbing, or engagement is dropping unexpectedly across the list, deliverability is declining even if the dashboard doesn't show it directly.

Click-to-open rate (CTOR). Clicks divided by opens, expressed as a percentage. This is the single most underused metric. CTOR tells you what percentage of the people who actually saw your email did something with it. Even if "opens" are inflated by Apple's Mail Privacy Protection (which they are for most senders post-2021), CTOR remains directionally useful because both opens and clicks are measured the same way and the inflation cancels.

A healthy CTOR is industry-dependent. For most senders, 5-15% is normal. Above 20% is excellent. Below 3% suggests subject lines that promise more than the body delivers, or content that doesn't match the audience's interest.

Conversion rate. The percentage of subscribers who completed the action the email was driving (purchase, signup, download, booking). Conversion rate is closer to the bottom line than any upstream metric, and it's the closest thing to ground truth.

The conversion that matters depends on the email's purpose. A newsletter might have no conversion at all (the goal is engagement). A promotional email's conversion is the purchase. A re-engagement email's conversion is the click that signals "yes, keep me on the list." Define the conversion before measuring it.

Revenue per send. Total revenue attributed to a campaign, divided by the number of emails sent. This is the metric that survives every other distortion. Open rates can be inflated. Click rates can be misleading. Conversion rates can be skewed by a few large purchases. Revenue per send tells you whether the send was worth doing.

Revenue per send has a whole article of its own (see article 4). The short version: most senders don't track it, and the ones who do tend to make better decisions than the ones who don't.

What not to track (or to track less)

A few common dashboard items that are popular and not very useful.

Total opens. A vanity number. Total opens go up when the list grows. They go up when you send more campaigns. They don't tell you whether the program is healthy.

Total clicks. Same problem. Useful for gross-volume reporting, not for decision-making.

List size. Goes up over time as long as you're acquiring subscribers. Doesn't tell you whether the list is healthy. A list of 500,000 with 3% engagement is worse than a list of 50,000 with 30% engagement, but the headline number suggests the opposite.

Open rate (in isolation). Since 2021, open rates are largely fictional for senders with significant Apple Mail audiences. The number can stay flat or even rise while real engagement collapses. Open rate is still useful when paired with CTOR, but on its own it misleads.

Engagement-to-list-size ratio. Some dashboards show "engagement rate" as engaged subscribers divided by total subscribers. This number drops when you stop suppressing dormant subscribers, which is the wrong direction. Track absolute engagement count or per-segment engagement instead.

How to read the dashboard

The discipline I use:

First, check the trend. Are the four core metrics (deliverability, CTOR, conversion, revenue per send) trending up, flat, or down over the past four weeks? Trend is the first read.

Second, check by cohort or segment. Even if the overall trend looks fine, are specific cohorts or segments showing different patterns? Average can hide a divergent subgroup.

Third, check the correlation. If open rate went up, did revenue per send go up correspondingly? If yes, the move is real. If no, the open rate change is a distraction and the revenue is the truth.

Fourth, check for anomalies. Anything dramatically different from baseline? A spike in unsubscribes? A drop in deliverability? Anomalies are signals to investigate.

Fifth, decide. Based on the read, what (if anything) should change? Most weeks the answer is "nothing, hold steady." Some weeks the answer is "investigate this specific anomaly." Rarely the answer is "act on this trend."

Where to start

If you've never run analytics with this kind of discipline, start small.

One, define your four metrics. Use the four above unless your business has a specific reason to vary.

Two, find them in your ESP's dashboard. Most ESPs surface CTOR but bury it. Make sure you can find each metric in 30 seconds.

Three, schedule a weekly review. Pick a day and time. Block it. Review the four metrics over the past four weeks. Write down anything notable.

Four, do this for one quarter. By the end of the quarter, you'll know your real baseline, your typical week-to-week variation, and what counts as a real change.

Five, build cohort analysis once the baseline is established. (See article 5.)

That's the framework. The rest is practice.

Why this matters

Senders who run on intuition and screenshots eventually meet someone with data. The meeting goes badly for the screenshot side. Senders who measure their own program honestly are the ones who can defend their decisions, anticipate problems, and improve over time. The accuracy is the point.

Stars are where they are. Numbers are what they say. Read them.

- Vega

Vega's intro:

A retail brand celebrated a campaign with a 31% open rate against their 22% baseline. The marketing director sent the screenshot to the CEO. The CEO replied with applause. I asked about the revenue numbers. Revenue per send was $0.87. The baseline weekly newsletter ran $1.04. The "successful" campaign was producing 19% less revenue per send than the routine cadence. The campaign was killed before it scaled. Most senders don't track revenue per send. Most senders should.

---

Revenue per send is exactly what it sounds like. Total revenue attributed to a campaign, divided by the number of emails sent. The number is in dollars (or your local currency).

A campaign that sent 50,000 emails and produced $50,000 in attributable revenue has a revenue per send of $1.00. A campaign that sent 100,000 emails and produced $40,000 has a revenue per send of $0.40.

The first campaign was more profitable. The second campaign produced more total revenue but worse per-send economics. Most dashboards report total revenue and let you draw your own conclusions. Most senders draw the wrong conclusion (the bigger total wins). Revenue per send is the metric that surfaces which campaign was actually better.

Why this metric beats almost every other one:

It survives MPP. Open rates are inflated. Revenue isn't.

It survives subject-line A/B noise. A subject line that boosts opens but attracts non-buyers shows up here as a worse number.

It captures conversion friction. A campaign that drives lots of clicks but the landing page converts poorly shows up here.

It accounts for list quality. A small engaged list often beats a large dormant list on revenue per send. The metric rewards the work of keeping a list clean.

It maps directly to business outcomes. Marketing's job is eventually to produce revenue. This is the metric that says whether the work is producing it.

How to calculate it

Two pieces of data:

Total revenue attributed to the campaign. This usually comes from your ecommerce platform or CRM, with the campaign tagged via UTM parameters or unique tracking links. The attribution model matters (more on this below).

Total emails sent. Available from your ESP. Easy.

The math: revenue divided by sends.

If a campaign sent 25,000 emails and produced $18,500 in attributable orders, revenue per send is $18,500 / 25,000 = $0.74.

That's the basic version. The harder version is making sure "attributable revenue" is calculated honestly.

Attribution models

The attribution question is: how do you know a purchase came from a specific email?

Last-click attribution assigns the revenue to whichever marketing channel produced the last click before the purchase. If the customer clicked an email, then visited the site again from a Google search, then bought, last-click gives the revenue to Google search, not the email.

First-click attribution assigns it to the first touch. The customer who first heard of you through an email gets the email credited even if they later came back through other channels.

Multi-touch attribution spreads the credit across multiple touches, weighted somehow. The model varies (linear, time-decay, U-shaped, etc.).

For revenue per send specifically, the cleanest version uses a "view-through plus click" attribution window of 7-14 days. Any purchase made within 7-14 days of clicking the email gets credited to the email. This is generous enough to catch the email's real influence and tight enough to filter out coincidental purchases.

Most ESPs (Klaviyo, Mailchimp, Brevo, Cakemail, etc.) handle this attribution automatically when configured correctly. Set up UTM parameters on your campaign links. Configure the ESP's attribution window. The dashboard will show attributed revenue per campaign.

What "good" revenue per send looks like

The honest answer: it depends on your business.

Ecommerce DTC: $1-5 per send for healthy brands. Above $5 is excellent. Below $0.50 means the list isn't producing.

B2B SaaS: harder to attribute directly because the sales cycle is longer. Track downstream revenue with a 30-90 day attribution window. Healthy ranges vary widely by ACV, but $5-50 per send for inbound nurture is common.

Newsletter / content businesses: revenue per send might be very low ($0.01-$0.10) because the email itself isn't selling anything. The metric still matters as a relative measure across campaigns, but the absolute number isn't comparable to ecommerce.

Nonprofits: donations per send. Wide variation by appeal type and audience. Annual averages are more meaningful than per-campaign.

The point isn't hitting an industry benchmark. The point is comparing your campaigns to each other consistently and tracking the trend over time.

What revenue per send tells you

Three patterns to watch.

Trend over time. Is revenue per send going up, flat, or down quarter over quarter? A flat or rising trend means the program is healthy. A declining trend means something is wrong, even if other metrics look fine.

Variance across campaigns. Some campaigns will outperform others. The high performers tell you what works for your audience. The low performers tell you what doesn't. Repeat the patterns from the winners.

Variance across segments. When you can break revenue per send down by segment (new vs engaged vs lapsed, by demographic, by acquisition source), you find which audiences are actually paying for the program. This often surfaces uncomfortable truths: a small high-value segment might be subsidizing a large low-value one.

Why senders don't track this

Three reasons, in order of how often I see them.

One: they don't know how. Setting up attribution correctly is moderately technical. Senders without analytics expertise default to whatever the ESP shows by default, which is usually total revenue, not per-send.

Two: it makes campaigns look worse. A campaign with an exciting open rate but a mediocre revenue per send is uncomfortable to report. Some teams avoid the metric to avoid the conversation.

Three: the marketing team isn't measured on revenue. When the marketing team's KPI is "open rate" or "list growth," they don't track revenue per send because nobody asks for it. The CFO would, but the marketing dashboard isn't built for the CFO.

The senders who do track it are usually the ones whose marketing function is held accountable for revenue, not for activity metrics. The accountability is what produces the discipline.

How to start tracking it

If you don't already track revenue per send, three steps.

Set up UTM parameters on every campaign link. Source, medium, campaign name, content. Most ESPs auto-generate these if you tag campaigns properly. Without UTMs, you can't attribute revenue back to specific sends.

Configure attribution in your ESP or analytics platform. A 7-14 day click attribution window is the default for most senders. If your ESP doesn't surface revenue per send directly, you can calculate it from the campaign-level data.

Add revenue per send to your weekly review. Look at the trend. Look at variance across campaigns. Look at variance across segments if you have the data.

The first month, the numbers might look surprising. The high-open campaigns might be the low-revenue ones. The boring weekly newsletter might be the highest revenue per send in the program. These are the patterns the metric is designed to surface.

What to do once you know

Use revenue per send to make decisions. Kill campaigns that consistently underperform on revenue, even if their open rates look fine. Double down on patterns that consistently produce revenue, even if they look unexciting. Match send frequency to revenue capacity (sending more often only helps if revenue per send holds; if it falls, you're cannibalizing the program).

The metric is honest. The decisions that follow from it are usually better than the decisions made on opens, clicks, or list size alone.

Stars don't lie. Neither does revenue per send. Read it.

- Vega

Vega's intro:

A SaaS sender had stable engagement metrics for three years. Open rate flat at 28%. Click rate flat at 4%. Dashboards looked healthy. The team was confident. I ran a cohort analysis. Subscribers from 2022 were opening at 8%. Subscribers from 2023 were at 12%. Subscribers from 2024 were at 22%. New subscribers from the past 90 days were at 41%. The flat overall average was hiding a steep collapse within each cohort, masked by new subscribers replacing what the older cohorts were losing. The list was a leaky bucket. The team rebuilt their measurement around cohort tracking. Eighteen months later, the decay was arrested.

---

A cohort is a group of subscribers who share something. The most useful cohort for email is "subscribers who joined in the same time period." January 2024 cohort. February 2024 cohort. Q3 2025 cohort.

Cohort analysis is the practice of tracking each cohort's behavior separately over time, instead of looking at the full list as one undifferentiated mass.

The reason this matters: averages lie when the underlying population is changing. An overall metric like "list-wide open rate" is the weighted average of every subscriber's behavior, regardless of when they joined. If new subscribers are joining at high engagement and old subscribers are decaying at the same rate, the average can stay completely flat while the actual experience of every individual cohort is getting worse.

This is the leaky bucket problem. The bucket has water (subscribers). New water is pouring in (signups). Water is leaking out the bottom (engagement decay). If the inflow matches the outflow, the water level (overall metric) stays the same. The bucket is still leaking. You just can't tell from the level.

How cohorts reveal what averages hide

Three patterns cohort analysis surfaces that overall averages don't.

Engagement decay. Every cohort's engagement decays over time. Healthy senders have slow decay (engagement holds for 12-24 months before noticeably dropping). Unhealthy senders have fast decay (engagement drops 50% within 6 months). Without cohort tracking, you can't see the decay rate. You see only the average, which gets refreshed by new signups.

Quality differences across cohorts. Subscribers acquired through different sources or in different periods often behave differently. Subscribers from a Black Friday acquisition campaign might have lower long-term engagement than subscribers from organic referrals. Cohort analysis surfaces this. The aggregate metric doesn't.

The honest size of the engaged list. Most senders overestimate their engaged audience because they look at the recent cohort's engagement and assume it applies to the whole list. Cohort analysis shows that the recent cohort might be 41% engaged while the cohort from two years ago is at 8%. The overall "engaged audience" is much smaller than recent cohort numbers suggest.

How to build a cohort analysis

The math is simple. The setup takes a couple hours the first time and almost no time afterwards.

Step one: pull subscriber list with signup dates.

Export from your ESP. You need at minimum: subscriber ID, signup date, and an engagement signal (last open or last click date works).

Step two: bucket subscribers by signup month or quarter.

Each bucket is a cohort. For monthly cohorts, group by signup year-month (e.g., "2024-01," "2024-02"). For quarterly cohorts, group by year-quarter (e.g., "2024-Q1"). Pick the granularity that makes the chart readable.

Step three: calculate engagement rate per cohort, by month.

For each cohort, calculate "what percentage of this cohort opened or clicked at least once in the most recent 30 days?" Repeat for the previous 30 days. And the 30 days before that. You're building a time series for each cohort.

The output is a chart with months on the x-axis and engagement percentage on the y-axis, with one line per cohort. The newest cohort starts on the right side at high engagement. Older cohorts are lines that started at high engagement at signup time and have drifted down since.

Step four: read the chart.

Look at three things.

The starting engagement rate. Where does each cohort begin? If newer cohorts start lower than older cohorts did, your acquisition quality is declining over time. If they start higher, acquisition is improving.

The decay rate. How fast does each cohort drift down? Steep decay means subscribers are losing interest fast. Gentle decay means engagement holds. Different cohorts can have different decay rates depending on what was happening in the program when they joined.

The cross-over point. When does each cohort drop below your "Engaged" threshold (typically 90 days without engagement)? The earlier the cross-over, the worse the cohort.

What cohort analysis tells you to do

Three actions, depending on what the chart shows.

If new cohorts start at lower engagement than old cohorts did, acquisition quality is declining. Investigate the acquisition channels. The newer signups might be coming from sources that produce less engaged subscribers. Common culprits: incentive-based signups, list buys, low-quality referrals.

If decay is steep across all cohorts, content quality is the issue. Subscribers signed up for what you promised, then disengaged when the actual sends didn't match. Audit the welcome series. Audit content alignment with signup promise. Audit cadence (sending too often is a common cause of fast decay).

If decay is steep only in older cohorts, the program drifted. The brand or content may have shifted in ways that don't match what older subscribers signed up for. A re-permission campaign on the oldest cohorts gives them a chance to re-confirm or graceful exit.

Cohort analysis for revenue, not just engagement

The same technique works for revenue. Group subscribers by signup period. Track per-cohort revenue per send over time. The cohorts that produce more revenue per send are the high-quality cohorts. The ones that produce less are the lower-quality acquisitions.

This is one of the most useful business questions in email: which acquisition sources produce subscribers who actually pay? Cohort analysis surfaces it. The answer is sometimes uncomfortable. The Q4 holiday campaign that drove 40,000 signups might be the cohort that produces the lowest lifetime revenue, while the slower organic-referral pipeline produces the highest. The marketing team that brings this up to leadership is the team that gets resources to fund the right acquisition.

How often to run it

Quarterly is the cadence I recommend. Build the cohort chart once, save the methodology, and re-run it every quarter. Three quarters of cohort tracking shows you the structural pattern. By a year in, you can predict next quarter's behavior with reasonable accuracy.

For senders with a lot of activity, monthly cohort review can surface trends faster, but the per-month variation is noisier. Quarterly is the sweet spot for most senders.

The cohort-aware sender

The senders who run cohort analysis make different decisions than senders who don't. They invest in retention because they see the decay. They cut acquisition channels that produce low-quality cohorts. They redesign welcome series after seeing how fast new cohorts decay without good onboarding. They act on patterns the surface metrics never reveal.

Most senders never run cohort analysis. The ones who do are usually the ones with the healthiest programs five years out, because they caught structural issues that the headline metrics hid.

The leaky bucket isn't a metaphor for a small subset of senders. It's the default condition of email lists. New subscribers replace the engagement that older subscribers lose. The replacement is invisible until you separate the streams.

Separate the streams. Read each one. The patterns are there.

- Vega