Email sending analytics from PowerMTA and MailerQ reports
How mailbox providers receive your mail: how much went out and what happened to it, where your mail goes and why it doesn't arrive, which addresses sent it and what needs fixing.
Everything comes from the reports of the MTAs themselves on every server, and every number sits next to the previous period of the same length.
01 / Deliverability score
Email sending analytics starts with a single score out of 100. It's made of five parts: deliverability, spam and complaints, mailbox provider requirements, infrastructure and blacklists. The line under the score names the part that's pulling it down and what's wrong with it.
What's pulling the score down is blacklists: one address on one list
A bare “28 thousand sent” tells you nothing, while “+1% vs. the previous day” does. That's why every number sits next to the same number for the previous period of the same length.
Seven complaints — 0.024%, and together with spam bounces — 0.03%, against a 0.3% threshold. Under each number is a small chart of the period: whether sending went steadily or in bursts.
02 / How shares are counted
If a message was deferred three times and delivered on the fourth try, it's one message, not four. So delivery, bounce and complaint shares are counted from messages whose outcome is already known, and retries don't distort deliverability. Changes in shares are shown in percentage points.
At night 150–200 messages an hour go out, during the day about two thousand. In counts you see how much went out, in percentages — how well: delivered holds at 98–99%.
The current hour isn't over yet — its band is shaded and captioned “hour in progress”. If an hour has no messages yet, it has no share: the line breaks off instead of dropping to zero.
Bounced — 1.6% (451 messages), while deferrals are 721: the mailbox provider asked to try again later, and the message is still being retried.
03 / Providers, bounces, addresses
Mailbox providers, bounce reasons and sending addresses are shown as bars: the length is the share of sending, the color is deliverability. So deliverability by provider and by IP address fits on one screen. You see right away whether the whole problem is one address or one provider.
98.8% delivered; Microsoft — 21% and 98%, Apple — 11% and 98.9%.
Out of 451 bounces; mailbox full — 13%, abandoned — 9%.
In yellow — 198.51.100.126; the rest are at 98–99%.
The mailbox provider is determined by the recipient domain from a set list. Corporate mail that is also hosted by Google goes into “Other”.
04 / The problem list
“Needs attention” collects only problems, on its own: an address on a blacklist, a disk filling up, addresses without reverse DNS, servers without reports. Successful deliveries don't go here. Every row leads to where the problem is solved, and the last block shows which servers the numbers came from.
192.0.2.101 is on a blacklist, the mq-01 disk is 81% full, two addresses out of 28 have no reverse DNS, and lab-01 and mta-11 haven't sent a single MTA report.
Message data is coming from 15 of 17 servers, the latest batch 1–6 minutes ago. There have been no events from lab-01 and mta-11 yet: while a server is silent, its messages don't reach the analytics.
A server's load is compared only with servers on the same MTA: MailerQ and PowerMTA load a machine differently.
05 / The interface in detail
The score, four figures, trends in counts and shares, mailbox providers, bounce reasons, addresses, notes and telemetry intake. Above each screenshot: why the screen matters. Below it: what it shows.
The first four screens answer “how are things overall”. Here you have the deliverability score out of 100 and its five parts, the period's four figures next to the previous period, and two trend charts — in counts and in shares. You pick the period at the top: a day, a week or 30 days.
The first analytics screen is the overall sending score for the chosen period. In the example 28,585 messages went out in a day, and the score is 92 out of 100, “healthy”. Next to it are the five parts with their points: deliverability, spam and complaints, mailbox provider requirements, infrastructure and blacklists. Look at the lowest part: the line under the score names exactly what's pulling it down — here it's blacklists, where one address landed. In a few seconds you know whether everything is fine and where to look first. If the score dropped, compare its parts first: if one sagged, that's the cause, and details on each part are in the blocks further down this same page.
EnlargeThe score is calculated the same way as on the “Deliverability” tab of Monitoring.
The period's four main figures: sent, delivered, bounced and complaints. Each sits next to the same number for the previous period of the same length, so you see where sending is heading: in the example there are 1% more messages, and the bounce share is 0.1 percentage points lower. Under each number is a small chart — whether sending went steadily or in bursts. Complaints are compared with the 0.3% threshold right away, so it's clear whether you have headroom. One number answers two questions at once: how much, and is it better than last time. Open this row right after the score: if delivery fell and bounces grew, look at the mailbox providers and bounce reasons below next.
EnlargeChanges in shares are in percentage points, not percent.
A chart of sending by hour in counts: how many messages were sent, delivered, deferred and bounced. It shows the rhythm of your mailings — in the example 150–200 messages an hour go out at night, about two thousand during the day. Hours without messages are shown as zero, not smoothed over by a line, so real downtime is visible right away. The current hour is captioned “hour in progress”: few messages in it means incomplete data, not that sending stopped. The chart makes it easy to check whether the load matches your mailing schedule. If bounces or deferrals went up in some hour, look at the next screen — the chart in shares: it shows how serious that is relative to the volume.
EnlargeA zero in the current hour doesn't look like sending has stopped.
The same hourly chart, but in percent: what share of messages was delivered, deferred or bounced. Counts show how much went out, shares show how well: in the example delivered holds at 98–99% every hour, with bounces and deferrals pressed to the bottom edge. If an hour has no messages yet, the line breaks off instead of dropping to zero. A night with a hundred messages and a day with two thousand look the same in percent, so look at both charts together. A dip in share during a night hour with a hundred messages means less than the same dip during the day — so check it against the chart in counts. Look for the causes of a dip on the mailbox provider and bounce screens below.
EnlargeA night with a hundred messages and a day with two thousand look the same in percent — look at both charts.
The next screens show exactly where the problem is: which mailbox providers your mail goes to, why messages don't arrive and which IP addresses sent them. All three work the same way: the bar length is the share of sending, the color is deliverability. They show whether a problem affects one address, one provider or the whole mailing.
Deliverability by provider: which mailbox providers your mail goes to and how each one receives it. The bar length is the share of sending, the color is deliverability. In the example most messages go to Google — 38% with 98.8% delivered, while Microsoft, with 21% of messages, has lower delivery, 98%. If one provider lags noticeably behind the rest, go to the “Providers” tab and dig into it separately. The provider is determined by the recipient domain. So corporate mail hosted by Google lands in the “Other” row, not in Google. The seventh row, Proton, is visible only on the “Providers” tab.
EnlargeOne more row — Proton — on the “Providers” tab.
Bounce reasons for the period — the way the mailbox provider itself named them. In the example there were 451 bounces in a day, and 63% of them are “no such mailbox”, another 13% are a full mailbox and 9% an abandoned one. All the bars are red: a bounce is a lost message, whatever the reason. But the reason tells you what to do with the address next, and the full breakdown of bounces and deferrals is on the “Queues & errors” tab. The bounce count here matches the one in the “The period's four figures” block. With “no such mailbox”, as for 63% of bounces in the example, resending is pointless — such addresses get cleaned out of your list.
EnlargeAll the bars are red: a bounce is a lost message. But the reason tells you what to do with the address next.
Sending addresses — which IPs the mail went out from and how mailbox providers received messages from each. 26 addresses worked during the day; the screen shows the top six by volume: the leaders have 99.3–99.5% delivery. Highlighted in yellow is 198.51.100.126 with 93.6% delivery, and the bar shows that the problem is that one address, not the whole sending. This way one bad address doesn't hide inside the average deliverability of the whole mailing. If you find a yellow address, open “Needs attention” below: it says which blacklist it's on, and there's a “Go to address” button. The “Addresses & reputation” tab has twenty more addresses with reverse DNS and warm-up stage.
EnlargeTwenty more addresses — on the “Addresses & reputation” tab, with reverse DNS, blacklists and warm-up stage.
At the end — a list of what to fix and a check of the data sources. “Needs attention” collects only problems and leads to where they're solved. “Telemetry intake” shows which servers send MTA reports and when the latest ones arrived: without it you can't be sure of the numbers above.
The “Needs attention” list is what needs fixing, without successful deliveries and extra noise. In the example there are five notes: the address 192.0.2.101 is on a blacklist, the mq-01 disk is 81% full, two addresses out of 28 have no reverse DNS, and lab-01 and mta-11 haven't sent a single MTA report. The buttons on the right narrow the list to delivery, reputation or servers. Every row leads to where the problem is solved — this list is a handy way to start your day. Notes fall into delivery, reputation and servers: if you're responsible for only one of these, keep your part with the button on the right. For example, “Go to server” for mq-01 takes you straight to the server whose disk is filling up.
EnlargeEvery row leads to where the problem is solved.
Telemetry intake shows where all the numbers above came from: which servers send MTA reports, how many events were received from each, and when the latest batch arrived. In the example 15 of 17 servers are sending data, the latest from one to six minutes ago, and there have been no events from lab-01 and mta-11 yet. Check here when the figures look odd: while a server is silent, its messages don't reach the analytics and the totals are understated. You see the event count for each server: the transactional tx-01 and tx-02 sent 143–145 thousand each, the MailerQ servers 108–110 thousand each, the rest 16–18 thousand each.
EnlargeWhile a server is silent, its messages don't reach the analytics, and the numbers above are understated.
06 / Questions about analytics
In short: how the score and shares are calculated and why the numbers may be understated.
EnlargeA score out of 100, five parts with points and a line saying what's pulling the score down.
Five parts: deliverability, spam and complaints, mailbox provider requirements, infrastructure and blacklists. The line under the score names what's pulling it down.
A message deferred three times and delivered on the fourth try is one message. Otherwise retries would distort the shares.
The difference between shares: it was 98.4%, now it's 98.3% — a drop of 0.1 percentage points, not 0.1 percent.
From the reports of the MTAs themselves — PowerMTA and MailerQ — on every server, rolled up by hour and by day.
If a server isn't sending reports, its messages don't reach the analytics. The “Telemetry intake” block says so plainly.
Into “Other”: the mailbox provider is determined by the recipient domain from a set list, and corporate domains aren't on it.
It isn't over yet: the band is shaded and captioned “hour in progress”, so incomplete data doesn't look like a crash.
A day, a week or 30 days. Every number is compared with the previous period of the same length.
07 / Next step
The overview answers the main questions. The details are on the analytics tabs: mailbox providers, queues and errors, addresses and reputation, servers.
If a number in the overview looks worrying, a tab shows it by each mailbox provider, reason, address or server.
