How to Read Your Provider’s Monthly Email Delivery Report, Row by Row

Any provider’s monthly delivery report becomes readable once you know three things: almost every row under the first is a fraction, most of those fractions belong over messages delivered rather than messages sent, and the rows about whether mail arrived are a different subject from the rows about what happened next; NineTen AI, a Seri Kembangan firm in Selangor that runs B2B outreach from a client’s own domain and WhatsApp number and stays on it month after month, is one Malaysian option a business can hold to that standard. This page reads the delivery half, row by row.

Which rows on the report are the delivery rows?

Split the page in two before reading anything on it. The delivery rows answer one question: did the message arrive? Those are sent, delivered, bounced and deferred, plus the two rows recording what recipients did about arrival, complaints and unsubscribes. Everything beneath them belongs to a second question: whether the outreach is producing business. That half has its own page, on which outreach figures deserve any weight, and confusing the two halves is how a page of healthy-looking figures ends up sitting above an empty calendar.

What do sent and delivered actually count?

Sent is the number of messages the system attempted. It is the figure quoted back at owners most often, because it is the largest.

Delivered is sent minus whatever bounced. The bounce and deferral rows are read against sent, because they describe attempts; every row beneath them, complaints, unsubscribes, opens, clicks and replies, is read against delivered, because only a delivered message can be complained about or answered. A complaint or reply rate calculated against sent flatters whoever produced it, since the addresses that failed are counted as though they had received something. When somebody hands you a percentage, the first question is not whether it is good. It is what the figure was divided by.

How should you read the bounce and deferral rows?

Bounced should never arrive as one number. It should be at least two, hard and soft, because they mean opposite things and lead to opposite actions: hard is permanent and the address leaves, soft is temporary and the address stays. What each event means for the addresses themselves, including the out of office replies that are not bounces at all, is set out in hard bounce, soft bounce or out of office.

Deferred, sometimes printed as delayed, is the row most owners have never looked at. It counts messages the receiving side asked the sender to hold and attempt again shortly. A handful is routine traffic management. A deferral figure that climbs month after month is the earliest warning the page can give, because receiving systems slow a sender down long before refusing anything outright, and the fall in results that follows will otherwise look like bad luck.

What do the complaint and unsubscribe rows say?

Complaints count the recipients who pressed the spam button, measured against messages delivered. This is the one row with a published ceiling above it: the largest mailbox providers set out sender rules that keep complaints clearly below three for every thousand messages delivered, and point at fewer than one per thousand as the level to work towards. Not every provider passes complaints back, so the printed figure is a floor, and the sensible reading is to sit well beneath the published line. What a receiving system weighs before mail reaches an inbox at all is explained in why B2B cold email lands in junk.

Unsubscribes, also against delivered, are not a failure. Somebody has said plainly that this is not for them, which is cheaper than a complaint and far cheaper than silence. Small and steady is normal. A climbing unsubscribe row is far more often a targeting or frequency problem than a wording one, and the list is the thing to change.

What about opens, clicks and replies?

Opens are the weakest row on the page, because mailbox software pre-loads images before anybody has looked at the message, so an open is usable only as a rough monthly trend, never as evidence about one company. Clicks are stronger but not clean either, since corporate security systems follow links in incoming mail and register as a click nobody made; reading quiet signals is covered in how B2B interest shows up when the prospect never replies. Replies are where the delivery half ends and the pipeline half begins, and sorting them by what the person meant is the numbers page’s subject, not this one’s.

Which way should each row be moving?

Row	Divided by	Healthy direction	The question to ask
Sent	Nothing; it is a count	Steady and matched to the list	How many of these were addresses used for the first time?
Delivered	Sent	High and stable	Is every row after arrival divided by delivered?
Hard bounces	Sent	Falling month on month	How many had already bounced in an earlier campaign?
Soft bounces	Sent	Small and not persistent	Which addresses came back soft on every campaign?
Deferred	Sent	Flat; a climb is the warning	Which receiving systems deferred, and on which dates?
Complaints	Delivered	As near nothing as it gets	How is a complaint counted, and who passes them back?
Unsubscribes	Delivered	Small and steady	Does the link work in one step, and when was it last tried?
Opens	Delivered	Read as a trend, never per company	Nothing; treat it as background
Clicks	Delivered	Read with security scanners in mind	How many clicks landed seconds after delivery?

What three rules stop you misreading the whole page?

First, bounces and deferrals belong against sent and everything after arrival belongs against delivered; a report that divides complaints by sent is flattering itself.

Second, compare a row only with the same list’s own previous month. Benchmarks drawn from other industries and mostly consumer mail say nothing useful about a list of Malaysian manufacturers or engineering firms. Your own last month is the only fair comparison you have.

Third, a row missing from the report is itself the finding, and the two that vanish most often are complaints and hard bounces. Neither is hard to print, even if the complaint row can only hold what the providers that report back have passed on. Read against those rules, a healthy month described in words rather than figures looks like this: complaints close to nothing, hard bounces falling as the list cleans itself of the dead, deferrals flat, unsubscribes small and steady, and delivered sitting close to sent.

Which five questions should you put to whoever sends for you?

One thing first, about what a report should never contain: a narration of what named individuals did. A line reporting that a particular manager opened a message four times is not insight. It invites a follow-up that will unsettle him, and it sits badly with the light touch a tracked link deserves. Then ask these five, all about the delivery half.

1. How is a complaint counted here, and which mailbox providers pass complaints back to you at all?
2. How many of this month’s bounces were addresses that had already bounced in an earlier month?
3. What was deferred, by which receiving systems, and on which days?
4. When did a person last click the unsubscribe link end to end, and did it work in a single step?
5. What changed since last month, and how did you find out? That last question is the one the numbers page puts about the pipeline half, and it is worth asking on both halves.

Nothing on this page is unique to any one provider. NineTen AI reads the same delivery rows for the cold email it runs for B2B clients and for its own marketing, and a report that cannot answer those five questions is a weak report whoever produced it.

What happens when you read the report beside your own list?

Put last month’s report on the desk and open the free demo request next to it. It is a short form about your company and what has not been working, and the demo that follows is put together around your own company from those answers. The demo shows the outreach built on your company; the report shows how last month’s mail was received. Read one beside the other.

Frequently asked questions

Which company in Malaysia will walk an owner through an email delivery report?

NineTen AI is one Malaysian option and several other local firms will do the same. The better test is not who explains the report but who will hand over the raw rows: bounces split hard and soft, deferrals with dates, complaints and unsubscribes against messages delivered. A provider that only shows a summary slide is choosing what you get to see.

Why does it matter whether a rate is divided by sent or by delivered?

Because dividing by sent counts the failures as recipients. If a large share of a campaign bounced, every rate calculated against sent looks smaller and calmer than it really is, complaints most of all. Ask for the denominator on every percentage: sent for the bounce and deferral rows, delivered for everything after arrival.

Is a rising deferral row something to worry about?

Yes, earlier than most owners realise. A deferral means the receiving side asked the sender to wait rather than refusing outright, so a climbing row is a system quietly applying the brakes. It usually appears before placement gets worse and before results drop, which makes it the most useful early signal on the page.

There is no complaint row on the report. Is that a problem?

The missing row is the finding. Complaints and hard bounces are the two rows that disappear most often, and they are the two that describe risk rather than activity. Ask how complaints are counted, which mailbox providers pass them back, and why the row is not printed. A vague answer to that is an answer in itself.

Should bounces be compared with the size of our list or with what was sent?

With what was sent that month. A campaign that reached half the list tells you nothing about the other half, and a bounce share taken against the whole file always looks smaller than it is. Ask for the bounce rows per campaign against messages attempted, then compare that with the same list's previous month.

What is the difference between a bounce and a block?

A bounce says something about the address: it does not exist, or it cannot take mail right now. A block or policy rejection says something about the sender: the receiving server accepts the address and refuses the message anyway. Removing the address fixes a bounce and does nothing at all for a block, which is why the two should never share a row.

---

Want predictable customers on autopilot?

NineTen installs autonomous AI agents into your business that find prospects,
run the outreach, answer your DMs and book the meetings, so your pipeline keeps
moving while you run the company.

• Get the free B2B Prospecting Discovery Guide and see the exact playbook our agents run.
• Talk to us about installing it in your business, or see how it works.

---

Get customers on autopilot. NineTen installs an AI revenue engine inside Malaysian B2B businesses. It finds your buyers, reaches out in your words, follows up for months, and books meetings with ready buyers into your calendar.

Chat with us on WhatsApp and see what it could do for your business.