> ## Documentation Index
> Fetch the complete documentation index at: https://notifizz.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Deliverability — Fifteen Levers, and Who Pulls Them
> The full list of what decides whether your email reaches the inbox: which levers Notifizz operates for you, which three are yours, the spam score, reputation monitoring and the automatic brake on marketing sends.
# Deliverability
Deliverability is not a setting. It is a dozen small guarantees that have to hold at once, most of which have nothing to do with the words in your email. **Settings → Email → Deliverability** lists all of them on one screen, each with its real state for your account — so the question stops being "are we doing the right things?" and becomes "which row is amber?".
**Written for all three roles.** Marketing needs to know what is already handled and what is not; Dev owns two of the three configurable rows; Product/Ops reads the reputation half of this page. The screen itself is deliberately readable by all three.
## TL;DR
* Fifteen levers, on one page. **Nine are active by construction** — they are how the platform works, with nothing to configure. **Three are yours to set up.** **Three are not shipped or are on request.**
* The score at the top reads *N levers active or configured out of 12* — the three unavailable ones are excluded rather than counted against you.
* Every campaign is **scored for spam risk while it is still in review**, on the real rendered content, plus structural checks the content scanner cannot see. The score informs; it never blocks.
* Your **complaint rate is evaluated continuously** against the published bulk-sender thresholds: **0.10 %** raises a warning, **0.30 %** is critical.
* At the critical level, **marketing sends stop automatically** and resume on their own. **Transactional mail is never affected.**
* Your sending domain is **scanned against the major anti-spam blocklists** several times a day, and a broken authentication record is raised as an incident.
## The fifteen levers
### Nine that are already on
These need no configuration. They are the platform's behaviour, and they are listed rather than hidden precisely because you should be able to tell a prospect, an auditor or a worried CMO what is actually in place.
| Lever | What it does |
| -------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Separate transactional and promotional streams** | Your confirmations and password resets travel separately from your marketing mail, so a bad complaint week on marketing cannot contaminate the mail your users are waiting for. The [category](/docs/concepts/campaigns#campaign-categories) is classified by AI when a campaign is created, with a protective bias: at the slightest doubt the campaign goes promotional, because a wrong *transactional* call is the only dangerous error. |
| **Sends in adapted steps** | Bulk sends are split into paced steps instead of fired in one block, with a cautious regime on a recently verified domain or a first high-volume send. See [pacing](/docs/concepts/broadcasts#pacing). |
| **One-click unsubscribe and managed link** | The one-click unsubscribe header and the footer link are injected on promotional mail automatically, and opt-outs take effect immediately. Since 2024 this is a requirement from the major mailbox providers for volume senders, not an option. |
| **Bounces and complaints handled, addresses suppressed** | Every hard rejection and every spam report is consumed continuously and removed from future sends. Continuing to write to a dead address, or to someone who reported you, is one of the costliest signals a sender can emit. |
| **Promotions restricted to consenting contacts** | Promotional sends only reach contacts whose consent is on record; transactional mail follows its own legal basis. The check runs per recipient at send time, so an opt-out registered a second earlier is honoured. |
| **Spam detection before sending** | A content score plus structural checks on every campaign before it goes out. [Detailed below](#the-spam-score). |
| **Sending infrastructure with policed reputation** | The shared sending infrastructure Notifizz uses vets senders at the door and offers no free high volume, which is what makes sharing viable in the first place. |
| **Shared domains monitored** | The domains placed inside your email bodies are checked against the major anti-spam lists several times a day; a flagged domain is swapped for a backup within minutes, with nothing to do on your side. The last check time is shown on the row. |
| **Your reputation, watched** | Complaint rate, rejections and domain health, evaluated continuously per mailbox provider, with an automatic brake at the critical level. [Detailed below](#reputation-monitoring). |
### Three that are yours
| Lever | Where | Notes |
| -------------------------------------- | -------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Sending domain authentication** | [Settings → Email → Domains](/docs/operations/sending-domain) | The one that matters most. Amber until a domain reads *Verified*; the row then shows which domain. |
| **Custom content and tracking domain** | [Settings → Email → Domains](/docs/operations/tracking-domain) | Available on the top plan. Without it you use the shared domains above, which Notifizz monitors and rotates itself. |
| **Notification categories** | Settings → Unsubscribe categories | Lets a tired recipient turn off one kind of mail instead of all of it. Without categories, their only choices are unsubscribing from everything or reporting you as spam. The row shows how many you have defined. |
### Three that are not levers you can pull today
| Lever | Availability |
| ----------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Notification pressure** | Announced, not shipped. A cap on how many messages one contact receives over a period, counted across email, notification centre and push together. |
| **Dedicated IP address** | On request. Relevant from roughly a million emails a month — below that it actively hurts, because a low-volume dedicated IP never accumulates enough history to be recognised. |
| **Tailored deliverability support** | On request. Configuration audit, preparing a volume ramp-up, incident handling, brand logo in inboxes (BIMI). |
These three are shown with a plan or availability chip and **excluded from the score**. A screen that counted an unshipped feature as a failure would be lying to you every day.
## The spam score
Every email produced by a **non-live run** — a review simulation, a test — is scored automatically, on the message exactly as it would ship. The verdict appears on the run panel next to that message, labelled *Spam risk*. Live sends are not re-scored: by then the question has been answered.
### What it measures
Two independent things, deliberately.
**A content score**, computed with SpamAssassin, the reference open-source filter. It reads the rendered message exactly as it would ship. Bands:
| Band | Promotional campaigns | Transactional campaigns |
| -------------------------------------------- | --------------------- | ----------------------- |
| Green | below 2 | below 2.5 |
| Orange — worth a look | 2 and above | 2.5 and above |
| Red — content filters would classify as spam | 5 and above | 5 and above |
The transactional cut is higher for a reason: a short receipt (little text, a logo, a tracking pixel) legitimately measures around 1.8. That is the normal shape of a receipt, not a risk, and alerting it would train you to ignore the indicator.
**Structural checks**, which are data-independent and catch what a content score cannot:
| Finding | Why it matters |
| ------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Shortened link | A shortener hides the destination from the reader *and* from filters, and its domain carries every other user's reputation. |
| Almost no text for the images present | Filters cannot read the message, and it looks empty to anyone who blocks images. Promotional campaigns only — a short receipt is legitimately text-poor. |
| Subject in capitals, or several exclamation marks | A classic promotional signal — checked on promotional campaigns only. |
| Message over 102 KB | Gmail clips it and hides everything after the cut, unsubscribe footer included. |
A structural finding lifts a green verdict to orange. It never reaches red — red stays the content filter's own verdict.
### What it does not measure
The score reads **content only**. Your domain reputation, your IP reputation and the live blocklist state are not in it — those live on the Reputation page. A green score on a domain with a 0.4 % complaint rate means the copy is fine and the account is not.
**The score never blocks anything.** It is informational by design: a red verdict is a strong reason to rewrite, not a gate. You can turn scoring off entirely for your organisation in **Settings → Email → Sender**, under *Review spam score*. It is on by default.
## Reputation monitoring
**Settings → Email → Reputation** answers a question no single number can: how do the mailbox providers currently see you? Notifizz evaluates it continuously and states a verdict rather than leaving you to interpret raw counts.
### What is measured
* **Complaint rate over a rolling 7-day window** — spam reports divided by delivered mail.
* **A per-provider breakdown** over 7 and 30 days — Gmail, Outlook & Microsoft, Yahoo & AOL, Apple, and others — with delivered, bounces and complaints for each. Daily counters cover closed days, so the table runs through yesterday.
* **Volume anomaly** — a sudden spike relative to the preceding week. A brutal ramp looks like a compromised account to a receiver, even when it is a legitimate launch.
* **Provider blocks** — the share of bounces that are reputation or policy refusals rather than dead mailboxes. This is usually the *first* visible symptom of a degradation, well before any spam rate moves.
* **Blocklist listings** on your sending domain, from the several-times-a-day scan.
* **Authentication drift** — a previously verified sending domain that no longer authenticates. See [sending domain](/docs/operations/sending-domain#after-verification-it-keeps-being-checked).
### The verdict
| Level | Trigger | What it means |
| ------------------- | --------------------------- | ----------------------------------------------------------------------------------------------------------------------------- |
| **Healthy** | Complaint rate below 0.10 % | Nothing to do. |
| **Needs attention** | 0.10 % and above | Nothing is blocked. Space out sends to your least engaged recipients — that is the single gesture that brings this rate down. |
| **Critical** | 0.30 % and above | Marketing sends are paused automatically. See below. |
Those two thresholds are not ours: they are the numbers Gmail and Yahoo published with their 2024 bulk-sender requirements. Using anyone else's would leave you compliant with Notifizz and non-compliant with the people who actually deliver your mail.
**Below 100 delivered messages in the rolling window, the rate is not treated as a verdict.** One complaint out of thirty is a 3.3 % rate and means nothing at all — so the level stays *Healthy* and the rate is shown for information only. And until you have sent anything, there is no verdict at all: the page says so rather than showing a reassuring green.
**Gmail is a blind spot that has to be closed separately.** Gmail does not forward individual spam reports to senders, so our own complaint rate cannot see them. Linking [Google Postmaster Tools](/docs/operations/gmail-postmaster) gives you the rate Gmail actually measures — and that rate can *raise* your Notifizz reputation level, including to critical. It never lowers it.
## The automatic brake on marketing sends
This is the part worth reading before you need it.
When the complaint rate reaches the critical threshold, Notifizz **stops marketing sends for the organisation by itself**. No ticket, no human in the loop, no waiting for someone to notice.
What that means concretely, per path:
| Path | Behaviour under the brake |
| ------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------- |
| A [broadcast](/docs/concepts/broadcasts) you launch manually | Refused, with an explicit reason. Nothing partial leaves. |
| A broadcast scheduled for later | Deferred, not failed. It stays scheduled and is retried on the next pass, so it goes out on its own once the brake releases. |
| Continuous, segment-driven campaigns | New journeys are not started while the brake holds. |
| **Transactional mail** | **Unaffected.** It runs on a separate path, and a password reset is not the thing to hold back because a newsletter went badly. |
**Release is automatic, and deliberately not symmetric.** The brake engages at 0.30 % but only releases once the rate falls back **below 0.20 %**. Without that gap the brake would flicker on and off around the threshold as feedback trickles in — stopping and restarting a mailing repeatedly, which is worse for your reputation than either state.
The brake is visible as a banner on the Reputation page, and a launch attempted during that window tells you why it was refused rather than queueing silently.
A brake is a symptom, not a cure. It buys you the time to fix what caused the complaints — usually a stale list, an audience that never really opted in, or a send frequency nobody signed up for. It will re-engage on the next campaign if the underlying list is unchanged.
## What to do at each level
| Situation | The useful move |
| ------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------- |
| Needs attention, no other signal | Reduce frequency to your least engaged segment. Complaint rates fall faster from sending less than from writing better. |
| Provider blocks detected | A provider is filtering you. Check your recent content and the freshness of your lists — these bounces are refusals, not bad addresses. |
| Volume anomaly | If the spike was intentional, ramp in stages over a few weeks instead. If it was not, find out what fired. |
| Sending domain listed on a blocklist | Request delisting from the list that carries it, then find the send that put you there — otherwise the next one repeats it. |
| Authentication drift | Restore the DNS records immediately. Nothing else on this page matters while receivers cannot verify your mail. |
| Critical / brake engaged | Clean the list before doing anything else. The brake releases on its own; sending the same list again re-engages it. |
## FAQ
Because the maximum is 12. The three levers that are unshipped or on request are excluded from the count rather than held against you. A 12 of 12 means everything available to your account is in place.
Expected, and the two are not in contradiction. The score reads content; placement is decided mostly by reputation, authentication and recipient engagement — none of which content scoring can see. Start on the Reputation page, then check your [sending domain](/docs/operations/sending-domain) records and, for Gmail recipients, [what Gmail sees](/docs/operations/gmail-postmaster).
Not from the dashboard, no. The brake exists because a critical complaint rate damages the same sending reputation every one of your future campaigns depends on, and because the account emitting the complaints is rarely the one best placed to judge in the moment. If you believe the measurement is wrong for your account, talk to us — that conversation is the intended path.
No. Nothing about the score is a gate — it is shown so a human can decide. What does block a live send is the absence of a verified sending domain and a default sender.
Because those counters aggregate closed days. A partial current day would make every rate look different depending on the hour you opened the page. The continuous verdict at the top of the screen does include today.
Not yet, and that is honest rather than flattering. Below one hundred delivered messages in the rolling window the rate is not statistically meaningful, so the level stays *Healthy* and the number beside it is informational. The signals that do not depend on volume — a blocklist listing, a broken authentication record — are raised at any volume.
They are part of the measured traffic, yes — a spam report on transactional mail is a real and serious signal, and it suppresses that address permanently at the provider. What transactional mail is exempt from is the brake: it keeps going even when marketing is paused.
## See also
The records that authenticate your mail, and what breaks when they lapse.
Close the one blind spot our own complaint rate cannot cover.
Move the links and images inside your emails onto your own domain.
Pacing, stopping, and how the brake behaves at launch time.