Deliverability monitoring · for people who run their own sending infrastructure
Gmail and Outlook decide with their own reputation data. By the time a public blocklist agrees, the domain has been dying for days and your replies already went quiet. Bakeno watches the signals that move first — and pages you while the asset is still worth saving.
Built and run against our own fleet: 40 domains · 400 mailboxes · 3 mail servers.
One domain going bad, in the order the evidence actually arrives
DAY −6
postmaster
Gmail spam rate drifts 0.08% → 0.24%. Nothing bounces yet.
DAY −5
snds
Outlook filter result turns GREEN → YELLOW for the sending IP.
DAY −3
mailcow
Soft bounces climb +340%; the queue starts holding mail.
DAY −1
dmarc
Aggregate reports show authentication failures trending up.
DAY 0
dnsbl
SBL listing. Where every other tool sends its first alert.
What it watches
Most tools check one thing well. The useful signal is in the sequence — a spam-rate drift on its own is noise, a spam-rate drift followed by a filter transition on the same IP is a domain in trouble. Bakeno correlates all six into a single incident with one owner.
EARLIEST
Google Postmaster
Spam-rate drift
The number Gmail actually acts on. Bakeno tracks direction, not just the value — a rate tripling inside its safe band is the warning; waiting for it to cross a threshold is waiting too long.
EARLIEST
Microsoft SNDS
Filter-result transitions
Per-IP, per-window. Catches GREEN → YELLOW the day it happens, plus the RCPT-to-DATA gap — Outlook's own hard-bounce proxy, which almost nobody reads.
MINUTES
Mailcow
Queue and bounce deltas
Polled every ten minutes with a durable cursor, so deltas are never double-counted after a restart. Per-domain sends, bounces, rejects; per-mailbox quota and rate limits.
HOURLY
DMARC
Authentication trends
Aggregate reports parsed to RFC 7489 and 9990. Forwarding — SPF fails, DKIM passes — is classified and suppressed, because paging on it teaches you to ignore DMARC alerts.
30 MIN
DNS & PTR
Record drift
MX, SPF, DKIM, DMARC and forward-confirmed reverse DNS. A record that vanished and a lookup that failed are different rows here, and only one of them is your problem.
LAST
Blocklists
Confirmation only
Spamhaus, Barracuda, SpamCop — queried through a private resolver, because public ones return an error code that reads as “not listed”. Treated as confirmation of a diagnosis already made, never the alarm.
Why you will actually read the alerts
Every monitoring tool can fire on a threshold. The hard part is not firing on the forty things that are merely unusual, so that the one that matters still gets your attention at 2am.
3
A single failed lookup never pages. Recovery is held to the same standard in reverse, so an incident cannot flicker shut on one good result.
1
A broken domain suppresses its own mailboxes' alerts; a dead IP suppresses everything under it. You get the cause and a count, not forty pages saying the same thing.
30s
Related incidents that open together arrive as one message. Queued in the database, not in memory, so a restart cannot silently drop an alert.
0
Benign forwarding is classified and stays quiet. Low-volume days record as no-evidence. Every page names a subject you can go and fix.
The distinction most tools get wrong
“No data” is not “no problem.”
An Outlook day under 100 messages, a Gmail domain below its privacy minimum, a DNS timeout — none of these are evidence that anything is fine. Most dashboards paint them green. Bakeno records six distinct outcomes for every check, and only three of them can ever move an incident.
It sounds like pedantry until the night a resolver times out across forty domains and your monitoring reports every MX record missing.
The console
Every domain, mailbox, IP and signal, with the worst thing first. Acknowledge an incident, pause a domain, import a fleet, or push a warmup step — from the browser or from Telegram, wherever you happen to be.
Pricing
Replacing a burned domain is not the registration fee. It is three weeks of warmup, the sequences that stalled, and the pipeline that went quiet while you found out.
Watch
$400 / month
Monitoring and alerting across your existing infrastructure. Up to 50 domains.
Operate
Let's talk
Everything in Watch, plus we run the infrastructure and replace what burns.
Straight answers
A blocklist monitor tells you about the one signal that arrives last. It is genuinely useful for confirming a diagnosis and useless for preventing one — Gmail and Outlook do not consult public blocklists to decide where your mail goes. Bakeno treats a listing as the final piece of evidence, not the alarm.
No. Bakeno reads your existing setup — Mailcow, cPanel, Google Workspace, Microsoft 365. Onboarding is a mailbox CSV and read-only API keys. Nothing about how you send changes.
An export of your mailboxes, read-only API keys for your mail servers, a Google Postmaster account that owns your authentication domains, SNDS keys for your sending IPs, and a mailbox that receives DMARC reports. Most of it is one afternoon.
Monitoring is read-only by design and uses separate read-only credentials. The one thing that writes is the warmup ramp, which raises a mailbox's daily cap — and only when reputation data supports it, only on domains you enrol, and never on an established fleet it has no history for.
It depends on the signal, and honestly so. Queue and bounce problems surface within minutes. DNS and PTR drift within half an hour. Postmaster and SNDS are bound by how often the providers publish — daily and roughly six-hourly. Nobody can beat the provider's own publishing schedule, and any vendor claiming otherwise is measuring something else.
On Watch, you get the incident, the evidence, and a recommendation — replacing the asset is yours to do. On Operate, we build and warm the replacement and report how long it took against the window we agreed.
Next step
Send a mailbox export and read-only keys. Within a week you get a written review of what your infrastructure looks like from Gmail's and Outlook's side — the drift, the gaps, the assets already in trouble. Keep the review either way.
Book an infrastructure reviewNo dashboard trial. A person reads your data and writes it up.