Analytics

Analytics rolls every send, open, click, reply, bounce, and complaint into rates you can act on, by workspace, campaign, and mailbox. For health bands and the thresholds Warmbly acts on, see Deliverability.

Campaign analytics are shared across the workspace. With View analytics, you can see the overview, daily and hourly statistics, step performance and comparisons for campaigns created by any member of the selected workspace. The campaign author does not need to grant separate access.

What is counted

Seven rules govern the counts:

• Tracking must be on. Opens and clicks need open or link tracking enabled on the campaign, and a tracking host on the install. A custom tracking domain per mailbox is optional; without one they go through the shared host and still count. With tracking off they stay at zero while sends and replies still count.
• Reading in Warmbly does not count as an open. Unibox removes Warmbly's tracking pixels from the displayed message, including quoted history. This applies to sent messages and every other folder; recipient copies retain their pixels.
• Replies are human replies. Out-of-office and autoresponders never count, never stamp the contact as replied, and never trip stop-on-reply. A reply is attributed by the thread it answers, so a person who replies from a different address than the one you mailed (a "send as" alias, a forward, a colleague answering on their behalf) still counts as a reply for that lead. In a campaign that rotates across several mailboxes, each lead's reply is expected in the mailbox that wrote to them (the one their follow-ups come from too), so that mailbox has to be connected with inbox sync in the same workspace for the reply to be seen; a reply that reaches another mailbox in the pool stays in the Unibox and is not counted. Reply detection is always on; the reply-intent switch in Settings governs only what happens after a reply is detected. A reply that was missed is not lost: once a day, and whenever it starts, the platform re-checks recent inbound mail that was never attributed and marks the lead replied if it should have been.
• Bots are filtered. Crawlers, CLI agents, prefetches, chat link previews, and security gateways that announce themselves are served normally but never counted. The ones that do not announce themselves are caught by where they came from and by what they do. Requests from a known mail-filtering network are treated as automated whatever their user agent claims, and a click from one is still redirected, just without the ticket that would file the visit against the recipient. Microsoft 365's own filtering layer, where Safe Links and delivery-time link scanning run, is recognised out of the box, as is Barracuda's. Proofpoint, Mimecast and Cisco Secure Email are recognised too, with one difference: those three also offer browser isolation, which renders a clicked page in the vendor's cloud for the recipient to read, so a click from them can be a real person. A match there is treated as automated only when it also arrives soon enough after the send to be the delivery-time scan, and a recipient who clicks through isolation later still counts as a person. The rest are caught by behaviour: an open within a minute of the send, or a click within thirty seconds, and clicks on two or more links of one email from the same source within five seconds (a scanner walking the message). The clock on the first two starts when the send is handed to a worker, before the mail has even been delivered, so those windows cover the sending provider's queue and the transit to the recipient as well as reading time. A self-hosted instance can change those windows under Instance settings, including the one the isolation vendors are measured against, when its recipients' gateways are slower or faster than that. Otherwise one corporate scanner would "click" every link seconds after delivery. A self-hosted instance can name its own scanner networks; see tracking service configuration.
• Auto-opens are logged, not counted. Privacy proxies like Apple Mail Privacy Protection, and instant opens, are kept as proof of delivery and shown separately (12 auto, not counted), and an open counts only once per email however many times it is fetched. Every open count and open rate, on the campaign overview, per step, per day and per hour, and on the dashboard, counts a person's opens only, which is how the Leads tab already reads a lead: not opened until a person opens it. Apple Mail Privacy Protection's relay (a user agent of just Mozilla/5.0) and the stripped WebKit user agent that Mail on a Mac and the new Outlook for Windows both send are treated as a prefetch only when they arrive inside the automated-open window. Later, the same request is a person's open. Auto-opens never trigger opened-based branches or automations.
• A click is an open. Opens need the mail client to load images, and many clients block them. A click by the person proves the email was read, so it counts as an open as well, and a lead can never read "clicked, not opened".
• Auto-clicks are kept but never counted. A click classified as automated is logged on the contact's activity with the link it hit and the rule that caught it, and the campaign overview shows how many steps had only automated clicks (3 auto). It never makes the step clicked, never fires a clicked branch or automation, and never sends a webhook. A person clicking the same link later counts normally. A burst is only recognisable from its second click, so a click that looks human waits for the burst window plus one second (six seconds) before it fires anything (clicked branches, automations, webhooks, the live feed): if a second link follows inside that window, both clicks are relabeled, the stamp is withdrawn and nothing fires.

Workspace dashboard

Live, scoped by 7d, 30d, or 90d tabs.

Sending totals, trends, comparisons, and top-campaign rates count email steps only. Completed wait and action steps move a sequence forward but never inflate the number of messages sent.

Direct mail

The rest of this page measures campaigns. Direct mail measures what you write by hand in the unified inbox, which most workspaces do far more of than they expect.

It is built from two different sources, and they are shown apart because they cover different sets of messages.

Volume and replies come from your synced mailboxes, so they cover direct messages written in Warmbly, Gmail, or your phone and include history from before this feature existed. Synced campaign and warmup sends are excluded from the outgoing figures. Nothing has to be switched on.

A thread counts as yours when the first message in it was one you sent, and as answered when a message came back after it. The delay is measured from your first send to their first reply, so a long exchange is one data point rather than many.

Opens and clicks are opt-in and measured differently. They come from completed send records, so they only cover HTML mail sent through Warmbly from a mailbox with tracking switched on, and only from the moment you switched it on. Plain-text-only messages cannot carry a pixel or wrapped link and remain untracked. That is why the card states its own denominator rather than reusing "sent by hand" above: dividing opens by messages that were never tracked would invent a rate.

To switch it on, open the mailbox and turn on Track opens and clicks on direct mail. See Mailboxes.

Per mailbox ranks your mailboxes by how many messages each mailbox sent and received, and marks the ones that are tracking. Top correspondents ranks the people you exchange the most mail with. Your own mailboxes are excluded from that list, so a copy to yourself is not a correspondent.

Per-campaign

Each campaign reports its own totals: enrolled contacts, sent email steps, remaining planned email-step sends, unique opens and clicks, replies, bounces, unsubscribes, and the four rates. Wait and action steps do not count as messages.

Per-step stats break sent, opens, clicks, replies, and bounces down by sequence step, which is how you find the touch pulling replies and the follow-ups that are dead weight. Email steps are listed in the order the sequence runs them and numbered the way the canvas numbers them (an unnamed step shows as Email 2), so action steps are left out of the count. Each count carries the step's own rate underneath it, measured against that step's sends rather than the campaign's, so a fourth touch that reached 11 contacts still compares against a first touch that reached 337. A step that has sent nothing shows a dash rather than 0%, and a dotted rate means some of those events were automated: hover it for the split. Daily stats drive the trend chart, and hourly stats show when sends land and get engagement. Campaigns can also be compared side by side on the same metrics, over the same kind of period.

Reporting period

The Performance bar on a campaign's overview picks the period every figure on the page covers: 7d, 30d, 90d, All time or Custom (a first and a last day). The headline rates, the totals, step performance, the daily chart and who engaged all change together, and the shared image names the period under the campaign. All time is the default and charts from the day of the first send. The last preset you picked is remembered for every campaign.

A period is a send cohort: it selects the emails sent on its days and counts every open, click, reply and bounce those emails earned, whenever they arrived. An email sent on September 5 that is answered on September 10 counts as a reply in a September 1 to 7 period, and a reply that arrives on September 3 to an email sent in August does not. That keeps every rate a share of the same emails, so the reply rate of last week's sends is comparable with the week before, before and after you change a mailbox, a domain or the copy. A recent period's rates can still rise as late replies come in. Days are UTC calendar days, and the last day counts in full. Enrolled contacts and remaining sends describe the campaign as it stands now and do not change with the period.

Who engaged, and from where lists the countries, mail clients and devices people opened and clicked from, counting each contact once per bucket. The device column says how the email was read as well as on what: Mobile app, Desktop app, Tablet app, Webmail, or Hidden by image proxy when the mailbox provider loaded the image itself (see How an open was read). A click counts under the device it came from. Only human events feed it, so a security scanner's data centre never tops the country list.

What is recorded per open and click

Every open and every click is kept as its own record, repeats included, with what the request said about itself: the mail client, whether it was an installed app or webmail, the browser, operating system and device type (see How an open was read); the country, region and city resolved from the source network; for clicks the exact link; and whether it was the person or an automated fetch, with the rule that caught it. The address itself never leaves the tracking service: it publishes the network the address belongs to (the last IPv4 octet zeroed, an IPv6 address cut to its first 48 bits), which is enough for a city-level lookup and cannot single out a host, plus a keyed token of the full address used to tell one source from another, which cannot be turned back into the address. The consumer resolves the network to a location and stores neither. Location needs the GeoLite2 database on the consumer (GEODB_PATH); without it the country, region and city stay empty and everything else is still recorded. These records are kept for a year, then pruned; the counts and the first open and click per step live on the lead itself and are not affected.

How an open was read

Each open says what it was read on, as far as the request can tell: iPhone · Apple Mail app, Windows PC · Outlook app, Mac · Webmail in Chrome. The contact's Activity tab shows it on every open and click, the live feed on a running campaign and the dashboard's recent activity show it next to the contact, and the contact's Overview sums it up under How they read: each client and device their opens came from, with how many and how recently.

What the request can tell depends on the mail client:

A click names the browser the link opened in and the device it opened on, because the link leaves the mail client. Nothing is guessed: when the request cannot say something, the field is left empty rather than filled with the proxy's own details. Opens and clicks recorded before this was added are read again by the same rules in the background after the instance upgrades, a small batch at a time, so older history reads the same way once it finishes.

These records back the engagement rows on a contact's activity timeline and the per-campaign breakdown above.

Per-mailbox

The overview calls the configured percentage Reply target because it controls how much synthetic warmup mail should receive a reply. The Analytics tab reports the measured reply rate from actual sends and replies. Its Target met percentage is actual warmup sends divided by planned target volume across the active days in the selected range; it can exceed 100% when volume beats the plan.

Use this to find the one mailbox dragging down a campaign's bounce rate instead of blaming the campaign.

A/B analysis

Variants can be scoped to one sequence step or the whole campaign. Traffic splits across the active arms, and the step's own email is the Original control arm, itself a first-class arm with its own share, so 70% Original against 30% of a new subject is a valid test. Drag the split bar or type exact percentages; shares always total 100%.

A/B is on by default in outreach settings, but a campaign only runs an experiment with at least two active variants.

Assignment is deterministic per contact and step for step-scoped variants, so the same contact sees the same variant across retries. Campaign-level variants assign on first send by weighted random draw and keep the contact there. Pausing a variant drops it from the split without deleting it.

Picking a winner

The default winning rule is reply rate; you can rank by open or click rate instead. Warmbly will not call a winner on a tiny sample: it needs a minimum (default 30 sends for the leader) and reports confidence.

How to read your numbers

1. Check the four headline rates for your window.
2. If one looks off, switch the chart to that metric and find the day it moved. That usually points at one campaign or list.
3. Open that campaign and read per-step stats to see which touch is responsible.
4. For bounces or complaints, drop into the per-mailbox breakdown and Deliverability to find the sender at fault.
5. Use A/B analysis for copy decisions, and wait for real confidence.

A clean reply rate with low bounces and complaints beats a big open-rate screenshot every time.

Related guides