Expected = the sum of the configured cold-email daily_limit of every SEND-READY account carrying one of the six infra parent tags: Reseller · Outreach Today · MilkBox · MailIn · MailIn SMTP · Koldify. Send-ready = Instantly status == 1 (so never paused, never in a connection / soft-bounce / sending error state) and carrying no status_message — the second half matters because an inbox whose OAuth has permanently failed still reports status == 1 while being physically unable to send. Not every workspace carries every lane — which lanes a workspace holds on the day you are looking at is what the per-workspace breakdown below shows, and the sub-heading names the workspaces on the Koldify lane that day. Each account contributes its own configured limit; nothing is ever averaged or multiplied out across a pool. Deduped by email, so an account on two lanes is counted once in the workspace total. Warm-up volume is a separate quota and is never counted. (Sam's spec, 2026-08-28. Maxify is excluded entirely. Google Panel and Microsoft Panel are still censused but are NOT counted here — which is why this tab's fleet total will not reconcile to the Infra tab's, where Google Panel IS counted.)

⛔ Resting domains expect ZERO. A workspace deliberately resting its domains is not under-sending — it is doing what was planned, so its expected is 0 for every day of the rest and it drops out of the fleet total both here and in the feed itself. The census still measures those inboxes, so the capacity comes straight back when the rest ends. A rest is always labeled; if a resting workspace does send, that is flagged as a live campaign somebody forgot to pause, never as a percentage. Windows live in rest_windows.json and cover past days too, so turning one on corrects the whole week on the next hourly publish. (Sam, 2026-09-17: "we were resting then, so it shouldn't be added. It's incorrect to say that it should have sent 224,000.")

⛔ Each day shows that day's own expected. The census snapshots at 06:00Z daily and the number on a past day is that day's snapshot, not today's. Until 2026-08-28 the latest snapshot was stamped onto all seven days, so "yesterday" silently moved every time an account was re-limited today — 08-27 rendered F5 at 942,313 when its own census that morning said 797,980.

⛔ Parent tags only. Campaigns bind the parent tag, so capacity counted by tag family (“Outreach Today Active”, “… Warming”, “… Weekend”, “OTD”) is not capacity a campaign can reach — counting the family overstates what the fleet can actually send.

Sent = the workspace-grain daily total from Instantly analytics (/campaigns/analytics/daily with no campaign_id), split into lanes by the per-campaign rows. ⛔ A per-campaign sum is not the workspace total — it drops every campaign deleted or renamed out of the window. That hole was worth 560,103 sends on 08-27 in F5 alone (rendered 110,045; the workspace actually sent 670,148). Volume the listed campaigns cannot account for now appears as its own unattributed lane instead of vanishing — and it is not simply deleted campaigns: F5's campaign list is complete and its lifetime per-campaign totals are smaller than four days of its workspace sends, while two independent Instantly aggregates (workspace-grain daily and the analytics overview) agree with each other to the digit. Instantly's per-campaign breakdown is incomplete at the source, so the residual is parked, never proportioned across the real lanes. A campaign bound to two infra parent tags cannot be split truthfully at campaign grain, so it is reported separately rather than proportioned.

% of expected = Sent ÷ Expected. Nothing else. ⚠ Weekends legitimately run a reduced schedule, so a Saturday or Sunday reads low against a weekday-configured fleet — that is the schedule, not an outage.

Weekend campaigns are included here on purpose. The weekend exclusion governs performance analysis (reply and booking rates are not comparable); this tab answers “did the infra send what it could”, and a weekend send is a real send. Weekend volume is broken out in its own column.