|
Northwind Financial · Report Sentinel
Morning digest
Monday 03 August 2026 · sent 09:00 ET · checks run 06:40 ET
Open 3 of 7 today. Service Levels and Queue Health, Acquisition Channel Mix, Collections and Payments
|
|
207 checks across 33 metric series and 7 reports
· nine predefined rules · nothing tuned this morning
|
|
This morning
Three of the seven reports cleared the attention threshold this morning. Service Levels and Queue Health is the one to open first: abandon rate printed 7.19% against a 5.75% trailing mean, four sigma above its band, and average speed of answer has now closed above its center line eight business days running. Acquisition Channel Mix is second, where the source mix moved 5.93 share points off its trailing shape in three days, paid social up 5.38 points and paid search down 3.25, with blended cost barely moving. Collections and Payments is third and narrower: the return rate printed 3.93% against a 3.46% mean, a single day step at 3.4 sigma, and nothing else on that report left its band. The thread worth pulling connects the contact center to workforce management. Calls offered sat 2.5 sigma below its trailing mean on Friday while answer speed and end of day backlog both ran high, so the queue got slower on lighter volume, not heavier. Schedule adherence has fallen six business days without a pause over the same window, from 92.11% to 91.00%, which is inside its band every single day and therefore invisible on a daily glance. Neither report can establish cause, and the checks make no attempt to. The question for the floor and the contact center owners is whether the same schedule change sits behind both, and that is answerable this morning with the shift template history. Revenue Funnel Daily and Enrollment Pipeline and Aging came through with no rule firing at all and can be skipped. One limit on this morning's read: Marketing Spend and Acquisition Cost is a business day behind the rest of the fleet, so its bands are computed on stale input and the channel mix shift above has not yet been priced against settled spend. Treat any cost per acquisition read from that report as provisional until the feed lands.
Narrative source: claude-opus-5, committed sample. Written after the ranking, never before it.
|
|
Open these 3 today Ranked by attention score. The reason is stated before you click. | Contact Center · data through 2026-07-31 | OPEN TODAY · 100 |
Beyond 3 sigma x22 of 3 beyond 2 sigma x28 day run, one side4 of 5 beyond 1 sigma x2 Abandon rate printed 7.19% against a 5.75% trailing mean, 4.0 sigma above the band. Average speed of answer printed 40.7s against a 34.8s trailing mean, 3.1 sigma above the band. Average speed of answer broke two sigma on 3 of the last 3 business days, all above the center line. | Metric | Latest | 20d mean | Z | | Abandon rate | 7.19% | 5.75% | +3.99σ | | Average speed of answer | 40.7s | 34.8s | +3.13σ | | End of day backlog | 468 | 417 | +2.11σ |
|
Growth Marketing · data through 2026-07-31 | OPEN TODAY · 61 |
Mix shift Channel mix moved 5.93 share points off its trailing shape (+5.06 sigma): Paid Social +5.38pp, Paid Search -3.25pp. |
Servicing · data through 2026-07-31 | OPEN TODAY · 56 |
Beyond 3 sigma Return rate printed 3.93% against a 3.46% trailing mean, 3.4 sigma above the band. | Metric | Latest | 20d mean | Z | | Return rate | 3.93% | 3.46% | +3.41σ |
|
| Worth a skim (2) Inside the warning band or running on late input. Not urgent. | Growth Marketing · data through 2026-07-30 | SKIM · 28 |
Feed did not refresh Latest data is 2026-07-30, 1 business day behind the fleet. Every band on this report is judged on stale input. |
Workforce Management · data through 2026-07-31 | SKIM · 26 |
6 day monotone trend Schedule adherence moved lower six business days without a pause, 92.11% to 91.00% (-1.20%). |
| Clear, skip today (2) Every check passed. Listed so the silence is auditable, not assumed. | |
|
|
What the bands claim
Each metric is compared with its own trailing 20 business days. Sigma is estimated from the
average moving range, so a step change does not inflate the limit that is supposed to catch
it. A flag means the reading is unusual for that series. It is not a target miss, not a
forecast, and not a cause. With 33 series under nine rules, one flag every two
to three business days is expected to be noise, which is why this digest ranks and never
pages. One report is running on late input this morning: Marketing Spend and Acquisition Cost.
Synthetic demonstration. Every figure is generated by generate_fleet.py
from seed 20260803. No employer data, schema, figure, or branding appears anywhere in this
repository. Regenerate the whole space with
python3 generate_fleet.py && python3 sentinel.py.
|
|