▸ CompassValue–Irritant classification, root causes and action plans now live in Compass — the same intents, placed by value to the company × value to the customer, then turned into owned actions.
The six outcomes, the two biggest movers with their drivers, this week's signals, fatal errors, and where actions stand — one page, every figure linked back into the product.
The month in CX: the six outcomes against May and target, demand shifts, quality and fatal trends, the emotion journey, initiative expected-vs-realized, and customer journeys.
Identity not verified before account change on call #4239 — Voice · In-house Pod B · Sat Jul 11. The audit auto-zeroed. This is the fourth identity-verification fatal this week.
AeroShip was named in 490 conversations this week, up 144 on the prior week (+42%). Negative share 83% — the highest of the three carriers. Missed transit SLA remains the leading probable cause.
Reports restate figures from the product — one number, one source. Sections a tenant's data doesn't support are omitted, not shown locked. Watches fire from the same tests as the pages.
8,470 service contacts carried $66,902 in total cost to serve this week — $41,249 handling + $25,653 downstream (modeled). First contact resolution rose a point to 68 against a target of 80 — +0.6 from rate, +0.4 from mix; quality held at 87 with 7 fatal errors; experience held at −11. Volume is the mover — contacts grew 3% on the prior week and cost rose $1,364. Two initiatives are live worth $565K expected; two breaching movers remain uncovered.
What moved
First contact resolution rose 1.0 point — 67 to 68 · rate +0.6 · mix +0.4Summit ↗
Volume grew 3% week over week — 8,470 contacts vs 8,190 · cost of CX rose $1,364Registry ↗
Signals
Orders held for verification spiked to 20× baseline — 328 contacts, first seen Jul 7Evidence ↗
Churn language at 3.1× baseline — 501 conversations, watch week 2 of 4Watchlist ↗
Field keeps recovering deliveries — 96% resolution on 100 visitsEvidence ↗
June carried roughly 34,320 service contacts at $61,614 a week in total cost to serve — $37,986 handling + $23,628 downstream (modeled). Resolution rose three points to 65, still fifteen points from target. Quality rose to 86 with fatal errors falling from 36 to 26. The negative emotion margin narrowed from −14 to −12, and delivery-update contacts grew 23% on May. One initiative shipped in May is realizing $38K a year against $32K expected; two more went live in the last week of June.
The six outcomes — June vs May
Demand — what shifted
Get a delivery update rose 23% — 651 weekly contacts by month end vs 530 · carrier transit misses deepenedDemand ↗
Cancel my order rose 9% · 38% of cancellations cite delivery delay — the save-offer play sits unadoptedShelf ↗
Check product availability eased 6% as June restocks landedDemand ↗
Quality
Program quality 86 vs 85 in May · defect-free contacts 67% vs 65% · target 85 held both monthsQuality ↗
Fatal violations 26 vs 36 — identity verification remains the largest classFatals ↗
Addresses every question worsened through June — 30% to 33% fail — concentrated at MK — EmailParameter ↗
Experience
Negative emotion margin −12 vs −14 in May · 22% of journeys worsened · save rate 45%Registry ↗
Churn language emerged in the price-change cohort late June — 3.1× baseline by Jul 12, under watchWatchlist ↗
Initiatives — expected vs realized
This table is the QBR section — the same figures as the Actions register. Open the register ↗
Customer journeys
41% of follow-up contacts changed channel · top reason: no reply within the promised windowJourneys ↗
Computed on the 85% of contacts with a linked customerIdentity ↗
Definitions and targets are governed in Studio ↗. Sections a tenant's data doesn't support are omitted from delivery.
inferred on every contact from friction events · survey CES shown beside it · 1–5, higher is easier · Jul 6 – 12 · definition & calibration in Studio ↗
3.9 inferred
+0.05vs prior week survey CES 3.8 · 1,420 responses this week · gap 0.1
↓ CSV · Social muted at 60 contacts (low n) · Field n/a — visits carry duration, not handle
Channel
Partner · conversational contacts
Drivers — minutes added by friction
events observed on every recording and thread · 4,043 minutes · 6% of the week's 65,147 handle minutes
Contacts in the shortest handle decile repeat at 1.8× the average within 7 days — association from the signals engine, not attribution. Repeat check in the FCR deep dive ↗
a journey = one customer, one issue, first contact to close · Jul 6 – 12 · computed on the 85% of contacts with a linked customer · identity in Studio ↗
Studio changes how numbers are computed and displayed from their effective date. Nothing here edits a conversation, a score already issued, or an initiative's frozen expectation.
Ask Intendra
Ask anything about your conversations
answers draw on the week's scored conversations · every figure opens into its surface
History
Fit
Who fits what — intent & customer to associate
Channel to Intent Fit
one-contact resolution · journeys in small print · cells under 10 journeys muted · ↓ CSV
— the intent does not arrive on that channel · each row weights back to the intent's overall first-contact resolution
Best Fit Associate to Intent / Associate to Customer
fit40 → 95 · shrunk toward team mean when evidence is thinthin evidenceclusters score on a 12-week window · situations and cluster × situation pairs on 26 weeks
Partner × Intent × Customer
same rubric, same model, every partner · routable pool = associates at ≥75 with confident evidence
How the mapping works
Intents are detected per contact at ask level. Agent fit is scored at the capability-cluster level — groups of asks that need the same skills — with customer situations (first-time, snowball, irate) overlaid on any ask. Cells with n below 35 are averaged toward the team mean and marked thin; clusters score on a 12-week window, rarer situations on 26 weeks.
Crossing situations with clusters
Focus a situation and the cluster columns re-score for that situation — First-time customers × Saves & cancellations reads: how each agent handles a save when the caller is new. Pairs score on the 26-week window; both detections must sit on the same contact. Thin pairs are averaged toward the team mean and are never used for routing or bench decisions on their own.