Foxglove Group

We watch every entry from your account for recurring themes — repeated tags, similar root causes, and entries that look like the start of something larger. A one-off issue gets fixed in its own entry and closed; a pattern is the opposite — the same root cause surfacing across several entries, where the real fix is structural, not another patch. Below: the categories that are moving this period, then the patterns themselves. Each one carries a small lifecycle: auto-detected patterns get reviewed by CSFWP before they reach you; ones you or your lead hand raise get reviewed by the auto-program first. Either way nothing becomes a project without your sign-off, and anything dismissed stays on the record with the closer named. This period: 4 awaiting your confirmation, 2 already scoped, 2 closed off.

1. Auto-detected · 3h agoModerate

   Guest booking emails are increasingly landing in spam

   Early read, not yet vetted: two recent entries describe guest-facing booking email going missing — follow-ups bouncing because they run through reps' personal Gmail, and the audience sync stalling on a departing marketer's personal key. The shared thread is guest-facing mail leaving via personal, unauthenticated identities rather than an authenticated service sender, which dents deliverability and sender reputation over time. Worth watching for one more instance; it may also turn out to be a facet of the stranded-automations work already scoped.

   1. Auto-detected3h ago
   2. CSFWP reviewpending
   3. Your confirmationpending
   4. Projectpending
   Supporting entries

   • ENT-3yc1bb
   • ENT-3yd2xx
2. Auto-detected · May 22Strong

   Re-tendering creative work every time is eroding the brand

   Three entries in the last 30 days point at the same friction: procurement's $15k tender threshold has no carve-out for a continuing creative partner. The spring venue-launch reel is the live example, but the same pattern cost the spring photography refresh its consistency (flagged in the brand review) and added two weeks to a functions-brochure timeline through repeated vendor onboarding. Each fresh tender saves roughly $3-4k on the bid but burns 8-10 days of vendor onboarding and erodes the brand cohesion we then spend post-production hours trying to recover. Net cost is well above the savings. The fix is a one-meeting policy change with procurement.

   1. Auto-detectedMay 22
   2. CSFWP reviewedNik · May 22
   3. Client confirmedMaya · yesterday
   4. ScopedPRJ-08
   Supporting entries

   • ENT-3y8a0p
   • ENT-3y2k09
   • ENT-3y1m22
3. Auto-detected · this morningStrong

   Duplicate guest records are costing the team hours every week

   The matching rules catch about 73% of duplicate guest records cleanly. The remaining 27% are fuzzy matches — slight name spellings, missing accents, and multi-touch arrivals where a web booking form and an Instagram DM fire within seconds. The rules don't do fuzzy matching by design; that's a different class of problem. The functions team currently absorbs this as 4+ hours of manual merging per week. We've patched two specific incidents this quarter (ENT-3y3n44 most recently) but they keep recurring because the root cause is architectural, not configuration. The clean fix is a small fuzzy-match pre-pass that flags candidates before the strict rules evaluate — about a week of work, no disruption to the existing rules.

   1. Auto-detectedthis morning
   2. CSFWP reviewedNik · this morning
   3. Your confirmationawaiting
   4. Projectpending
   Supporting entries

   • ENT-3y7f12
   • ENT-3y3n44
4. Auto-detected · 3 days agoModerate

   Out-of-date content is reaching customers and partners

   Three Knowledge-staleness entries this month: last season's package quoted to a client by a functions agent, a retired package list in a live proposal deck, and a booking article referencing the retired widget UI. Spot-check on the broader base: of the top 40 most-viewed articles, 31 have not been edited in 6+ months and 11 have not been touched in over a year. There is no review SLA, no per-article owner, no flag for stale-after-X. We are spending roughly 2 hours per entry putting out fires — and these are only the visible ones; the agent incident only surfaced because the client chased the price. This also gates two bigger moves: the guest-concierge agent (AN-28) can't be trusted until Knowledge is clean, and the offer-sync SOP (AN-29) needs Knowledge as a managed surface. Better folded into those than stood up alone.

   1. Auto-detected3 days ago
   2. CSFWP reviewedNik · 3 days ago
   3. Your confirmationheld
   4. Watchingmore evidence needed
   Supporting entries

   • ENT-3y6r88
   • ENT-3y4s17
   • ENT-3xz8s7
5. Auto-detected · yesterdayStrong

   Critical systems are tied to one person — who's on leave in a month

   Two entries this week surfaced the same architectural smell from different angles. The Mailchimp→Salesforce audience sync runs under Priya's personal API key on her work laptop (ENT-3yd2xx); the functions team's booking follow-ups authenticate as each rep's individual Gmail (ENT-3yc1bb). A quick audit of the broader stack found three more integrations in the same shape — a lead-enrichment webhook, a BI export, and the Collins→Salesforce booking sync — none of them on a service account. The failure mode is silent and asymmetric: everything works until someone leaves, takes leave, or hits a re-auth prompt nobody else can see. We've patched both immediate entries, but the right move is a deliberate migration of the dependency layer onto service accounts and shared sending infrastructure. Priya's leave landing in four weeks is the natural forcing function — if we don't move now, we'll move under pressure.

   1. Auto-detectedyesterday
   2. CSFWP reviewedNik · yesterday
   3. Client confirmedMaya · yesterday
   4. ScopedPRJ-07
   Supporting entries

   • ENT-3yd2xx
   • ENT-3yc1bb
6. Raised by Maya Tan · May 20Moderate

   We keep losing the files guests send us

   Two file-loss incidents this month, from different intake channels, point at one root cause: payloads over ~5MB are silently dropped on the way into Salesforce. A wedding planner's 7MB site map uploaded through the website enquiry form never arrived on the record (ENT-3yh2k8); a corporate client's 9MB run-sheet emailed to the functions inbox never attached to the booking (ENT-3yj5n1). Each was patched in the moment by asking the guest to re-send something smaller — but the functions team is now quietly coaching every client to 'keep it under 5MB', which isn't a fix, it's a workaround calcifying into policy. Maya raised it as 'we keep losing attachments'; the auto-program cross-referenced 90 days of ingest and confirmed the shape: 6.4% of submissions with an attachment arrive empty, climbing to 18% once the file is over 5MB, and the failure is silent — the record is created, the body is fine, the attachment is just missing. This is the ingest layer, not any one channel, so the fix is one routing change (hand >5MB payloads through a storage step) that covers the web form and email intake at once. Worth confirming the framing before we scope.

   1. Maya Tan raised itMay 20
   2. Auto-program reviewedyesterday
   3. Your confirmationawaiting
   4. Projectpending
   Supporting entries

   • ENT-3yh2k8
   • ENT-3yj5n1
7. Auto-detected · this morningStrong

   After-hours enquiries go unanswered — and guests book elsewhere

   A steady stream of after-hours enquiries arrives on Instagram DM and website chat — hours, parking, walk-ins, group availability — and sits unanswered until the morning, by which point a measurable share of those guests have booked elsewhere. Across the last 30 days, ~68% of after-hours enquiries are repeat questions already answered in Knowledge, peaking 7-11pm when nobody's monitoring. The move is a guest-concierge agent on Agentforce, grounded on the Knowledge base, live on web chat and Instagram DM: it answers the FAQ tier instantly and hands any qualified function or group enquiry to the team with the guest's details attached. One hard dependency — it's only as trustworthy as the Knowledge behind it, so it rides on the staleness fix (AN-23), not ahead of it.

   1. Auto-detectedthis morning
   2. CSFWP reviewedNik · this morning
   3. Your confirmationawaiting
   4. Projectpending
   Supporting entries

   • ENT-3yf4q2
   • ENT-3y6r88
8. Auto-detected · yesterdayStrong

   Every offer is updated by hand in four places — and one always slips

   This is the throughline under three separate threads. The procurement-tender chaos (creative re-tendered every campaign), the Knowledge staleness (offers going out of date in the articles staff and agents quote from), and this week's winter-offer miss (the price updated in Webflow, Mailchimp and Collins but not Salesforce Knowledge, so staff and the website chat quoted the old number for two days) are all the same root cause: an offer lives in four places and is reconciled by hand. The fix is to make the offer a governed object. Run the graphic-design tender and offer process as an SOP on a Monday board — brief → bids → award → asset approval — and on approval, automations plus the integration layer publish the finished offer to Webflow CMS (the site), Marketing Cloud (campaigns), and the Salesforce Knowledge article (staff + the concierge agent) in a single move. Approve once, every surface updates; the brand stays cohesive and Knowledge can't drift. It also retires the manual-propagation work that's quietly eating the content coordinator's launch weeks.

   1. Auto-detectedyesterday
   2. CSFWP reviewedNik · yesterday
   3. Your confirmationawaiting
   4. Projectpending
   Supporting entries

   • ENT-3yg7m5
   • ENT-3y8a0p
   • ENT-3y6r88
   • ENT-3y4s17
9. Auto-detected · May 12Moderate

   Enquiries spike every weekend

   Pattern detector flagged a recurring 2-3x spike in Friday-evening and Saturday inbound across the last six weeks, well outside the weekday distribution. Worth bringing to Maya as a possible coverage or demand pattern.

   1. Auto-detectedMay 12
   2. CSFWP reviewedNik · May 13
   3. Client confirmation—
   4. Closed by MayaFoxglove Group · May 14
10. Auto-detected · May 16Weak

    Replies on emailed-in entries are getting slower

    Pattern detector flagged a 40% increase in time-to-first-response on emailed-in entries vs Portal across the last three weeks.

    1. Auto-detectedMay 16
    2. CSFWP review—
    3. Client confirmation—
    4. Closed by NikCSFWP · May 18

Pattern detection runs nightly across your last 90 days of entries, projects, and Throughlines. Strengths and recommendations are first-pass; talk back to any of them and your lead hand at CSFWP will rework the analysis or open the project for you.