Greenbox Capital Goodwood Consulting
Meeting Summary
Box ↔ HubSpot Alignment

Box ↔ HubSpot Alignment — Meeting Summary

Greenbox × Goodwood alignment meeting to scope the Goodwood-owned middleware that replaces the legacy Box Update Service. Architecture greenlit; ownership confirmed; endpoint shape agreed; mapping deliverable due to Kris by EOD 5/14.
2026-05-13 · 2:30pm ET · 47 min

Attendees

PersonOrg / Role
Sarah LackeyGreenbox — VP Technology (host)
Kris GlaittliGreenbox — VP Sales
Brian LaMureGreenbox — Enterprise Solutions Architect (5 yrs)
Kusha KapoorGreenbox — owner of Box-side codebase
Nova GuliyevGreenbox — Senior Data Engineer (week 3)
Steven LentzGreenbox — Program/Project Mgmt (reports to Sarah)
Ryan ThibodeauxGoodwood — President

What was decided

DecisionOutcome
Middleware ownership Goodwood owns the integration end-to-end — transformer, reconciler, audit, drift detection. Greenbox tech team is not on the hook for configuring HubSpot itself; they own the source-data interface only. Confirmed
Read path Greenbox exposes REST endpoints via API M gateway — one or multiple endpoints, scoped to specific fields, hitting the read replica. Service-principal credentials. Brian and Nova/Kusha own implementation. Confirmed
Sync cadence Near real-time. Pull cadence = a setting (15-min, 10-min, etc.) gated only by cycle time, not by additional Greenbox work. Brian: “That’s a setting”. Confirmed
Round-trip HubSpot IDs Retired. No more pushing hs_object_id back to Box. Kusha confirmed: the only reason the old pattern existed was v1-API constraints; modern v4 supports custom-ID lookup. Eliminates the merge-creates-dangling-pointer problem. Confirmed
ID scheme Brian announced a standardized prefixed-string ID format that wraps Box-side raw IDs (preserving the original number) to avoid collisions across object types and align with Salesforce + downstream analytics. Adopted as the canonical interface. Confirmed
Picklist sync Pull-based lookups endpoint — Goodwood queries Box-side lookup tables (decline reasons, AM/RAM list, processing status, employee list, etc.) before pushing data, updates HubSpot picklists, then pushes records. Removes Roque-as-human-in-the-loop. Confirmed
Tech stack Goodwood proposed Firebase (Cloud Functions + Firestore). Sarah flagged she doesn’t have capacity to onboard another cloud env. Kusha noted Greenbox already uses Databricks (which has a HubSpot connector). Sarah + Brian taking offline; SOW pricing anchored to Firebase. Open — offline
PII classification Brian flagged: every field exposed through the gateway needs PII classification. Greenbox security review owns this. Pending

Architecture confirmed

Action items

ActionOwnerDue
Deliver HubSpot ↔ Box field mapping (CREATE + UPDATE) so Kusha & Nova can scope endpoint payloads Ryan (Goodwood) EOD 2026-05-14
Design REST endpoint payloads + API M wrappers; resolve secondary-owner JOIN at source Brian + Kusha + Nova (Greenbox) Following 2 weeks
Size the engineering work on the Greenbox side; slot against the “model two” priority queue Sarah + Steven (Greenbox) Following 1–2 weeks (likely end-of-month)
Tech stack decision (Firebase vs Databricks) — offline business-strategy discussion Sarah + Brian (Greenbox) Before sprint kickoff
PII classification per field exposed through API M Greenbox security review Before production cutover

Notable moments

“In your new scenario flowing back to box at all — you’re just basically polling box to find out what data needs to be synced over to Hubspot, and if that’s the case we no longer need Hubspot data back in box at all.” Brian LaMure · 14:00
“The only reason we require the hub spot IDs today is so that we can push that information to hub spot, and we are actually using the first version of the API. So that was the reason — they did not support custom IDs to be used and had to be hub spot IDs.” Kusha Kapoor · 14:20
“Close to real time is what we want to get to, but there’s points of diminishing return.” Kris Glaittli · 16:41
“Part of our standard interfacing schema now includes a completely standardized way to identify any object, whether it’s a merchant, an owner, a loan, whatever, as a string. We proxy all keys with one standardized string format... and that’s what we would expose in the APIs.” Brian LaMure · 35:36
“I’ll get that to you tomorrow — that works.” Ryan Thibodeaux · 45:20 (commitment on field mapping deliverable)

Sentiment

The room is fully aligned. Kris (the impacted business owner) has been waiting on this for 18 months and visibly relieved when Brian volunteered the Greenbox engineering work as a "fairly straightforward parallel-type task." Sarah explicitly priced the project against renewal strategy (“high priority”) but flagged a competing initiative — "model two" — that may push start into end-of-month. Brian's confidence is high; he and Nova will scope the endpoints. Kusha is on board with the new pattern and welcomes removing the eventing burden.

Open items going into Sprint 1

  1. Tech stack decision — Firebase vs Databricks. Anchored to Firebase in the SOW until decided.
  2. API M auth scheme — service-principal credentials, but specific flow (OAuth client-creds vs HMAC vs API key) TBD in Week 1 contract review.
  3. PII classification per field — Greenbox security review owns; gates production cutover.
  4. HubSpot daily API limit — current 1M/day cap is tight under build + cutover load. Greenbox to request increase before Week 3.
  5. Greenbox-side prioritization against the "model two" track — Sarah to slot.
Next reference: see the full transcript for verbatim discussion, and the project plan for the formalized 4-week sprint scope, deliverables, timeline, and field mapping that flows from this meeting.