Data Platform capacity gap
Warehouse migration · gap opens September 28, 2026 · owner Priya Raghunathan, VP Engineering
Scenario
The three constraints the plan is scored against.
Change a constraint and the recommendation is recomputed. Same inputs, same plan — this is a scoring function, not a model call.
Sets the start of the delivery window Current scores against.
Spend beyond what the FY26 plan already approved. The open requisition is already budgeted, so it counts as zero.
Capacity that must still exist after the migration ships. A contractor contributes none of it.
Four ways to close a 3.0 FTE gap
Ranges, not false precision. Toggle a path to model a different plan — the recommendation stays visible so you can see what you changed.
| In plan | Path | Time to capacity | Incremental cost | Capacity recovered | Durable | Reversibility | Key risk | Confidence |
|---|---|---|---|---|---|---|---|---|
BuildRecommended Redeploy Maya Chen into the migration workstream | 1.5–6 wks first → full | $19K–$28K Beyond approved budget | 0.83 FTE 28% of the gap | 0.9 FTE after cutover | Reversible | Leaves a partial hole in Product Engineering’s billing pipeline work for six weeks. | High Skill adjacency is evidenced by delivered work, and internal redeployments in this org have ramped within 6 weeks in 4 of the last 5 cases. | |
BorrowRecommended Engage one senior data migration contractor for 12 weeks | 2–3 wks first → full | $96K–$132K Beyond approved budget | 0.97 FTE 32% of the gap | — ends at cutover | Reversible | Contractor context does not stay with the team. Capacity disappears at week 14 whether or not the migration has landed. | High Both MSA vendors have placed comparable roles for Northstar within 3 weeks in the last 18 months. | |
BuyRecommended Keep the approved Senior Data Engineer requisition active | 11–17 wks first → full | $0 incremental Already in the FY26 plan | 0.19 FTE 6% of the gap | 1.0 FTE after cutover | Reversible with cost | Median time-to-fill for this role is 11 weeks, then 6 weeks to ramp. Hiring alone cannot close the near-term gap. | Medium Time-to-fill is drawn from 7 comparable senior data roles over 24 months. The sample is small and the market has moved. | |
AutomateNot authorized Automate migration QA regression and release coordination | 5–6 wks first → full | $9K–$16K Beyond approved budget | 0.36 FTE 12% of the gap | 0.4 FTE after cutover | Reversible | The automatable slice is real but small. It cannot carry the schema translation work, which is where the gap actually sits. | Low The 0.4 FTE estimate comes from ticket-duration data on 6 weeks of history — too short a window to trust, and the task mix changes at cutover. |
Build + Borrow + Buy — a mixed plan
Recovers 1.99 FTE of the 3.0 FTE gap across the delivery window, taking the shortfall from 18% to 6% for $115K–$160K of incremental spend.
A higher-scoring plan was ruled out: Durable capacity of 0.9 FTE is below the 1.0 FTE floor.
Closest alternative: Build + Buy — 34% of the gap for $19K–$28K, leaving 12% uncovered.
Capacity against the delivery window
The shaded band is the window being scored. The hatched area is what the plan still does not cover.
How this plan scored
Every term, weight and contribution.
| Term | Score | Weight | Contribution |
|---|---|---|---|
Gap coverage Share of the 3.0 FTE gap the plan recovers across the window. | 0.66 | ×0.45 | 0.298 |
Speed to capacity Coverage-weighted mean time before each path supplies capacity. | 0.74 | ×0.18 | 0.134 |
Cost headroom How much of the spend ceiling the plan leaves unused. | 0.08 | ×0.15 | 0.013 |
Reversibility How cleanly the plan can be undone if the situation changes. | 0.96 | ×0.10 | 0.096 |
Confidence Calibrated from the evidence behind each path, not decoration. | 0.88 | ×0.12 | 0.105 |
Coordination penalty Charged per extra path. Every path added is another thing to run. | — | 2 × 0.03 | −0.060 |
| Plan score | 0.586 | ||
Modeled, not recommended
Automate — Automate migration QA regression and release coordination. Worth 0.36 FTE across the window, 12% of the gap.
The regression suite and release calendar are owned by Platform Quality, outside the requesting org’s approval scope. Current cannot include a path it is not authorized to change.
Unblocked by: Scope change approved by Sana Iqbal, Director of Platform Quality
It also could not carry this gap alone — the schema translation work is where the shortfall actually sits, and that is not automatable.
Protected attributes are excluded from every ranking on this screen.
Internal adjacency is computed from delivered work, learning records and role history. Performance ratings are not an input to adjacency, because rating data in this org correlates with tenure and team, not with capability for this workstream.
- Selection scope
- 2 candidates surfaced from 96 in-family engineers. No candidate was excluded by the model — the other 94 fell below the adjacency threshold on delivered work.
- Fairness review
- Reviewed by People Analytics, 2026-07-14