E7 Behavioral Cheat Sheet
Print this. Read it 10 minutes before walking in.
Before you open your mouth: the E7 litmus test
Run this against every story you’re about to tell. If you can’t check all four, pick a different story.
- I identified the problem — nobody assigned this to me
- I set the direction — the strategy was mine, not my manager’s
- I influenced executives — VPs/directors changed resource allocation because of me
- I changed multiple orgs — the outcome crossed org boundaries, not just teams
Story skeleton (A0–A4 at E7)
| Step | What to say | E7 signal |
|---|---|---|
| A0 – Spotted it | ”I noticed [company-wide problem] that no one was solving” | You found it, not assigned |
| A1 – Built the case | ”I quantified the cost: [dollars/eng-hours/risk]“ | Data-driven, exec-legible |
| A2 – Designed the strategy | ”I proposed [approach] with a [timeline] roadmap” | You authored the plan |
| A3 – Aligned leadership | ”I presented to [N VPs], negotiated [resources/ownership]“ | Exec influence |
| A4 – Enabled at scale | ”It became the standard — [N teams/orgs] adopted it” | Company-wide multiplier |
Each A-step = one What sentence + 2–3 How details. Example: “I aligned leadership (WHAT) by mapping each VP’s priorities, presenting ROI in their terms, and proposing shared ownership so no one org bore all the cost (HOWs).”
Your 19 stories — recall guide
Strongest E7 stories (company-wide strategy + exec influence): 2, 9, 12, 15, 16, 18
1. Snapshot Generator (Cap One, May–Jul 2024)
10 teams (~80 engs) blocked: CDC init took 30 days per topic. Found nightly exports weren’t point-in-time. Built Spark-based snapshot generator using event history → rebuild state in minutes. Parameterized across 33 topics, added monitoring + runbooks. Result: 30d → 10min init. 10 teams unblocked. Technique reused in 3 modernization programs.
2. VP Trust / Temporal (Cap One, Jan–Sep 2024)
Peer VP (20 teams) resisted Temporal orchestration — feared delay if tech failed. Did 1-on-1s with VP + managers to understand fear. Built POC: deploy 2wk → 3d speedup. Engaged trusted managers as co-owners. Low-risk pilot 2wk → success → rolled out to all 20 teams. Result: VP became champion. Temporal standardized across 20 teams. Deploy time -70%. Used same playbook on 2 other VPs.
3. Claims Cross-Team Leadership (Cap One, Sep 2024–Feb 2025)
16 contractor teams (~250 eng) rushing after Discover deal. Ambiguous ownership. Audited 5 handoffs + 70 PRs → found testing/ownership gaps. Drafted RACI for VPs. Ran code reviews showing Ports/Adapters via live refactor. Introduced mandatory unit + contract tests. Flagged Dec delay 6wk early → reset scope for Feb. Result: Feb on time. First phase eliminated stuck claims ($34M/yr). Patterns replicated in 2 later programs.
4. Claims Legacy Decommission (Cap One, Q4 2024–Q1 2025)
Legacy Java monolith: 1M claims, $500M exposure, 10% stuck ($50M risk). Cataloged 5 resolution strategies, found dual-write fault zones. Designed workflow-agnostic Lambda Router + Step Functions with idempotent payloads. Piloted 2 strategies (40% volume), shadow diffs verified 0 state drift. Trained 16 teams in 2 weeks.
Result: Stuck-claim rate → 0% on migrated flows ($20M risk removed). Pattern became Fraud Tech standard.
5. Claims Delay & Recovery (Cap One, Sep–Feb 2025)
Joined Sep into project started Jun; Dec deadline already at risk. Found over-granular Step Functions (>30 states) → slow + fragile. Reported to VP end-Sep with quality gaps + mitigation plan. Rebuilt 2nd workflow (Ports/Adapters, Lambda OOP, Step Fn for durability only). Result: Dec delay acknowledged early (no surprise). Feb delivered on schedule. 0 stuck claims (~$20M saved).
6. Claims Conflict Resolution (Cap One, Jan 2025)
Two tech leads deadlocked on Lambda granularity — blocking workflow release. Facilitated joint session reviewing Step Fn→Lambda RPC semantics. Compared to gRPC oneOf pattern; showed state cost of extra Lambdas. Consensus: single Lambda per activity type. Result: Conflict closed pre-launch. Published design standard repo used by 5 teams org-wide.
7. Fraud Team Transformation (Cap One, Q3 2024–Q1 2025)
Team of 12 migrating EMR→Glue without big-data skills. System unstable, 5000-line YAML pipelines. 1wk audit found: Spark UI unused, jobs ran 1 executor (QA config copy-pasted to prod). Built hands-on Spark/Glue curriculum + weekly labs. Introduced dbt for SQL defenses. Replaced custom StepFn orchestration with Glue Workflows (-60% complexity). Paired tech lead through 3 cycles → independent. Result: Incidents 40→12/mo (-70%). Runtime -35%. $25M fraud prevented. Training kit rolled to 2 more teams.
8. Fraud Pipeline Incident (Cap One, Oct 2024)
Spark pipeline failed 1 week after credit-card launch → ~$3M fraud loss accumulating. Formed SWAT same day (Data + Infra + Fraud teams). Root cause: 30:1 partition skew → OOM on 1 executor. Fix: repartitioned data + autoscale executors → service restored in 18h. Then wrote Spark tuning guide + post-mortem template + skew alerting dashboard. Led debug workshops for 3 teams. Result: Resolved <24h. 0 recurrence in 12 months. Playbook adopted by 4 teams + Fraud Platform on-call.
9. Temporal as Enterprise Pattern (Cap One, Q4 2024)
Org leaning on SQS/Lambda for core account workflows. Mapped 15 async steps / 8 integrations → fragile state mgmt, months of debug risk. Built Temporal Saga POC for account booking (auditable + rollback in minutes). Demoed strong-read API via Temporal signal. Presented to execs: latency -70%, failure -90%. Reframed as risk reduction, not tech debate. Result: Adopted as enterprise standard. 20+ teams migrated. Code complexity -40%.
10. Temporal Evaluation (Pinterest, May–Jun 2023)
Partner team building entitlement platform leaned toward SQS/Lambda choreography. In 3 weeks: mapped SQS vs Airflow vs Temporal, benchmarked latency/observability/error recovery. Built Temporal POC for budget reservation Saga (<200ms vs 3min Airflow). Let them run both approaches 1 week — Temporal won on observability. Wrote trade-off doc for architecture review board. Result: Team adopted Temporal. 6mo debug saved. Decision template reused for 2 later evaluations.
11. Budget Planning / Trust (Pinterest, Aug–Dec 2023)
Month 1 onboarding. 80 tables + 50 pipelines duplicated, no tests, contractors underperforming. Met Finance (wanted auditability), Platform (accuracy), Service owners (simplicity). Chose Ports/Adapters for testability. POC’d GraphQL hierarchical queries + Cedar authz. Built 6-stage approval workflow with audit changelog + real-time subscriptions. Result: Prod on schedule. 0 incidents. Pattern reused by 2 systems. Cross-org trust established in first quarter.
12. dbt Company-Wide Introduction (Pinterest, 2023)
Pinterest had zero dbt. 150+ tables, monolithic SQL (30+ CTEs, 500+ lines), no unit testing. System allocating $700M/yr in cloud spend couldn’t be safely changed. Diagnosed missing modeling discipline as root cause. Wrote paper decoupling attribution from aggregation. Piloted dbt, proved value. Partnered with Data Platform team on DAG renderer, extended with Iceberg WAP support. Result: dbt became standard for Infra Gov. 6 platforms migrated in 6mo with 2 engineers. Pages 200/mo→5. SLA breaches→0.
13. Developer Tooling (Pinterest, 2023)
Spark-based dev workflow: 5-min minimum per iteration, engineers avoided testing. Built custom dbt-Presto adapter (wrapping official adapter for Pinterest auth). Wrote DuckDB sampling scripts for offline local dev. Designed tag-based project structure for isolated testing of independent platforms. Result: ~100x faster feedback. Adopted org-wide. Testing went from avoided to default.
14. PagerDuty Reliability (Pinterest, 2023)
~200 pages/mo. Email-based paging with regex parsing. No routing to owning teams. Mapped failure modes — most were noise (upstream platform issues triggering attribution alerts). Paired with junior engineer on design: walked them through Airflow callbacks, PagerDuty API, routing logic. Junior owned implementation. Built classification: upstream→platform team, attribution→our team. Result: 200→5 pages/mo. Junior went from zero infra experience to independently owning operational tooling.
15. Cross-Org Platform Migration (Pinterest, 2023–2024)
12 platform teams each with bespoke SQL for cost attribution. ~1 year to onboard each. Defined repeatable migration pattern: onboard usage data → validate chargeback model → replace legacy SQL → verify cost attribution for C-level reporting. Met each platform team for data contracts. Executed 6 platforms in 6mo with 2 engineers. Result: Onboarding 1yr→1mo. 50% of system migrated. SLA breaches→0.
16. Cost Hierarchy / Beyond OKR (Pinterest, Mar–Jun 2023)
Org reorgs broke 80 cost pipelines. Finance spent weeks manual-fixing. Not on my OKR. Mapped roll-up logic across 50 pipelines (duplication everywhere). Designed cost_center_hierarchy table (parent/child + effective dates). Met Finance lead, demoed simple org-change update. Built dbt dimension + traversal macros.
Result: Org change weeks→minutes. Pattern reused across 50 pipelines, replicated to 2 other hierarchies.
17. dbt Migration / Productivity (Pinterest, 2023)
50 Python/SQL pipelines with 90% duplication → velocity collapse. Cataloged pipelines, grouped by pattern (cleanup, join, agg). 3-phase migration plan. Piloted AWS platform subset (15 pipelines), validated lineage + CI/CD. Built macro library + auto-validation on PRs. Result: Dev cycle weeks→days. Velocity x3. Pattern adopted by adjacent teams.
18. COVID Scaling (Credimi, Mar–May 2020)
CTO at digital lender (€2B issued). COVID lockdown paralyzed banks; SMBs needed funds in days. Met Risk & Ops, segmented requests: auto-approve (good history + <€50k) / manual / reject. Built auto-decision engine. Presented to 5 bank partners, aligned on daily risk reporting. Launched with 1 bank, validated defaults, scaled to all 5 in 3 weeks. Result: 400→1300 loans/mo (3x). 97% satisfaction. Became Generali’s emergency-funding platform.
19. Stakeholder Misalignment / Failure (Sapient, 2018–2019)
$15M Gates Foundation project. Funder wanted analytics; grantees needed simple data tools. Focused solely on Foundation feedback → ignored end users. Released MVP meeting funder criteria but grantees couldn’t use it. Recovery: bi-weekly office hours + shadow sessions, logged pain points. Redesigned flows, joint UAT with grantees. Result: Platform adopted 3mo later. 40% features redesigned. Lesson: always involve end users when funder≠user.
Influence moves — quick ref
| Currency | What you trade | Drop this phrase |
|---|---|---|
| Recognition | Give visibility to their leadership | ”I highlighted your team’s contribution in my exec update” |
| Assistance | Solve their problem first | ”Before asking for buy-in, I helped them with [their pain]“ |
| Information | Share data they need | ”I gave them early benchmarks so they could plan” |
| Network | Connect them to people | ”I introduced them to [person] who unblocked [thing]” |
The 4-step formula: Engage (ask questions) → Align (shared goals) → Credibility (your track record) → Facilitate (make yes easy)
Scope phrases: E6 vs E7
| E6 (will get you down-leveled) | E7 (what to say instead) |
|---|---|
| “across 3 teams" | "across 4 orgs" |
| "I led the project" | "I set the technical direction" |
| "6-month timeline" | "12-month+ roadmap I defined" |
| "convinced the tech leads" | "presented to VPs and negotiated resources" |
| "became our team’s standard" | "became the company standard" |
| "resolved a disagreement between teams" | "resolved competing priorities between orgs" |
| "I built it" | "I shaped what 100+ engineers built” |
Per-company adjustments (2 lines each)
Amazon: Name the Leadership Principle explicitly in your answer. Every result needs a number. Bar Raiser can veto everything — behavioral matters more than technical here.
Meta: Your 2-min answer is just the opener — they’ll drill for 10-15 min. Prepare: “alternatives considered?”, “what would you do differently?”, “how did others react?” E7 signal = operated independently at org level for 12+ months.
Netflix: Own your failures harder than your successes. Use their language: “Freedom & Responsibility,” “Context not Control.” One negative vote = rejection. 40-50% of eval is culture fit.
Google: Committee of 5+ reviews your packet. Show intellectual humility + collaboration. “Googleyness” matters — creativity, openness to feedback, no ego.
Emergency recovery phrases
If you’re rambling: “Let me cut to the outcome — [result in one sentence].”
If you told an E6 story by accident: “…and that’s how it started. What made this E7-scale was that I then [escalate to org-level impact].”
If you blank on details: “The specific numbers were [approximate confidently]. The key decision point was [pivot to the strategic choice you made].”
If interrupted: “Great question. [15s answer]. Want me to continue, or move on?”