**Drop 6 – Roadmap & ROI: Your First 90 Stores**
*Series capstone: “Edge Renaissance—putting compute (and the customer) back where they belong.”*

---

### ☕ Executive espresso (60‑second read)

* **Three phases, ninety stores, one playbook.** Pilot (≤5), Prove (≤30), Scale (≤90).
* **ROI is real—and fast.** 100 ms faster + higher uptime → +\$X M revenue; hardware amortizes in \~18 months.
* **Govern once, repeat everywhere.** Immutable builds, GitOps, patch waves. Treat each store like a tiny region.
* **Message to the board:** “We’re not reinventing—just putting compute back where customers are.”

---

## 1️⃣ The 0‑5‑30‑90 rollout map

```
PHASE 0: Prep (4–6 weeks)
Goal: Tooling, golden images, KPIs defined
Outputs: Git repo, CI/CD, BOM finalized, latency baseline

PHASE 1: Pilot 5 (6–8 weeks)
Goal: Prove CX & $ impact with 2 workloads (e.g., /inventory API + image cache)
KPIs: TTFB ↓100 ms, conversion ↑≥5%, no POS downtime during WAN blips
Deliverables: Exec brief, ops runbooks, security sign‑off

PHASE 2: Prove 30 (8–12 weeks)
Goal: Add 2–3 more edge workloads (vision AI, promos), test patch waves
KPIs: Cache hit ≥90%, alert MTTR <15 min, patch drift ≤N‑1
Deliverables: Audit evidence automation, SLO dashboard, budget ask for scale

PHASE 3: Scale 90 (12–16 weeks)
Goal: Industrialized rollout kit; 3 waves of 30 stores
KPIs: 95% stores on latest bundle, <0.1% rollback rate
Deliverables: Playbook v1.0, board update, expansion roadmap (300+)
```

---

## 2️⃣ Quick‑math ROI you can defend

**Inputs (example):**

* Online revenue = \$500 M/year, base conversion 2.5%
* Latency cut = 100 ms → +7–8% conversion (industry studies)
* Cap‑ex per store cluster (Good/Better tier) = \$5–6 k (4‑yr life)
* CDN/egress avoided = \$0.045/GB × 100 TB/mo ≈ \$4.5 k/mo

**Back‑of‑napkin:**

```
Revenue lift:  $500M × 7% = +$35M/year
Hardware:      90 × $6k =  $540k  (=$135k/yr amortized)
CDN savings:   ~$54k/yr (after expansion)
ROI (Year 1):  (~$35M + $54k) / $675k ≈ 51×
```

*Swap in your numbers; the order of magnitude rarely changes.*

---

## 3️⃣ People & process: who does what?

```
CORE EDGE PLATFORM SQUAD (8–10 FTE)
• Platform Lead (1) – roadmap, budget, exec comms
• SRE/Automation (2–3) – CI/CD, GitOps, patch waves
• Security/Compliance Eng (1–2) – policy as code, audits
• Edge App Owners (2–3) – /inventory API, promos, vision AI

REGIONAL / STORE TECH PARTNERS (as‑needed)
• “Smart hands” for racking, swaps
• Local networking tweaks

BUSINESS STAKEHOLDERS
• Digital Product, Merch, Ops – define the CX wins
• Finance – track realized savings & uplift
```

> **Rule:** Central team builds the pattern, sites pull it. No snowflakes.

---

## 4️⃣ Operate like a cloud, even if it’s a closet

**Your top 6 SLOs (track weekly):**

1. **TTFB p95 (ms)** for key endpoints (PDP image, `/inventory`)
2. **Cache hit rate (%)** on Varnish/Nginx
3. **POS/API latency (ms)** during WAN events
4. **Patch drift (# stores on N‑2 or older)**
5. **Backup success rate (%)** (PBS snapshots)
6. **MTTR for node failure (mins/hours)**

Dashboards: Grafana/Loki summaries pushed centrally nightly; alerts on trends.

---

## 5️⃣ Board‑ready narrative (5 slides)

1. **Why now:** Latency is a silent tax; we’re over‑spending on cloud/CDN; Amazon wins on speed.
2. **What we’re doing:** Put compute and delivery where customers are—stores & DCs—using Proxmox on commodity gear.
3. **Proof:** Pilot results—100 ms faster → +X% revenue, \$Y saved in egress, 0 outages on WAN loss.
4. **Plan & risk:** 0‑5‑30‑90 rollout, governance pillars, automated compliance.
5. **Ask:** \$Z cap‑ex, 10 FTE cross‑functional team, payback <18 months.

---

## 6️⃣ Risk register (and the auto‑mitigations)

| Risk                    | Mitigation baked in                        |
| ----------------------- | ------------------------------------------ |
| Node dies in peak hours | Ceph 3× replica + live‑migration           |
| Patch breaks store      | Canary waves + fast rollback via Git tag   |
| WAN outage              | Local DNS/APIs, queued sync jobs           |
| Audit fails             | Automated evidence exports; policy‑as‑code |
| Drift / config sprawl   | Immutable bundles, Git as SOoT             |

---

## 7️⃣ Timeline & dollars (visual you can paste)

```
Q1  Prep & Pilot (5 stores)       $  75k  (gear + labor)
Q2  Prove @ 30 stores             $ 180k
Q3  Scale wave 1 (30 stores)      $ 180k
Q4  Scale wave 2 (30 stores)      $ 180k
-----------------------------------------
YEAR 1 CAP-EX / OPEX              ≈ $615k

Projected Year-1 Upside           $35M+ revenue lift + $54k cost saved
```

*(Adjust “gear” line if you choose GPU nodes or higher tiers.)*

---

## 8️⃣ This week’s action list

1. **Finalize KPIs & baselines** (latency, conversion, uptime).
2. **Draft the board deck** with the 5‑slide narrative above.
3. **Lock the pilot scope**: 2 workloads, 5 stores, 8‑week clock.
4. **Stand up the repo & CI/CD** (even if empty) to anchor governance.
5. **Book the Phase‑gate reviews** now (Pilot exit, 30‑store exit).

---

### 🎁 Series wrap & what’s next

* **Appendix pack (on request):** BOM spreadsheet, CI pipeline YAML, audit evidence script.
* **Office hours / webinar?** If there’s interest, I’ll host a live walkthrough.
* **Spin‑offs:**

    * “Edge ML Ops: packaging & shipping models to 500 sites”
    * “From closets to parking lots: EV chargers & edge compute”
    * “Design patterns for zero‑trust store networks”

> **Thank you for riding along.** This wasn’t about “new tech”—it was about rediscovering what made the Internet (and great retail) work: put value as close to the customer as you can, and don’t pay rent on distance.

*Hit follow, drop questions in the comments, or DM for the appendix. Your closets are ready to scale.*