GRAI Retrospective Model

Four-Step Template

Retrospective Topic: 618 Mid-Year Sale Campaign Retrospective
Retrospective Date: 2026-06-25
Participants: Operations team, Product team, Design team, Engineering team

---

GRAI Four-Step Method	My Notes & Answers
Goal — Review Goals • What is the purpose goal? • What is the task goal (quantified)? • What is the action goal (method)?	Purpose goal: Boost brand influence through 618 sale, acquire and reactivate users, lay foundation for H2 growth Task goal: GMV reach $5M, 20K new users, 8% campaign conversion rate Action goal: Prepare 1 month in advance, 3 rounds of pre-heat, main venue + sub-venue format, partner with 5 brand collaborators
Result — Evaluate Results • Completion of purpose / task / action goals? • 1-3 highlights in process / results? • 1-3 shortcomings in process / results?	Goal completion: GMV $6.2M (124%), 23K new users (115%), 9.2% conversion rate (115%) Highlights: 1. First attempt at live-stream commerce, contributed 30% of GMV, exceeded expectations 2. Existing user repurchase rate reached 35%, indicating improving user stickiness 3. Cross-department collaboration was smoother than last year; no last-minute requirement changes like before Shortcomings: 1. Server crashed for 20 minutes on the first day, estimated loss around $20K GMV 2. Some product inventory was inaccurately estimated, best-sellers sold out early 3. Customer support couldn't keep up; user complaints tripled from usual
Analysis — Root Cause Analysis • What caused success? • What caused failure? • Which analysis method was used?	Success reasons: 1. Precise product selection for live stream, high host cooperation, plus platform traffic support (main factor) 2. Pre-planned re-engagement campaign for existing users, reactivated a group of dormant users 3. Started stress testing and full-link rehearsal 2 weeks early; overall stability much better than last year Failure reasons: 1. Server capacity estimation was too conservative; prepared for 1.5x last year's peak, but actual traffic was 2.3x (live-stream traffic exceeded expectations) 2. Inventory forecast model only referenced last year's data, didn't account for live-stream uplift 3. Customer support team only added 2 staff, insufficient estimation Analysis method: ☑ 5 Whys Analysis □ Fishbone diagram □ SWOT analysis □ Other:
Insight — Lessons Learned • What methodologies / SOPs can be documented? • What Q&As can be compiled?	Methodology / SOP: 1. Full-link stress testing SOP before big sales (documented) 2. Live-stream product selection and operations SOP (validated effective this time) 3. Cross-department collaboration mechanism (weekly meeting + daily standup + real-time chat group) Q&A: Q: How to estimate server capacity for big sales? A: Reference historical peak × 2x, multiply by another 1.5x if there's a new traffic channel (e.g., live-stream). It's better to over-prepare than under-prepare. Q: What if inventory forecast is inaccurate? A: Phased stock — 60% in first batch, replenish dynamically based on first 2 days' sales data, agree on 48-hour replenishment mechanism with suppliers. Q: How to estimate customer support staffing? A: Prepare 3-5x usual staffing; also prepare auto-reply scripts and FAQ to improve efficiency.

---

G — Goal Review

Purpose Goal

Why are we doing this? What is the underlying purpose and meaning?

Achieve three objectives through the 618 mid-year sale:

1. Drive performance, hit 40% of H1 GMV target
2. Acquire new users through promotional discounts
3. Brand exposure, increase industry visibility

Task Goal (Quantified)

What specific measurable results are to be achieved?

Metric	Target	Unit
GMV	500	$10K (5M)
New users	20	K users
Conversion rate	8	%
AOV	200	$

Action Goal (Method)

What methods and paths are planned to achieve the goals?

• Launch preparation 1 month early, establish cross-department task force
• 3 rounds of pre-heat: Blog post → Community referral → Splash-screen ads
• Main venue + 5 themed sub-venues
• First attempt at live-stream commerce: 2 sessions daily, invite 3 top influencers
• Partner with 5 brand collaborators for joint promotion

---

R — Result Evaluation

Goal Completion

Goal type	Target	Actual	Completion	Score
Purpose goal	Hit 40% of H1 target	Hit 48% of H1 target	120%	⭐⭐⭐⭐⭐
Task goal	GMV $5M	GMV $6.2M	124%	⭐⭐⭐⭐⭐
Action goal	Execute as planned	Mostly on plan, live-stream effect exceeded expectations	110%	⭐⭐⭐⭐

Highlights (1-3)

1. Live-stream exceeded expectations: Contributed 30% of GMV — the biggest pleasant surprise of this campaign, validated the feasibility of live-stream e-commerce
2. High existing user repurchase rate: Reached 35%, indicating daily user operations are working and user stickiness is improving
3. Smooth cross-department collaboration: No chaotic situations like last year; advance rehearsal and clear division of labor worked

Shortcomings (1-3)

1. Server downtime: Server crashed for 20 minutes at 10:00 on day one; conservative estimate $20K GMV loss, also damaged user experience
2. Best-sellers out of stock: TOP 3 products sold out 2 days early; traffic came later but nothing to sell — estimated loss $15K
3. Customer support lagged: User wait time went from 30 seconds usual to 5 minutes; complaints tripled, damaging brand reputation

---

A — Analysis

Success Reasons

Reason	Type	Contribution
Precise live-stream product selection + platform traffic support	Combined objective+subjective	⭐⭐⭐⭐⭐
Well-planned existing user re-engagement, reactivated dormant users	Subjective	⭐⭐⭐⭐
Advance rehearsal, mature cross-department collaboration	Subjective	⭐⭐⭐⭐

Failure Reasons

Reason	Type	Impact
Server capacity estimate too conservative; didn't account for live-stream uplift traffic	Subjective (insufficient experience)	🔴 High
Inventory forecast only referenced historical data, no new variable model	Subjective (methodology issue)	🟡 Medium
Insufficient support staffing; misjudged campaign scale	Subjective (inaccurate estimate)	🟡 Medium

5 Whys Deep Dive (Optional)

Problem: Why did the server crash?
  → Because traffic peak exceeded server capacity
  → Why did traffic exceed?
    → Because live-stream brought far more traffic than expected
    → Why wasn't live-stream traffic anticipated?
      → Because there was no live-stream last year, no historical data to reference
      → Why not make a more conservative estimate?
        → Because of cost concerns, wanted "just enough" — and it backfired
        → Root cause: Insufficient traffic risk assessment for new channels; capacity strategy too conservative. Should be "better to over-prepare than under-prepare"

---

I — Insight

Reusable Methodology / SOP

Methods that can be reused next time:

1. Big sale full-link stress testing SOP — Start 2 weeks early, at least 3 rounds
2. Live-stream product selection and operations SOP — Selection criteria, host coordination, traffic handling
3. Cross-department collaboration mechanism — Weekly meeting + daily standup + real-time chat group + clear RACI

Processes to formalize:

• Step 1: 1 month before sale, complete all plan reviews and resource confirmation
• Step 2: 2 weeks before sale, begin full-link stress testing and simulation rehearsal
• Step 3: 1 week before sale, all staff on standby, do final checks
• Step 4: During sale, daily retrospective, quickly adjust strategy
• Step 5: Within 1 week after sale, complete full retrospective document

Q&A Compilation

Q Question	A Answer	Applicable scenario
How to estimate server capacity?	Historical peak × 2, × 1.5 more if new channels exist	All large-scale promotions
How much inventory to stock?	60% in first batch, replenish based on real-time data	E-commerce campaigns
How to staff customer support?	3-5x usual, prepare auto-replies	All large-scale events
How to handle live-stream traffic?	Prepare landing pages in advance, ensure server can handle	Live-stream commerce scenarios

Action Plan

Action Item	Owner	Deadline	Priority	Status
Compile big sale SOP document, company-wide rollout	Ops lead	2026-07-10	🔴 High	⬜
Upgrade server capacity strategy, introduce auto-scaling	Eng lead	2026-07-15	🔴 High	⬜
Build intelligent inventory forecast model	Data team	2026-08-01	🟡 Medium	⬜
Expand support team + establish flexible staffing mechanism	HR + Support lead	2026-07-31	🟡 Medium	⬜
Validate improvements during Double 11 retrospective	All	2026-11-20	🟢 Low	⬜

---

📌 Quick Reference Card

Letter	Chinese	English	Core Question
G	Review goals	Goal	What was the original goal?
R	Evaluate results	Result	How did actual results turn out? Highlights and shortcomings?
A	Root cause analysis	Analysis	Why did it succeed / fail? What's the root cause?
I	Lessons learned	Insight	What was learned? How to do better next time?