Marketing Ops · Q3 2026

Campaign
Incremental Lift

用数据证明每一场活动的真实价值Prove the real value of every campaign with data

Simon Ni · July 2026

核心问题CORE PROBLEM

Marketing Ops当前没有衡量
活动质量的关键指标Marketing Ops currently lacks
a key metric to measure campaign quality

问题1Issue 1

ROI只看"花了多少赚了多少"
但无法回答：不做这个活动会怎样？ROI only shows "spend vs earn"
Can't answer: What if we didn't run it?

问题2Issue 2

活动重叠时归因混乱
多活动同时跑，谁的功劳？Overlap = attribution chaos
Multiple campaigns, who gets credit?

问题3Issue 3

CEO问"活动到底值不值"
我们答不上来CEO asks "are campaigns worth it?"
We can't answer
解决方案SOLUTION

Incremental Lift = 因果推断Incremental Lift = Causal Inference

实验组（活动组）Treatment Group

看到活动 + 可以参与
正常活动体验See campaign + can participate
Normal experience

对照组（Holdout）Control Group (Holdout)

看不到活动
自然行为基线No campaign exposure
Natural behavior baseline

Lift% = （活动组指标 − 对照组指标）/ 对照组指标 × 100%(Treatment metric − Control metric) / Control metric × 100%

Holdout比例：5-10%用户随机分入对照组，不影响业务大盘Holdout ratio: 5-10% users randomly assigned to control, minimal business impact

Q3 2026 OKRs — Campaign Part

Key Results

#维度Dimension指标Metric公式FormulaTarget
KR1规模ScaleTotal Incremental ProfitΣ(Treatment Rev − Control Rev) − Cost[BI]
KR2质量QualityeFTD Incremental Lift %(Treatment eFTD% − Control eFTD%) / Control eFTD%≥ 30%
KR3效率EfficiencyRevenue / CostIncremental Revenue / Total Campaign Cost≥ 5x
min 3x/region
KR4EUEU Site Campaign ROI QoQ↑EU Campaign Rev / EU Campaign CostQoQ ↑
KR5基建InfraHoldout基建上线 Infra Live分组+过滤系统Segmentation + filterAug 31
区域化策略REGIONAL STRATEGY

不同区域不同打法Different Regions, Different Plays

阶段Stage区域Regions策略StrategyKPI Focus
成熟MatureCIS / Korea提效降本，ROI优先Efficiency first, cut wasteROI ≥ 5x · Lift ≥ 30%
新兴GrowthSEA / LATAM / MENA增长优先，允许较低ROIGrowth first, lower ROI okeFTD Lift ≥ 20% · ROI ≥ 3x
衰退Decline流失严重区域High-churn regions召回为主，控制预算Win-back focus, tight budgetRecall 30d active ≥ 50%
合规ComplianceEU / Japan / HK合规优先，品牌教育为主Compliance first, education ledROI QoQ↑ · No violations
TIMELINE

落地节奏Execution Timeline

Aug W1-2
BI建分组系统BI builds segmentation
→
Aug 31
Holdout基建上线Holdout infra live
→
Sep W2
第一场带Holdout活动执行First campaign w/ holdout
→
Dec
Q3复盘，Lift数据完整Q3 review, full Lift data

BI侧BI Side

用户分组 + Holdout过滤
+ 自动Lift计算看板User segmentation + Holdout filter
+ Auto Lift dashboard

Ops侧Ops Side

活动SOP改造
强制标注Holdout比例Campaign SOP reform
Mandatory holdout % label
活动治理GOVERNANCE

有了Lift数据后如何管How to govern with Lift data

优秀Excellent

Lift ≥ 30%
ROI ≥ 5x

→ 加预算复制到其他区域→ Scale budget + clone to regions

达标Pass

Lift 20-30%
ROI 3-5x

→ 维持运行，持续优化→ Maintain, keep optimizing

不合格Fail

Lift < 10%
ROI < 3x

→ 立刻停止，复盘问题→ Stop immediately, post-mortem
历史复盘PAST REVIEW

没有Lift时我们踩了什么坑What went wrong without Lift data

虚高ROIInflated ROI

全产品线收入算活动功劳
实际增量可能很小Full product revenue credited to campaign
Actual incremental may be tiny

活动打架Campaign Clash

同人群同时被多活动覆盖
哪个活动有效无法判断Same users in multiple campaigns
Can't tell which one works

低效活动续命Zombie Campaigns

没有"差"的标准
低效活动长期占用预算No "bad" benchmark
Low-ROI campaigns keep running

CEO质疑CEO Doubt

"活动到底有没有用？"
拿不出因果证据"Do campaigns even work?"
No causal evidence to show
价值VALUE

Incremental Lift带来的价值Value of Incremental Lift

CEO有答案CEO Gets Answers

"这场活动多赚了$X"
因果推断，不是相关性"This campaign made extra $X"
Causal, not correlation

砍掉废活动Kill Bad Campaigns

Lift < 10%立刻停
预算留给有效的Lift < 10% = stop now
Budget goes to winners

团队有方向Team Has Direction

每场活动目标清晰：
Lift ≥ 30%就是好活动Clear goal per campaign:
Lift ≥ 30% = good campaign

区域可比Cross-Region Compare

统一Lift指标
CIS vs SEA vs EU一目了然Unified Lift metric
CIS vs SEA vs EU at a glance
SUMMARY

一句话One Line

"不做活动会怎样" — 这是我们Q3要回答的问题"What if we didn't run it?" — This is the Q3 question

Aug 31
基建上线Infra live
→
Sep W2
首场执行First run
→
Dec
Q3复盘有数据Q3 review w/ data
1/10 · ← → · F fullscreen