performance-analytics（绩效分析）

Performance Analytics Skill

Frameworks for measuring, reporting, and optimizing marketing performance across channels and campaigns.

Key Marketing Metrics by Channel

Email Marketing

Social Media

Paid Advertising (Search and Social)

SEO / Organic Search

Content Marketing

Overall Marketing / Pipeline

Reporting Templates and Dashboards

Weekly Marketing Report

Quick-scan format for team standups:

• Top 3 metrics with week-over-week change
• What worked this week (1-2 bullet points with data)
• What needs attention (1-2 bullet points with data)
• This week's priorities (3-5 action items)

Monthly Marketing Report

Standard stakeholder report:

1. Executive summary (3-5 sentences)
2. Key metrics dashboard (table with MoM and target comparison)
3. Channel-by-channel performance summary
4. Campaign highlights and results
5. What worked and what did not (with hypotheses)
6. Recommendations and next month priorities
7. Budget spend vs. plan

Quarterly Business Review (QBR)

Strategic review for leadership:

1. Quarter performance vs. goals
2. Year-to-date trajectory
3. Channel ROI analysis
4. Campaign performance summary
5. Competitive and market observations
6. Strategic recommendations for next quarter
7. Budget request and allocation plan
8. Key experiments and learnings

Dashboard Design Principles

• Lead with the metrics that map to business objectives (not vanity metrics)
• Show trends over time, not just point-in-time snapshots
• Include comparison context: prior period, target, benchmark
• Use consistent color coding: green (on track), yellow (at risk), red (off track)
• Group metrics by funnel stage or business question
• Keep dashboards to one page/screen — detail goes in appendix
• Update cadence should match decision cadence (real-time for paid, weekly for content)

Trend Analysis and Forecasting

Trend Identification

When analyzing performance data, look for:

1. Directional trends: is the metric consistently going up, down, or flat over 4+ periods?
2. Inflection points: where did performance change direction and what happened then?
3. Seasonality: are there predictable patterns by day of week, month, or quarter?
4. Anomalies: one-time spikes or drops — what caused them and are they repeatable?
5. Leading indicators: which metrics change first and predict future outcomes?

Trend Analysis Process

1. Chart the metric over time (at least 8-12 data points for meaningful trends)
2. Identify the overall direction (upward, downward, flat, cyclical)
3. Calculate the rate of change (is it accelerating or decelerating?)
4. Overlay key events (campaigns launched, product changes, market events)
5. Compare to benchmarks or targets
6. Identify correlations with other metrics
7. Form hypotheses about causation (and plan tests to validate)

Simple Forecasting Approaches

• Linear projection: extend the current trend line forward (useful for stable metrics)
• Moving average: smooth out noise by averaging the last 3-6 periods
• Year-over-year comparison: use last year's pattern as a baseline, adjusted for growth rate
• Funnel math: forecast outputs from inputs (e.g., if we generate X leads at Y conversion rate, we will get Z customers)
• Scenario modeling: create best case, expected case, and worst case projections

Forecasting Caveats

• Short-term forecasts (1-3 months) are more reliable than long-term
• Forecasts based on fewer than 12 data points should be flagged as low confidence
• External factors (market shifts, competitive moves, economic changes) can invalidate trend-based forecasts
• Always present forecasts as ranges, not exact numbers

Attribution Modeling Basics

What Is Attribution?

Attribution determines which marketing touchpoints get credit for a conversion. This matters because buyers typically interact with multiple channels before converting.

Common Attribution Models

Attribution Practical Guidance

• Start with last-touch attribution if you have no model in place — it is the simplest and most actionable
• Compare first-touch and last-touch to understand which channels drive awareness vs. conversion
• Use position-based (U-shaped) as a reasonable middle ground for most B2B companies
• Data-driven attribution requires high conversion volume to be statistically meaningful
• No model is perfect — use attribution directionally, not as absolute truth
• Multi-touch attribution is better than single-touch, but any model is better than none

Attribution Pitfalls

• Do not optimize one channel in isolation based on single-touch attribution
• Awareness channels (display, social, PR) will always look bad in last-touch models
• Conversion channels (search, retargeting) will always look bad in first-touch models
• Self-reported attribution ("how did you hear about us?") provides useful qualitative color but is unreliable as quantitative data
• Cross-device and cross-channel tracking gaps mean attribution data is always incomplete

Optimization Recommendations Framework

Optimization Process

1. Identify: which metrics are underperforming vs. target or benchmark?
2. Diagnose: where in the funnel is the problem? (impressions, clicks, conversions, retention)
3. Hypothesize: what is causing the underperformance? (audience, message, creative, offer, timing, technical)
4. Prioritize: which fixes will have the biggest impact with the least effort?
5. Test: design an experiment to validate the hypothesis
6. Measure: did the change improve the metric?
7. Scale or iterate: roll out wins broadly; iterate on inconclusive or failed tests

Optimization Levers by Funnel Stage

Prioritization Framework

Rank optimization ideas on two dimensions:

Impact (how much will this move the metric?):

• High: directly addresses the primary bottleneck
• Medium: addresses a contributing factor
• Low: incremental improvement

Effort (how hard is this to implement?):

• Low: copy change, targeting adjustment, simple A/B test
• Medium: new creative, landing page redesign, workflow change
• High: new tool, cross-team project, major content production

Priority order:

1. High impact, low effort (do immediately)
2. High impact, high effort (plan and resource)
3. Low impact, low effort (do if capacity allows)
4. Low impact, high effort (deprioritize)

Testing Best Practices

• Test one variable at a time for clean results
• Define the success metric before launching the test
• Calculate required sample size before starting (do not end tests early)
• Run tests for a minimum of one full business cycle (typically one week for B2B)
• Document all tests and results, regardless of outcome
• Share learnings across the team — failed tests are valuable information
• A test that confirms the status quo is not a failure — it builds confidence in your current approach

Continuous Optimization Cadence

• Daily: monitor paid campaigns for budget pacing, anomalies, and disapproved ads
• Weekly: review channel performance, pause underperformers, scale winners
• Bi-weekly: refresh ad creative and test new variants
• Monthly: full performance review, identify new optimization opportunities, update forecasts
• Quarterly: strategic review of channel mix, budget allocation, and targeting strategy