示例产出 · Sample output — 这是 skill 的示例产出，完整方法论在付费 skill 内。 This page is a sample output of the skill; the full methodology lives inside the paid skill. All data below is fictional.
ANALYSIS REPORT · 数据分析报告

Emberline Coffee Roasters

Q3 2026 Sales Analysis — Why revenue dipped in August, and what to do about it

Period: Jul 1 – Sep 30, 2026 (vs Q2 2026) Dataset: orders.csv · 38,210 rows · 1 row = 1 order Prepared: October 6, 2026 Question: Why did revenue dip in August?
EXECUTIVE SUMMARY · 执行摘要

The short version

HEADLINE METRICS · 核心指标

Q3 2026 at a glance

REVENUE
$1.285M
−5.4% vs Q2 2026
ORDERS
38,210
−3.1% vs Q2 2026
AVG ORDER VALUE
$33.63
−2.3% vs Q2 2026
REPEAT PURCHASE RATE
41.8%
+1.4 pts vs Q2 2026
KEY FINDINGS · 关键发现

What the data shows

  1. $1.285M revenue, −5.4% vs Q2 — the first quarterly decline in two years, but concentrated in a single month (August).
  2. The August dip (−$56K) is fully reconciled: wholesale pause −$28K (50%), subscription pause spike −$19K (34%), DTC softness −$9K (16%). No unexplained remainder.
  3. The subscription base is healthy: repeat rate rose to 41.8%. The July cohort's Month-1 retention dipped to 71% (vs 76–78% baseline) — a pause-flow problem after the price change, not churn.
  4. Wholesale concentration risk: one partner represents 19% of Q3 revenue — the single largest swing factor in the business.
  5. September recovered to $445K — the dip looks like a one-off event, not the start of a trend (moderate evidence: one month of recovery so far).
EVIDENCE · 数据证据

The charts behind the findings

August was the whole story — September already bounced back
Monthly revenue, Apr – Sep 2026 · $ thousands
$360K $400K $440K $480K Apr May Jun Jul Aug Sep Aug: $392K −$56K (−12.5%) vs Jul

Jul $448K → Aug $392K → Sep $445K. The dip is isolated to August; the Q3 total ($1.285M) is August's story.

Subscription is the engine; wholesale is the swing factor
Q3 2026 revenue by channel · $ thousands · bars start at zero
0 $100K $200K $300K $400K $500K Subscription $486K · 37.8% DTC one-time $411K · 32.0% Wholesale $244K · 19.0% Retail café $143K · 11.2%

Channels sum to $1.285M — reconciled to the headline total.

Four in ten customers buy again
Customer mix, Q3 2026 · share of active customers
Q3 2026 active customers Repeat buyers · 42% New subscribers · 34% Wholesale partners · 15% Reactivated · 9%

Repeat rate 41.8% (+1.4 pts vs Q2) — the base is loyal; the gap is acquisition.

The August dip, decomposed: −$56K, fully explained
Revenue bridge Jul → Aug 2026 · $ thousands · segment changes sum to the total change
$460K $430K $400K $370K Jul $448K −$28K −$19K −$9K Aug $392K July baseline Wholesale partner paused orders Pause-rate spike after price change DTC softness heatwave shipping August actual

−28 − 19 − 9 = −$56K exactly. No unexplained remainder — the diagnostic equation balances.

Retention is stable — except the July cohort's first month
Subscription cohort retention · % of cohort still subscribed
Cohort (signup month)SizeMonth 0Month 1Month 2Month 3
Apr 20262,280100%77%62%53%
May 20262,310100%76%61%52%
Jun 20262,140100%78%64%55%
Jul 20262,050100%71% ↓——

The July cohort (first billed at the new price) retained 5–7 points worse in Month 1 — consistent with the pause-rate spike. Read vertically: the product retains; read the July row: the price change stung.

RECOMMENDATIONS · 建议

What to do, ranked by impact and effort

ActionExpected impactEffortTied to
Win-back call + volume incentive for the paused wholesale partner — founder-led, this week. +$28K/mo revenue recovery Low Finding 2, 4
Add a "skip-a-month" save offer to the subscription pause flow before the cancel/pause confirms. ≈ −30% pause rate ≈ +$6K/mo Medium Finding 3
Heat-season shipping upgrade — insulated packaging for DTC orders Jun–Aug so heatwaves don't force shipping pauses. Protects ≈ $9K seasonal dip Medium Finding 2
Reallocate 15% of DTC ad spend to subscription trial offers — acquisition is the gap, and trials convert at the healthy 41.8% repeat base. Lifts acquisition; LTV-tested Low Finding 5
METHODOLOGY & CAVEATS · 方法与局限

How this was done — and what it can't prove

METHODOLOGY

  • Data profiling: checked shape (38,210 orders), types, missingness (0.4% missing region codes — excluded and disclosed), duplicates (212 removed, latest kept), and date coverage (full months, Apr–Sep).
  • Transparent cleaning: every transformation logged with before/after row counts; no silent drops. Test orders and staff purchases removed by explicit filter.
  • Analysis: descriptive summaries first (totals, rates, quarter-over-quarter), then diagnostic segmentation — the August change was decomposed until segments summed exactly to −$56K.
  • Charting: one message per chart, axes labeled with units, bar charts on a zero baseline, titles state the finding.

CAVEATS

  • One month of recovery (September) is not a trend — the "one-off, not structural" read is moderate evidence, to be re-checked with October data.
  • Wholesale analysis rests on a single partner's pause — the $28K attribution comes from that partner's order history, not a controlled comparison.
  • Pause-rate spike is associated with the July price change (timing + July-cohort dip), not proven causal — a survey of paused subscribers would confirm.
  • External factors (competitor moves, weather beyond the shipping pause) are outside this dataset and not modeled.