أفضل ذكاء اصطناعي لـ Analyze a CSV file

Loaded 11,847 rows (153 dropped — see below). Three data quality issues to flag before any analysis: 1. The "signup_date" column is stored as text in mixed formats (M/D/YYYY and YYYY-MM-DD coexist). I parsed both — but you should standardize this in the source system, since 153 rows had unparseable dates and got dropped. 2. `order_count` has 412 zeros and 89 nulls. Treating "no orders" and "data missing" the same will skew retention numbers — I treated them separately below. 3. `region` has "West", "west", and "WEST" as three distinct values (1,043 / 489 / 12 rows). Normalized to lowercase; the 12 "WEST" rows look like an upstream import bug worth tracing. Five findings: 1. Customer cohort from Q3 2024 has 41% retention at month 12 — significantly higher than the 29% average across other quarters. (Code: `df.groupby('signup_quarter').agg({'retained_12mo': 'mean'})`) 2. Average order value is $156 — but median is $89. The mean is pulled by 312 orders above $1,000, mostly from a small enterprise segment (~3% of customers). Plan for two distinct customer types in any retention work. 3. The "West" region looks strong on average ($178 AOV) but has higher churn than the East ($142 AOV, lower churn). Suggests West attracts higher-spend but less-loyal customers. 4. Customers acquired through referral have 2.1× the LTV of paid-acquisition customers, but only 18% of total signups come from referral. There's a growth lever here. 5. 89 nulls in order_count correlate strongly with missing email (78 of 89). Likely a single broken integration in late 2025 — flag the data engineer. Best chart: a cohort retention heatmap by signup-quarter × month-since-signup. Code shown. What this data can't tell you: why Q3 2024 retention is so much higher. Could be product changes, a marketing campaign, or selection bias from how those customers were acquired. Pair with marketing spend data and product release notes before drawing conclusions.