Time-series trends (daily Sales & Traffic)

Use only when the artifact is validated as daily grain (one or more rows per calendar day). A weekly or monthly period can have a date field without supporting DoD, MA7, weekday/weekend, or seven-day WoW calculations.

When this section applies

• Portfolio-by-day: If rows are ASIN × day, sum ordered product sales, sessions, units (and B2B columns if present) per day before trend metrics. Document that double-counting is avoided by summing only at ASIN grain then grouping by date.
• Already portfolio daily: One row per day — use directly after validating date parsing.

Prerequisites

• Map date (or start/end — user may need to derive a single date)
• Map ordered_product_sales (or equivalent revenue for the trend — name in report)
• Optional: sessions, units, unit session % for parallel trend tables or narrative

Analysis window

• Default narrative: last N calendar days (legacy example N = 14 — user overrides).
• Require minimum days for MA/WoW (e.g. ≥ 3 for rolling MA with min_periods; ≥ 14 for classic last-7 vs prior-7 WoW). State when the window is shorter.

Core computations (agent-authored)

1. Sort by date ascending; drop rows with invalid dates.
2. Optional data-quality filter — e.g. exclude days with zero sessions if the user agrees it flags incomplete downloads — document if applied (do not apply silently by default).
3. Day-over-day (DoD): pct_change on revenue (and optionally sessions) × 100.
4. Moving average: e.g. 7-day rolling mean of revenue (min_periods 3–7 per user); label MA7 in output.
5. Deviation from MA: (revenue - MA7) / MA7 * 100 where MA7 > 0.
6. Anomaly flags (tunable): Default to ±20% deviation from MA7 unless the user chooses another threshold. Record the selected threshold in the metrics output and report.
7. Week-over-week (WoW): If ≥ 14 days in window: sum revenue last 7 vs prior 7; pct_change. If < 14 days, compute partial WoW or mark N/A.
8. Weekday vs weekend: Classify dayofweek >= 5 as weekend; compare average daily revenue weekend vs weekday; report gap % vs weekday average.
9. Day-of-week pattern: Mean revenue by Monday … Sunday (calendar order in table); call out best/worst day name.

Optional B2B on trends

• If B2B sales column exists and is summed per day, add B2B revenue trend or B2B share of day time series in the same section.

Deliverables (markdown; charts optional)

• Table: date, revenue, optional sessions/units, DoD %, MA7, deviation %, anomaly label
• Summary bullets: period total, daily average, peak day, trough day, WoW %, weekday vs weekend gap, anomaly count
• Day-of-week table: avg revenue per weekday name
• Charts (if tooling allows): line chart revenue + MA7; optional bar chart DOW averages — not required for skill compliance

Pitfalls

• Mixing currencies or marketplaces in one series without user confirmation.
• Rolling MA on non-contiguous dates — consider reindexing to calendar or flag gaps.
• ASIN-level daily file: portfolio trend must group by date first.