Visualisation des tendances mensuelles

Before plotting, aggregate the sales DataFrame to monthly totals. Group by the year_month Period column and sum revenue. Convert the PeriodIndex to a DatetimeIndex with .to_timestamp() so Matplotlib's date formatter renders tick labels correctly. Sorting by index ensures the line chart flows left to right chronologically.

Create a Figure and Axes with plt.subplots(figsize=(12, 5)) and call ax.plot() to draw the monthly revenue trend. Use marker='o' to place a dot at each month so individual data points are visible even when the line is smooth. Set the colour and line width to make the primary trend stand out from any overlays you add later.

A raw monthly revenue line can appear jagged due to seasonal noise. Overlay a 3-month rolling average computed with monthly.rolling(3).mean() to show the underlying trend more clearly. Plot it as a dashed line in a contrasting colour so it stands out from the raw series without obscuring it.

Use ax.annotate() to mark the month with the highest revenue. Find the peak with monthly.idxmax() and its value with monthly.max(). The annotation arrow points from a text label to the peak data point, making the insight immediately visible to any reader of the chart, including those who receive it as a PNG or PDF.

Similarly, mark the month with the lowest revenue using monthly.idxmin() to flag potential seasonality or disruptions. Annotations on both the peak and trough provide a complete story: they let stakeholders see not only the best month but also how far revenue can dip, which informs cash-flow planning.

Matplotlib's default date tick labels often overlap or use an unreadable format. Import mdates from matplotlib and set a MonthLocator and DateFormatter to control tick frequency and label format. Rotating tick labels by 45 degrees prevents overlap on a 12-month axis.

Highlight the Q4 period (October–December) with a shaded background using ax.axvspan(). This makes seasonal patterns immediately visible: if the peak consistently falls within the shaded zone, Q4 is a key selling period worth targeting with promotions. The alpha parameter controls the shading transparency.

Create a multi-line chart where each line represents a product category. Pivot the monthly-by-category DataFrame so columns are categories and the index is time, then iterate over columns and call ax.plot() for each. Matplotlib automatically assigns a different colour to each line when called repeatedly on the same Axes.

A stacked bar chart shows the total monthly revenue while also revealing each category's contribution to the stack. Use monthly_cat.plot(kind='bar', stacked=True, ax=ax) directly on the pivoted DataFrame. Stacked bars are best when you have fewer than six categories, otherwise the colours become hard to distinguish.

Plot the month-over-month percentage growth rate alongside the revenue to give context: a visually small bar in January might represent a 20 % growth jump from the prior month. Compute growth with monthly.pct_change() * 100 and plot it on a secondary y-axis using ax.twinx() to share the same x-axis without conflating units.

Use fig.savefig('monthly_trend.png', dpi=150, bbox_inches='tight') to produce a high-resolution PNG suitable for slide decks or web dashboards. The bbox_inches='tight' argument trims excess white space, and dpi=150 gives a sharp image without an excessively large file. Save to PDF as well for print-quality reports.

Test your understanding of Data Analysis concepts from this lesson.

In this lesson you learned: computing monthly revenue totals and plotting them as a line chart, overlaying a rolling average and annotating peak and trough months, and saving publication-quality figures to PNG and PDF. Next up we explore sessionisation and event sequencing in user behaviour data.