Support ticket deflection — pilot results¶
Four weeks of the new help-center widget, versus the four weeks before it shipped. Synthetic data for the ship.page Jupyter guide — /use-cases/jupyter-notebook.
In [1]:
import pandas as pd
import matplotlib.pyplot as plt
weeks = [f"W{i}" for i in range(33, 41)]
tickets = [412, 388, 401, 375, 301, 276, 264, 249]
widget = [0, 0, 0, 0, 118, 132, 141, 150]
df = pd.DataFrame({"week": weeks, "tickets": tickets, "widget_sessions": widget})
df["tickets_7d_avg"] = df["tickets"].rolling(2).mean().round(0)
df
Out[1]:
| week | tickets | widget_sessions | tickets_7d_avg | |
|---|---|---|---|---|
| 0 | W33 | 412 | 0 | NaN |
| 1 | W34 | 388 | 0 | 400.0 |
| 2 | W35 | 401 | 0 | 394.0 |
| 3 | W36 | 375 | 0 | 388.0 |
| 4 | W37 | 301 | 118 | 338.0 |
| 5 | W38 | 276 | 132 | 288.0 |
| 6 | W39 | 264 | 141 | 270.0 |
| 7 | W40 | 249 | 150 | 256.0 |
In [2]:
df.style.format({"tickets_7d_avg": lambda v: "—" if pd.isna(v) else f"{v:.0f}"})
Out[2]:
| week | tickets | widget_sessions | tickets_7d_avg | |
|---|---|---|---|---|
| 0 | W33 | 412 | 0 | — |
| 1 | W34 | 388 | 0 | 400 |
| 2 | W35 | 401 | 0 | 394 |
| 3 | W36 | 375 | 0 | 388 |
| 4 | W37 | 301 | 118 | 338 |
| 5 | W38 | 276 | 132 | 288 |
| 6 | W39 | 264 | 141 | 270 |
| 7 | W40 | 249 | 150 | 256 |
The picture¶
Tickets trend down from the week the widget shipped; widget sessions climb in step. Causation is a claim for the review meeting, not for this notebook.
In [3]:
fig, ax = plt.subplots(figsize=(7, 3.2))
ax.plot(df["week"], df["tickets"], marker="o", label="Tickets")
ax.plot(df["week"], df["widget_sessions"], marker="s", label="Widget sessions")
ax.axvline(3.5, color="grey", linestyle="--", linewidth=1)
ax.text(3.6, 400, "widget ships", fontsize=9, color="grey")
ax.set_ylabel("per week")
ax.legend()
ax.spines[["top", "right"]].set_visible(False)
plt.tight_layout()
plt.show()
