tsb — Marketing Attribution ROAS

Skills you'll use: melt, multi-column merge, KPI columns, groupby().agg(), pivotTableFull, and top-N selection.

1 · Normalize spend and join conversions

Turn wide channel spend into a tidy table, join conversion facts, and compute ROAS and cost per order.

TypeScript

import { DataFrame, melt, merge } from "tsb";

const joined = merge(spend, conversions, { on: ["date", "region", "channel"], how: "left" }).assign({
  roas: (df) => df.col("revenue").div(df.col("spend")),
  cost_per_order: (df) => df.col("spend").div(df.col("orders")),
});
console.log(joined.head(8).toString());

const channelRollup = joined
  .groupby("channel")
  .agg({ spend: "sum", revenue: "sum", orders: "sum", roas: "mean", cost_per_order: "mean" }, false)
  .sortValues("roas", false);
console.log("\nChannel rollup:");
console.log(channelRollup.toString());

import pandas as pd

spend_wide = pd.DataFrame({
    "date": ["2024-05-01", "2024-05-01", "2024-05-02", "2024-05-02"],
    "region": ["NA", "EU", "NA", "EU"],
    "search": [1200, 900, 1350, 980],
    "social": [800, 740, 880, 790],
    "email": [300, 260, 320, 280],
})
conversions = pd.DataFrame([
    {"date": "2024-05-01", "region": "NA", "channel": "search", "orders": 62, "revenue": 8680},
    {"date": "2024-05-01", "region": "NA", "channel": "social", "orders": 35, "revenue": 3920},
    {"date": "2024-05-01", "region": "NA", "channel": "email", "orders": 18, "revenue": 2520},
    {"date": "2024-05-01", "region": "EU", "channel": "search", "orders": 44, "revenue": 5280},
    {"date": "2024-05-01", "region": "EU", "channel": "social", "orders": 31, "revenue": 3100},
    {"date": "2024-05-01", "region": "EU", "channel": "email", "orders": 16, "revenue": 1760},
    {"date": "2024-05-02", "region": "NA", "channel": "search", "orders": 71, "revenue": 9940},
    {"date": "2024-05-02", "region": "NA", "channel": "social", "orders": 42, "revenue": 5040},
    {"date": "2024-05-02", "region": "NA", "channel": "email", "orders": 21, "revenue": 3150},
    {"date": "2024-05-02", "region": "EU", "channel": "search", "orders": 49, "revenue": 6370},
    {"date": "2024-05-02", "region": "EU", "channel": "social", "orders": 33, "revenue": 3465},
    {"date": "2024-05-02", "region": "EU", "channel": "email", "orders": 19, "revenue": 2280},
])
spend = spend_wide.melt(id_vars=["date", "region"], value_vars=["search", "social", "email"], var_name="channel", value_name="spend")
joined = spend.merge(conversions, on=["date", "region", "channel"], how="left")
joined["roas"] = joined["revenue"] / joined["spend"]
joined["cost_per_order"] = joined["spend"] / joined["orders"]
print(joined.head(8))
print(joined.groupby("channel", as_index=False).agg({"spend":"sum", "revenue":"sum", "orders":"sum", "roas":"mean", "cost_per_order":"mean"}))

Click ▶ Run to execute

Ctrl+Enter to run · Tab to indent

2 · Compare region/channel performance

Create a pivot-table scorecard and list the strongest daily placements.

TypeScript

import { DataFrame, nlargestDataFrame, pivotTableFull } from "tsb";

const joined = DataFrame.fromRecords([
  { region: "NA", channel: "search", spend: 1200, revenue: 8680, orders: 62 },
  { region: "NA", channel: "social", spend: 800, revenue: 3920, orders: 35 },
  { region: "NA", channel: "email", spend: 300, revenue: 2520, orders: 18 },
  { region: "EU", channel: "search", spend: 900, revenue: 5280, orders: 44 },
  { region: "EU", channel: "social", spend: 740, revenue: 3100, orders: 31 },
  { region: "EU", channel: "email", spend: 260, revenue: 1760, orders: 16 },
  { region: "NA", channel: "search", spend: 1350, revenue: 9940, orders: 71 },
  { region: "NA", channel: "social", spend: 880, revenue: 5040, orders: 42 },
  { region: "NA", channel: "email", spend: 320, revenue: 3150, orders: 21 },
  { region: "EU", channel: "search", spend: 980, revenue: 6370, orders: 49 },
  { region: "EU", channel: "social", spend: 790, revenue: 3465, orders: 33 },
  { region: "EU", channel: "email", spend: 280, revenue: 2280, orders: 19 },
]).assign({ roas: (df) => df.col("revenue").div(df.col("spend")) });

const roasTable = pivotTableFull(joined, {
  index: "region",
  columns: "channel",
  values: "roas",
  aggfunc: "mean",
  fill_value: 0,
  margins: true,
});
console.log("Mean ROAS by region/channel:");
console.log(roasTable.toString());

const efficient = nlargestDataFrame(joined, 4, { columns: "roas" }).select(["region", "channel", "spend", "revenue", "roas"]);
console.log("\nBest daily placements:");
console.log(efficient.toString());

import pandas as pd

joined = pd.DataFrame([
    {"region": "NA", "channel": "search", "spend": 1200, "revenue": 8680, "orders": 62},
    {"region": "NA", "channel": "social", "spend": 800, "revenue": 3920, "orders": 35},
    {"region": "NA", "channel": "email", "spend": 300, "revenue": 2520, "orders": 18},
    {"region": "EU", "channel": "search", "spend": 900, "revenue": 5280, "orders": 44},
    {"region": "EU", "channel": "social", "spend": 740, "revenue": 3100, "orders": 31},
    {"region": "EU", "channel": "email", "spend": 260, "revenue": 1760, "orders": 16},
    {"region": "NA", "channel": "search", "spend": 1350, "revenue": 9940, "orders": 71},
    {"region": "NA", "channel": "social", "spend": 880, "revenue": 5040, "orders": 42},
    {"region": "NA", "channel": "email", "spend": 320, "revenue": 3150, "orders": 21},
    {"region": "EU", "channel": "search", "spend": 980, "revenue": 6370, "orders": 49},
    {"region": "EU", "channel": "social", "spend": 790, "revenue": 3465, "orders": 33},
    {"region": "EU", "channel": "email", "spend": 280, "revenue": 2280, "orders": 19},
])
joined["roas"] = joined["revenue"] / joined["spend"]
print(pd.pivot_table(joined, index="region", columns="channel", values="roas", aggfunc="mean", fill_value=0, margins=True))
print(joined.nlargest(4, "roas")[["region", "channel", "spend", "revenue", "roas"]])

Click ▶ Run to execute

Ctrl+Enter to run · Tab to indent

📣 Marketing Attribution ROAS

1 · Normalize spend and join conversions

2 · Compare region/channel performance