Level 7 · Group coupon codes into campaigns

map: look every value up in a dictionary

Series.map(d) replaces each value with d[value]. It’s the pandas version of looking a code up in a reference table:

campaigns = {"WELCOME10": "Onboarding", "SPRING15": "Seasonal", "SUMMER15": "Seasonal",
             "BFCM25": "Holiday", "HOLIDAY20": "Holiday"}
orders["coupon_code"].map(campaigns).head()

The important detail is what happens to a value that isn’t a key. A plain Python dictionary raises KeyError for a missing key; map doesn’t. It quietly returns NaN. Orders with no coupon (NaN in coupon_code) also come back as NaN, since NaN isn’t a key either.

Make the gaps visible

Harbor & Pine has six coupon codes, but Jordan’s list has five. FREESHIP is missing, so its orders map to NaN:

coupon_orders = orders[orders["coupon_code"].notna()].copy()
mapped = coupon_orders["coupon_code"].map(campaigns)
mapped.isna().sum()      # 193, all FREESHIP

If you group now, those 193 orders disappear, because groupby leaves NaN keys out by default. The totals would add up to 1,530 instead of the 1,723 orders that used a coupon, and nobody would be told why. fillna("Unmapped") turns the silent gap into a row Jordan can see and act on.

This is why we filter to coupon orders first. Otherwise the 10,277 orders without any coupon would also become NaN and then “Unmapped”, mixing “no coupon” with “a code I forgot”.

The result

campaign	n_orders
Onboarding	587
Holiday	541
Seasonal	402
Unmapped	193

WELCOME10 alone is the single most used code. Holiday combines BFCM25 (348 orders) and HOLIDAY20 (193). When Jordan adds FREESHIP to the dictionary, the Unmapped row will disappear, which is exactly how a lookup table should behave. A quick reconciliation proves nothing was lost: 587 + 541 + 402 + 193 = 1,723, the number of orders with any coupon.

Common mistakes

• Grouping straight after map, so unmapped codes vanish from the totals.
• Using .replace(campaigns) instead: it leaves unknown codes unchanged, so FREESHIP would look like a campaign name.
• Mapping the whole orders table, so “no coupon” and “unknown coupon” end up in the same bucket.
• Expecting map to raise an error for a missing key, the way a plain dictionary lookup does.