# Key Insights — Marketing Campaign Performance (Jan–Jun 2026)

- **Cleaning changed the dataset materially, not cosmetically.** 798 raw rows → 780 after removing 18 duplicate CampaignIDs; 30 spelling/casing variants of Channel collapsed to 6 real channels and 21 Manager variants to 5; 103 blank cells across 5 columns were imputed by channel average or labelled "Unknown". Any analysis run on the raw file would have double-counted 18 campaigns and split "TikTok" across four different spellings.
- **Email's numbers are too good to be true and should be treated as a data-quality defect, not a win.** Email shows $2.81M revenue on just $31.4K spend — an 89.7x ROAS, versus 2.5–4.7x for every other channel. Email accounts for 45% of all revenue but 3% of all spend. That is almost certainly a units error in the source (spend recorded in the wrong scale), and it distorts the total 6.4x blended ROAS. Excluding Email, the portfolio returns 3.7x on $938.8K spend.
- **The biggest budget line is the worst-performing large channel.** Google Ads absorbs $345.5K — 36% of total spend, more than any other channel — but returns 3.64x against its 4.0x target, and 53 of its campaigns land Below Target. LinkedIn is worse on efficiency (2.46x, just under its 2.5x target, with 62 Below Target campaigns) but only carries $201.8K. Rebalancing spend out of Google Ads is the highest-dollar lever available.
- **One in three campaigns misses its channel benchmark.** 257 of 780 campaigns (33%) are classified Below Target, concentrated in LinkedIn (62), TikTok (54) and Google Ads (53). Facebook is the healthiest paid social channel at 4.67x against a 3.5x target with only 38 misses.
- **Performance is seasonal and top-heavy.** Six months of data (Jan–Jun 2026) peak in May at $1.35M revenue, 56% above the January low of $862.7K. Revenue is also concentrated at the campaign level: the single largest campaign, CMP-1089 (managed by R. Kim), booked $82.8K — more than double the largest Google Ads campaign at $37.4K. Conversion rates barely move across audiences (Returning 5.64%, High-Value 5.39%, New 5.20%), so channel and timing, not audience targeting, are where the variance lives.