Attribution is lying to you: why last-click overvalues branded search

If you pulled your attribution report this morning, branded search probably looks like your best channel. Cheap clicks, stellar ROAS, near-perfect conversion rates. The trouble is that branded search mostly harvests demand that something else created.

The harvest problem

Someone sees your TV spot, scrolls past your Meta ads for a fortnight, hears a podcast host read your promo code, and then types your brand name into Google. Last-click gives 100% of the credit to that final search. Every channel that actually built the intent gets nothing.

This is not a small distortion. When we run incrementality-calibrated models for DTC brands, branded search is routinely 60-80% less incremental than the attribution report claims. The demand was already there; the click was just the door handle.

Why this matters for budget

The damage happens when the report becomes the budget:

• Branded search gets more spend, because it "performs".
• Upper-funnel channels get cut, because they "don't".
• Total demand shrinks, but branded search still looks great, because it captures a larger share of a smaller pool.

It is a slow leak that reads as efficiency.

What to do instead

You do not need to abandon click data - you need to stop letting it referee. A Bayesian MMM works from aggregate spend and sales data, so it sees the relationship between TV going live and branded search volume rising a week later. It prices channels by what they cause, not what they touch.

Attribution tells you where the customer was standing when they converted. Modelling tells you what moved them there.

Start by asking one question of every channel in your report: if we switched this off tomorrow, how much revenue would actually disappear? If you cannot answer with confidence, the report is telling you less than you think.