Algorithmic (auto) bidding vs manual bidding: when to hand the wheel over

Algorithmic (auto) bidding vs manual bidding: when to hand the wheel over

DSPs offer automated bidding that optimizes toward a goal, and manual bidding where you set the price. The right choice depends on data density, not on which sounds more advanced.

1. Algorithmic bidding (target-CPA or target-ROAS style): the model adjusts each bid using conversion signals. It needs volume — a steady stream of conversions to learn from. On thin data it overfits noise and swings wildly.

2. Manual bidding: you set a fixed bid and adjust with multipliers. Predictable, slow, but it does not need a conversion firehose to behave. Correct for low-volume or long-consideration campaigns where the algorithm would starve.

3. The tradeoff: automation scales decisions you cannot make by hand across millions of auctions, but its quality collapses below a conversion threshold (commonly cited as tens of conversions per week, varies by DSP). Manual never collapses but never scales its intelligence either.

4. The transition trap: launching a fresh campaign straight into algorithmic bidding gives the model no history. It spends erratically during the learning phase, and that spend is the tuition.

Why it matters: teams enable auto-bidding on a campaign that will never feed it enough conversions, then blame the algorithm for volatility it was structurally guaranteed to produce. Check your conversions-per-week against your DSP's stated learning threshold before automating. Below it, manual bidding with multipliers is not the primitive choice — it is the correct one.