Bid Strategies and Budget Allocation
Date: 2026-08-16
Automated bidding trades control for efficiency, and the trade is usually worth taking. What you keep is the choice of what to optimise towards — and feeding it the wrong target is the most expensive mistake available in paid media.
What it is
Bid strategy is how much you pay per auction. Budget allocation is how spend is distributed across campaigns and channels.
Both are now largely automated. The platforms bid using signals you don’t have — device, time, history, predicted intent — and they beat manual bidding on efficiency in most cases.
What you actually control
Three things, and they’re the whole job:
| Lever | Effect |
|---|---|
| The conversion definition | What the optimiser is trying to produce |
| The value passed with it | How it trades off between conversions |
| The constraint — target CPA or ROAS | Where it stops bidding |
The conversion definition is the one that matters most. An optimiser maximising “purchases” will find you purchases — including refunded ones, test orders, and low-margin ones — because it optimises exactly what you asked for.
Feed contribution, not revenue
The single highest-return change available in most accounts, and it requires only that you know margin by product.
optimiser fed REVENUE optimiser fed CONTRIBUTION
£100 order, 10% margin = £100 = £10
£50 order, 40% margin = £50 = £20
bids harder for the £100 order bids harder for the £50 order
↑ the one that makes money
Passing revenue tells the machine to chase turnover. Passing contribution tells it to chase profit, and it will — Contribution Margin, Basket Composition.
Also worth excluding: refunded orders, and test orders. An optimiser trained on refunded conversions learns to find customers who return things.
Marginal, not average
Budget decisions are marginal decisions, and average CAC answers a different question:
spend customers avg CAC marginal CAC
£140k 4,000 £35 £40
£180k 4,500 £40 £80 ← the extra £40k
Average CAC at £40 looks affordable. The next £40,000 costs £80 per customer, and that’s the number the decision needs — Marginal Analysis, Customer Acquisition Cost.
The practical implication: allocate to the point where marginal CAC equals what a customer is worth, per channel, rather than distributing by historic ROAS.
Reallocating across channels
The common approach — shift budget towards the channel with the best reported return — is wrong for a specific reason: reported return isn’t comparable across platforms. Each uses its own attribution, window, view-through and modelling.
Google reports 4.2 ROAS
Meta reports 3.8 ROAS
→ shift budget to Google?
these numbers measure different things and
overlap on the same orders
Allocating on them systematically favours whichever platform attributes most generously — Walled Garden Reporting, Attribution Models.
The defensible routes: one common attribution model applied to every channel as a yardstick, and Incrementality Testing for anything material.
Practical rules
- Let the machine bid; you set the target. Manual bidding rarely wins now
- Give it enough conversions to learn. Automated strategies need volume; below a threshold they perform worse than manual — split campaigns too finely and none of them learn
- Change targets gradually. Large changes reset the learning phase and cost days of performance
- Set budgets by marginal return, not by last month’s ROAS
- Reserve a portion for testing — new creative, new audiences — that isn’t judged on immediate return, or the system optimises into a local maximum and stays there
- Watch new versus returning. An optimiser maximising conversions will happily find your existing customers, who are cheapest to convert — New vs Returning Customer Acquisition