The target you give Smart Bidding is the most important instruction in the account: «get me customers at €30» or «give me back €5 for every euro». Getting it wrong has two symmetrical failure modes: too demanding and the campaign stops entering auctions (little spend, few conversions, «it doesn't work»); too lax and it buys expensive customers for no reason. This lesson is the method for getting it right.
Step 1 · The business target: how much you can pay
Before you look at Google Ads, look at your profit and loss.
Lead generation (CPA)
You need three numbers: what a customer is worth (average revenue × margin), how many leads it takes to get a customer (close rate) and what share of the margin you accept spending to win them.
Average customer: €1,200 in revenue, 40% margin → €480 of margin. You close 1 in every 5 leads → each lead is worth up to €96. You want to keep at least half the margin → maximum CPA ≈ €48 per lead.
Ecommerce (ROAS)
With a 35% gross margin and fixed costs per order, break-even is
ROAS = 1 / margin → 1 / 0.35 ≈ 2.9. Below that you lose money on
every sale; to make money, the target ROAS has to be above it (from 4
onwards, in this example, it starts to pay). Important: Google's «target»
ROAS is calculated on the conversion value you send: if you send revenue
including VAT, the threshold changes.
Step 2 · The achievable target: what your history says
Now open the campaign with the last 30 days (or more, if there are few conversions) and look at its actual CPA (or actual ROAS). That number is what the campaign is capable of doing today. Three scenarios:
| History vs business | What to set |
|---|---|
| Actual CPA (€45) ≤ maximum CPA (€48) | Target = actual CPA, or slightly below (€40). It is already profitable; fine-tune |
| Actual CPA (€70) > maximum CPA (€48) | Target = actual CPA minus 10-20% (€60). Bring it down in steps to the maximum (lesson 5). If it never gets there, the problem is not the bid: it is tracking, ad, landing page or product |
| No history (new campaign) | Maximize conversions with no target for 3-4 weeks; then set from the actual figure |
The classic mistake: setting the desired target straight away (€48) when the campaign is doing €70. Google only enters the auctions it estimates at €48 or less: almost none. The campaign quietly switches itself off, and the diagnosis «Smart Bidding doesn't work» is false: you asked it for the impossible overnight.
Step 3 · Adjust by campaign and conversion type
- Brand allows a much lower CPA (it converts on its own); generic, higher. Different targets, different campaigns.
- If the conversion is a lead rather than a sale, the target CPA has to build in the close rate (step 1). A CPA of €20 per lead with a 2% close rate is a customer at €1,000.
- If you mix conversions of different value in the same action (a quote and a call, say), the average CPA misleads you. Better to use different values per type and bid for value (ROAS) — we'll see this in Measurement II.
Step 4 · Review, don't chase
The target is reviewed every 3-4 weeks with the whole chain (actual CPA, volume, lost IS (rank)). Signals:
- Actual CPA sustainably below the target and the campaign limited by budget → there is room: raise the budget, don't touch the target.
- Actual CPA below and spend nowhere near the budget → the target is too lax or the market has improved: lower it 10-15%.
- Actual CPA sustainably above → don't cut it in one go: first review search terms, negatives and the landing page; then adjust by 10-15%.
- Spend falling with «learning» showing → you have touched too much: stop and wait.
💡 Ninja trick: the target the strategy chases and the business target are two different things, and Google only knows the first one. Our Smart Bidding (SBNS) stores each campaign's business target (its actual CPA or ROAS, with its reference conversions) and judges compliance against that —never against the strategy's target— bringing the target closer to the real one in gradual steps and only when the data allows. It is exactly the method in this lesson, running every night.
What you should remember
- Work out how much you can pay from margin and close rate (CPA) or
from
1/margin(ROAS). - Set the target from your actual history, not from what you wish for: actual −10-20%, and come down in steps.
- Too demanding a target = a campaign that switches itself off; too lax = expensive customers.
- Review every 3-4 weeks with the whole chain; the target is not chased daily.