The single most expensive mistake I see on SaaS Google Ads accounts: running Target CPA on a campaign that doesn't qualify, while the account-level conversion data clearly points to Maximize Conversion Value being a better fit. The team picked tCPA because it's the default suggestion. Six months later, CAC is bouncing around like a pinball and nobody knows why.
This post is the decision framework I use to pick between the two, plus the migration path when you need to move from one to the other.
What they actually do
Target CPA (tCPA) tries to deliver conversions at or below a fixed cost. You tell it "I want demo requests at $200 each", and it bids on every auction to maximise the number of conversions it can deliver without exceeding $200 average CPA across the campaign.
Maximize Conversion Value (MCV) tries to deliver as much total conversion value as possible within your daily budget. You don't set a CPA target; you set a ROAS target (or no target at all, just a budget). The algorithm picks which auctions to win based on predicted value × predicted CVR.
The shorthand: tCPA optimises for cost, MCV optimises for revenue. Sounds the same. Isn't.
When tCPA works
Three conditions all need to be true. If even one is missing, tCPA underperforms and you should default to MCV.
- Conversion volume ≥30/month per campaign. Google says 30, but that's the floor. Healthy tCPA behaviour starts around 50/month. Below that, the algorithm sees random noise as signal and overcorrects.
- Conversion values are similar. A demo request from a 10-employee company and a 500-employee company have the same conversion value as far as tCPA cares. If one closes at $5k ACV and the other at $80k ACV, tCPA can't tell them apart — it just tries to keep CPA flat across both.
- You actually know your target CPA. Most teams guess. The number that should go into tCPA is
CAC ceiling = ACV × close rate × target LTV-to-CAC ratio. If you can't write that out, tCPA isn't your problem yet — unit economics is.
Concretely: a B2B SaaS with 40+ demo requests per month per campaign, similar ACVs across segments, and a clear $250 demo-CPA target derived from a $6k ACV and 22% close rate. That campaign will hum on tCPA.
When MCV works
Most of the rest of the time. Especially:
- Mixed conversion values. SMB and Enterprise on the same campaign, or trial signups + demo requests being counted together. MCV with proper conversion values is dramatically better than tCPA here.
- Lower volume campaigns (10-50 conversions/mo). MCV with a ROAS target is more forgiving of noise than tCPA. The algorithm has more degrees of freedom.
- Multi-product or multi-tier offerings. If you sell a $99/mo starter and a $999/mo enterprise tier, you want the algorithm to bid harder for enterprise traffic. tCPA can't do this.
- Anywhere you'd care about which conversion you win, not just how many. That's the whole point of MCV.
Setting conversion values
MCV is useless without conversion values. Pass actual dollar value on paid conversions. For lead-stage events, use a derived value based on close rates:
// Paid conversion — use the actual annual value
gtag('event', 'conversion', {
send_to: 'AW-XXXXXXXXX/PAID_ID',
value: subscription.annualValue,
currency: 'USD',
transaction_id: subscription.id,
});
// Demo request — use derived value from historical close rate
// If 22% of demos close at $6k ACV → demo request value = $1,320
gtag('event', 'conversion', {
send_to: 'AW-XXXXXXXXX/DEMO_ID',
value: 1320,
currency: 'USD',
transaction_id: leadId,
});Revisit derived values quarterly. If your close rate drifts up to 25% or your ACV drifts down to $5k, your derived value needs to change or the algorithm steers toward the wrong buyer profile.
tROAS: the MCV tuning knob
MCV has two modes: pure (no target) and with a Target ROAS. Start without a target — let it spend the budget freely for 4-6 weeks and gather data. Then read off the achieved ROAS from the report and set a tROAS target slightly above it (e.g. achieved 380%, set 400%).
Common mistake: setting an aspirational tROAS like 600% when the campaign has historically delivered 200%. The algorithm just throttles spend until it can hit the unrealistic target, and you wonder why your campaign suddenly stopped spending.
The tCPA → MCV migration path
Don't flip the switch and hope. Run this sequence:
- Week 0: Audit conversion values. If you don't have them on every event, set them — see our conversion-tracking guide for the wiring details.
- Week 1-2: Run for two more weeks on tCPA with values flowing. This builds a value-aware history the algorithm will use post-switch.
- Week 3: Switch bid strategy to MCV with no target. Expect CPA to bounce 10-30% for a learning period.
- Week 5-6: Read achieved ROAS. If healthy, set a tROAS at +5% over achieved. If not, investigate value mapping before tweaking.
Set a calendar reminder for week 4 — most teams panic during the learning period and roll back to tCPA, abandoning the migration before MCV has had a chance to learn.
When this decision really pays
On a small account spending $5k/month, the difference between tCPA and MCV might be 5-10% of revenue. Worth doing, not life-changing.
On a $50k+/month account with mixed conversion values, switching from tCPA to MCV typically moves revenue per dollar spent by 20-40% within two months. That's the kind of move that shows up in board meetings.
For specific situations — multi-product accounts, accounts where SMB and Enterprise pipeline mix matters, accounts where your CRM knows the close value at signup time — a focused consult can map the right strategy to your campaign mix in under an hour. Bring your last 90-day conversion report and we'll work backward from there.