Smart Bidding: How Much Data tCPA and tROAS Need

In short

Automated bidding learns only from primary conversions. If they’re dirty, no strategy will help.

For Target ROAS (tROAS) in Search and Shopping campaigns, Google requires at least 15 conversions with a value in the past 30 days. That’s the entry ticket, not a guarantee of stability.

Target CPA (tCPA) can be launched with no history, but you should only set targets once data has accumulated.

Value-based bidding isn’t just for e-commerce: for lead generation, we assign values to funnel stages and import sales from the CRM.

The most common mistake is changing targets often and sharply. Judge a strategy after one or two conversion cycles.

Automated bid strategies have long been the standard in Google Ads. Manual bidding is left for narrow use cases. But “automated” doesn’t mean “it does everything on its own.” The algorithm optimizes for what you give it, with whatever data quality you provide.

How automated bidding works

Smart Bidding sets a bid in every individual auction based on signals: device, location, time of day, language, query, browser, remarketing lists and many others. It predicts the likelihood of a conversion, or its value, for that specific impression and sets the bid to match your target.

Strategy	What it optimizes	When it fits
Maximize conversions	The number of conversions within budget	At launch, when there’s little data, or when all conversions have similar value
Maximize conversions with a target CPA (tCPA)	The number of conversions at a set average cost	There’s a steady flow of conversions and a clear acceptable cost per lead
Maximize conversion value	Total value within budget	You have dynamic values or funnel-stage values and are building up data
Maximize conversion value with a target ROAS (tROAS)	Value at a set return on ad spend	Enough conversions with values and clear target metrics
Target impression share	Impression share in a chosen part of the results page, not conversions	Standalone brand campaigns, not a core strategy for sales

In the current interface, tCPA and tROAS aren’t separate strategies but optional fields within Maximize conversions and Maximize conversion value.

How much data you need: the official thresholds

For tROAS, Google sets minimum requirements that depend on campaign type. Here are the main ones from the help article on Target ROAS:

Search and Shopping campaigns — at least 15 conversions in the past 30 days at the conversion-tracking level; only conversions with a value greater than zero count;
Demand Gen campaigns — at least 50 conversions in the past 35 days at the campaign level, with 10 of them in the past 7 days, or 100 conversions in 35 days across all Demand Gen campaigns in the account;
Video action campaigns — at least 30 conversions in the past 30 days.

tCPA has no hard minimum: Google lets you launch it even without conversion history. At the same time, the help center recommends evaluating results over the past 30 days and at least 30 conversions.

What this means in practice

The official minimum is the condition under which the strategy will run at all. Stable performance takes more. The fewer conversions you have, the more randomness goes into every decision the algorithm makes: a single large purchase or a week without sales changes the picture dramatically.

What we do when data is thin:

We consolidate. Fewer campaigns and ad groups with more data in each, instead of dozens of small campaigns.
We use portfolio bid strategies. One strategy across several campaigns with similar economics pools their data.
We don’t set targets at launch. Maximize conversions or conversion value without a target while data accumulates, then a target based on actual performance.
We check whether the budget is limiting. A target-based strategy in a campaign that hits its budget cap every day performs worse than expected.

Learning and the conversion cycle

After significant changes, the strategy enters learning status. It’s triggered by:

a change in bid strategy;
a significant change to the target;
a change to the set of primary conversions or their value;
major structural changes to the campaign.

How long learning lasts depends on conversion volume and on the conversion cycle, the time between a click and a conversion. For a store with impulse purchases, that’s hours; for B2B, it’s weeks. After a target change, Google advises giving the strategy one or two conversion cycles before drawing conclusions.

Another nuance is conversion delay. Google Ads attributes a conversion to the date of the click, so data for the most recent days is always incomplete. If you judge ROAS based on yesterday, it will almost always look worse than it really is. That’s why, when we analyze performance, we exclude the most recent days that fall within the typical delay.

How to choose your first target

The first tCPA or tROAS target is the most important one, because it sets the starting point for learning. Our sequence:

We take the campaign’s actual performance over the past few weeks, excluding the last few days where conversions haven’t caught up with clicks yet.
For tROAS, we make sure conversion values have been passed correctly for at least four weeks or one to two conversion cycles, whichever is longer. That’s exactly what Google recommends.
We set the target close to the actual figure, not the desired one. If actual ROAS is well below what the business needs, a sharply raised target won’t “force” the algorithm to do better. It will simply cut volume.
We check the target against Google’s recommendation in the interface. The recommendation is based on actual data and is a useful sanity check, but it doesn’t replace a business calculation.
From there, we move toward the business target gradually, watching how conversion volume and spend change.

An important trap: a target that looks “safe” can do damage. A tCPA that’s too low or a tROAS that’s too high limits auction participation, the campaign underspends its budget, and it looks as if demand has dried up. In reality, what has dried up are the auctions in which the algorithm can hit your target.

How to change targets

In small steps, not sharp jumps.
No more than once per conversion cycle.
Based on actual data over a sufficient period, not on the result you want.
We log every change in notes so we can later connect it to shifts in performance.

Value-based bidding

Strategies that optimize for the number of conversions treat every conversion as equal. For a business, that’s rarely true: a large order and a small one, a lead for a complex service and one for a cheap consultation all carry different value. Value-based bidding lets the algorithm account for that.

For an online store

Dynamic value from every order is the required baseline.
Margin instead of revenue, if margins vary widely across products. You can pass it through server-side data processing or post-order adjustments.
Conversion adjustments for returns and cancellations, so the algorithm doesn’t learn from orders that never happened.
A new customer acquisition goal, if acquiring new buyers matters more to you than bringing back existing ones.

For lead generation

This is where value-based bidding is especially underrated. Our approach:

We define the funnel stages: lead, qualified lead, deal.
We assign each stage a value proportional to its real worth to the business, based on stage-to-stage conversion rates and average deal size.
We import stages from the CRM into Google Ads through offline conversion import with GCLID and enhanced conversions for leads. Starting in June 2026, API uploads are moving to the Data Manager API, so it’s worth checking your integrations.
We use conversion value rules if value depends on region, device or audience.
We transition gradually from Maximize conversions to Maximize conversion value, and then to tROAS once there’s enough data.

It’s critical that the values are realistic. If every stage is assigned the same value, or the numbers are pulled out of thin air, the algorithm will optimize for those made-up numbers.

Smart Bidding Exploration

For Search campaigns using tROAS, Google offers Smart Bidding Exploration. It lets the strategy temporarily operate at a somewhat lower ROAS within a set range in order to find new queries the algorithm otherwise wouldn’t bid on.

What you should know:

the tolerance is set in the campaign or portfolio bid strategy settings;
it works best together with broad match or AI Max;
Google advises against raising the target ROAS to “compensate” for exploration, because that narrows the room to find new queries;
it’s ineffective in budget-limited campaigns.

Details are in the help article on Smart Bidding Exploration. We turn it on when a campaign is consistently hitting its target and the business is ready to temporarily accept a lower ROAS in exchange for more volume.

Control tools

Seasonality adjustments. For short events, such as a sale, when the conversion rate is expected to change. Not for regular seasonality, which the algorithm accounts for on its own.
Data exclusions. When conversions weren’t recorded, or were recorded incorrectly, because of a technical issue, those days need to be excluded so the algorithm doesn’t learn from them.
Maximum bid limits in portfolio bid strategies, as a safeguard against overly expensive clicks.
The bid strategy report, showing status, targets and top signals.
Experiments, to test a new strategy on split traffic instead of riskily switching the whole campaign.

Common mistakes

Primary conversions include micro-actions: phone number clicks, contact page views, scrolling.
Targets change every few days because of “one bad Tuesday.”
tROAS is set from the desired figure rather than actual data.
A tCPA campaign is constantly limited by budget.
Conversion values are identical or zero, while the strategy is Maximize conversion value.
Dozens of small campaigns, each with a handful of conversions per month.
Performance conclusions drawn from the last few days, ignoring conversion delay.
A technical measurement failure with no data exclusion, so the algorithm learned from zeros.

What to do next

Check three things in your account: which conversions are set as primary, how many conversions with a value each campaign gets in 30 days, and how often bid strategy targets changed over the last quarter. The answers will show whether the account is ready for tROAS or whether the data needs to be cleaned up first.

If you want to move to value-based bidding systematically, the AdsConcept team sets up measurement, CRM data imports, conversion values and bid strategies, and then manages the account with transparent reporting on every change.

Smart Bidding: How Much Data tCPA and tROAS Need, and How to Move to Value-Based Bidding