Automated Bidding Strategies Explained: tCPA, tROAS, and When to Bid Manually

Automated bidding sets a bid for each auction based on the goal you pick and the conversion data you send back. It follows that goal even when the goal is wrong, and it learns from your tracking even when the tracking is broken. This guide covers what each Google and Meta strategy optimizes and how much data it needs. The last sections cover common mistakes and the cases where a manual bid still works better.

What Smart Bidding optimizes

Automated bidding (Google calls its version Smart Bidding) sets a separate bid for every auction. The model behind it is trained on your conversion history and on signals from each query or impression: device, location, time of day, audience, browser and dozens more. Google calls this auction-time bidding. A manual keyword bid is one fixed number for every auction. The model recalculates the bid each time, based on how likely that user is to convert.

The model can only optimize the number you give it. If you ask for the cheapest conversions, it will find them, and they are often not your most valuable customers. If your conversion tracking fires on the wrong event, the algorithm will buy more of that wrong event. So the strategy you pick sets the goal. The model then works toward that goal with more patience and more data points than a person could.

Smart Bidding optimizes for the goal you set, even when that goal is the wrong one.

The five Google strategies

Google’s bid strategies fall into two groups. Volume-based strategies count conversions. Value-based strategies count revenue. Target CPA and Target ROAS are now optional targets inside the other two. That is why the current interface shows them as settings within Maximize Conversions and Maximize Conversion Value instead of as separate strategies.

Maximize Conversions

Spends your full budget to get as many conversions as possible and counts every conversion as worth the same. Use it when you want to spend the whole budget and every lead is worth about the same to you. The risk: it has no cost target, so it will spend up to the budget to get more volume. With a large budget, CPA can rise.

Target CPA

Now an optional target within Maximize Conversions. It tries to get conversions at the average cost you set. It bids higher in auctions that are likely to convert cheaply and lower in expensive ones. It still counts every conversion as worth the same. It only adds a cost limit to the volume goal. Use it for lead gen and for any funnel where all conversions have the same value.

Maximize Conversion Value

Spends the budget to get the highest total conversion value, meaning revenue. It will accept fewer conversions if they are worth more. Use it only when your conversions have different values and you send those values back to Google.

Target ROAS

An optional target within Maximize Conversion Value. It bids to reach a set ratio of conversion value to spend. A 400% ROAS target tells Google you want $4 back for every $1 spent. It needs accurate conversion values that differ from one conversion to the next, so its results depend on your conversion tracking and first-party data. If every conversion sends the same value, it works like a weaker Target CPA.

Enhanced CPC (retired)

ECPC was Google’s first Smart Bidding strategy. It raised or lowered your manual bids at auction time based on how likely a conversion was. Google turned it off for Search and Display campaigns in the week of March 31, 2025. Campaigns that were not moved to another strategy fell back to Manual CPC (Search Engine Land). If an older guide still recommends ECPC, that advice is out of date. The value-aware strategies above are its current equivalents.

How many conversions you need

Smart Bidding is a statistical model, so it needs enough examples to learn from. Google requires at least 15 conversions in the past 30 days before Search and Shopping campaigns can use value-aware Smart Bidding. Its long-standing guidance is that Target CPA works best with about 30 conversions in the last 30 days. Other campaign types have their own minimums. Demand Gen needs about 50 conversions in 35 days, Video Action campaigns about 30 per month, and App campaigns around 300 per month.

Below these numbers the algorithm is guessing. Two conversions from $2,000 of spend tell it almost nothing about which auctions to bid on, so bids move a lot and CPA jumps from week to week. If one campaign has too few conversions, you can consolidate: merge small ad groups or campaigns so their conversions add up to a volume the model can learn from. The other option is to stay on manual bidding until you have the data.

The learning phase: why you need to wait

When you launch a strategy or change its target, the algorithm goes into a learning period. It recalibrates, and results are less stable. Google recommends leaving the campaign alone for at least one to two conversion cycles before you judge it. It also recommends against changing the target often, because each change restarts the process. Any significant edit to budget, target or bid strategy can reset learning, and the model loses the progress it made.

Meta gives a specific number. An ad set leaves the learning phase after about 50 optimization events within a 7-day window. Ad sets that do not reach that volume get the “Learning Limited” status. Delivery there is less stable and cost per result is harder to predict (Meta Business Help Center). Significant edits reset the count. So frequent edits to Meta ad sets hurt results: after each change, the ad set starts learning from the beginning.

Meta’s bid strategies and their Google equivalents

Meta’s bid strategies use the same split between volume and value. They add a second choice: how strictly you limit cost. The names are different from Google’s, but the strategies work in a similar way.

• Highest volume (formerly Lowest cost). Spends the budget to get as many results as possible, with no cost target. Roughly the same as Maximize Conversions.
• Highest value. Spends the budget to get the highest total purchase value. It is meant for advertisers who send varied order values. It matches Maximize Conversion Value.
• Cost per result goal (cost cap). Keeps the average cost per result near your target while still trying to get volume. Its purpose is close to Target CPA. Meta treats the number as an average to aim for. It is not a hard limit.
• Bid cap. Sets a hard limit on the bid Meta can place in any auction. It is the most manual of the automated options. You get full cost control and give up some delivery. If the cap is too low, the ad set may barely deliver.
• ROAS goal. Bids to reach a target return on ad spend. It is the direct equivalent of Target ROAS and needs the same clean value data sent back to Meta.

The most common mistake on Meta is the same one people make on Google. If you set a cost cap or ROAS goal from the number you want instead of the number you have achieved, the ad set won’t deliver. If your actual cost per result is $40, a $25 cost cap asks Meta for volume that does not exist. Set caps and goals from your last 14 to 30 days of performance, then tighten them step by step.

Common mistakes

Most cases where “Smart Bidding failed” come from a few common mistakes. The algorithm is not the cause in any of them.

• Wrong conversion signal. If you optimize for an early-funnel event, such as a page view, an add-to-cart or a form open, the model learns to buy cheap actions that never turn into revenue.
• Too little data. Target ROAS on a campaign with eight conversions a month gives the model nothing to learn from. Its bids end up close to random within your limits.
• Impatience. Changing the target every few days restarts the learning period each time. The campaign never settles, and the operator decides automation is broken.
• Unrealistic targets. A Target CPA far below your historical CPA, or a ROAS goal above anything the account has reached, cuts delivery. The model will not overpay for volume it cannot buy at a profit.
• Wasted spend the model can’t see. Smart Bidding still bids on irrelevant queries if you never add negative keywords or exclusions. It sets bids for the traffic you allow in. Deciding which traffic is worth bidding on is still your job.

When manual or portfolio bidding works better

Automation is the default because at scale it prices auctions better than a person can. But in some situations a fixed bid, or portfolio control across several campaigns, still works better than leaving everything to the model.

• Few conversions. Below the learning thresholds, a bid set by a person based on experience is more predictable than a model working from a handful of data points.
• Unreliable tracking. If conversions fire on the wrong event, are deduplicated incorrectly, or go missing because of consent gaps, the model learns from bad data. Fix the tracking before you rely on automation.
• New campaigns. A new campaign with no history gives the algorithm nothing to model. It is often faster to bid manually first, collect a clean baseline, and switch to Smart Bidding once you have the volume. Launching straight into automation can waste budget during the learning phase.
• Hard cost limits. Smart Bidding aims for averages and does not guarantee a maximum. In regulated verticals or on thin-margin products, the cost per click or per acquisition sometimes cannot go over a set amount. In those cases a manual bid or a bid cap gives you control that average-based strategies do not.
• Trade-offs between campaigns that the platform can’t see. Smart Bidding in one campaign optimizes that campaign. It does not look at your whole account or at your combined margin across Google and Meta. Portfolio strategies and budget decisions made by people still handle those trade-offs. For the same reason, cross-channel budget pacing needs a person or a separate tool. A setting inside one platform can’t do it.

Running automated bidding on both platforms

In short: use automated bidding, send it clean conversion data, set targets from your real history, and give it enough time to learn. Keep manual control when you have few conversions, unreliable tracking, or a cost limit you cannot go over. These rules follow from how the strategies work: each one works toward exactly the target you set.

The platforms leave one decision to you: how much budget each campaign and each platform should get, hour by hour, as performance changes. Adriva automates that work across Google Ads and Meta from one workflow. The Smart Bidding models on each platform then bid inside a budget split that is actively managed. Read how the hourly rebalancing works to move spend toward the campaigns that convert.