Value-based bidding is a Google Ads Smart Bidding approach that predicts the monetary value of each conversion and bids more for clicks likely to produce high-value outcomes. You feed the algorithm a per-conversion value signal and it optimizes for total or target conversion value. The hard part is producing trustworthy values.
Key Takeaways
- Value-based bidding optimizes for predicted conversion value, not conversion count, so every conversion you feed to Google must carry a real number.
- The two value-based smart bidding strategies are Maximize Conversion Value (spend to the budget for max revenue) and Target ROAS (hit a return floor while scaling).
- B2B and long-sales-cycle startups must model lead values - assign stage-weighted or statistically modeled values rather than waiting for closed revenue months later.
- Conversion value rules and offline conversion imports are how you correct and enrich values after the click, closing the gap between the ad platform and your CRM.
- Migrating from Target CPA to a value goal is a ramp, not a flip - do it with blended values and a step-down of the old constraint to avoid tanking volume.
What Is Value-Based Bidding and How Does the Auction Use Predicted Conversion Value?
Value-based bidding is any Smart Bidding strategy that optimizes toward the value of conversions rather than their count. Cost-based bidding asks how much to bid for a conversion at a target cost; value-based bidding asks how much to bid given what the click is worth. That shift lets the auction pay more for a click predicting a $5,000 enterprise opportunity than one predicting a $50 self-serve signup.
Under the hood, the auction runs an auction-time prediction for every eligible query. The model estimates two things: the probability that a click converts, and the monetary value of that conversion. Multiply those and you get predicted conversion value per click. The bidder then translates that prediction into a bid - for Target ROAS it divides by your target, for Maximize Conversion Value it bids up to whatever your budget allows to capture the highest-value clicks. The mechanism is only as good as the value number you put on each conversion, which is why the rest of this post is about producing that number honestly.
The engine learns from your historical conversion values. If high-value conversions cluster on certain queries, devices, or audiences, the model bids more aggressively there. If your values are flat, stale, or missing, the model falls back to optimizing for conversion count and your value strategy becomes a pricier Target CPA. Our Smart Bidding overview covers the broader bidding system these strategies sit on top of.
What Are the Two Value-Based Smart Bidding Strategies and When Does Each Fit?
There are exactly two value-based smart bidding strategies, and the choice between them is about whether you want an efficiency guardrail.
Maximize Conversion Value spends your budget chasing the highest total conversion value it can find, with no ROAS floor. It is the right call when growth is the priority, your value tracking is trustworthy, and you are not bound by a strict return constraint. It is also the safer on-ramp to value-based bidding because it does not choke spend the way an over-ambitious ROAS target can.
Target ROAS imposes a return-on-ad-spend floor: it will not spend the full budget if it cannot find auctions that clear your target. Use it once you have a known margin floor and enough conversion volume for the model to learn. It is a margin-protection tool as much as a scaling tool. For the deep mechanics and migration sequencing, our Target ROAS guide covers how to set the target without collapsing volume.
How Do You Assign Values When You Do Not Sell Online?
If you sell software to enterprises and revenue lands 6 to 18 months after the click, you cannot wait for actual revenue to flow into Google. You have to assign a modeled lead value at conversion time. The goal is not to be perfectly right - it is to be directionally right so the auction bids more on leads that tend to become valuable customers.
Start with a worked example using clearly hypothetical numbers. Say a startup's trailing data shows: 1,000 leads became 200 MQLs, 60 SQLs, 30 opportunities, and 9 closed-won deals, with $450,000 in total closed revenue. That is a 0.9% lead-to-customer rate and $50,000 average deal size, so each raw lead is worth about $450 in expected value (0.009 x $50,000). But not all leads are equal. If only 30% of leads ever reach MQL stage, a pre-MQL lead should be valued at roughly 30% of that expected value, or about $135, while a lead already at SQL stage carries the value of a SQL-converted path - closer to $450 divided by the SQL rate of 6%, or about $1,500 in expected value.
The point of the arithmetic is to stop treating every lead as a flat $10. A lead that reached a sales-accepted stage has demonstrably higher expected value than a raw form fill, and the auction should bid accordingly. You pass these modeled values at conversion time, then correct them later with offline imports once real outcomes are known.
What Is Stage-Weighted Lead Valuation and How Do You Build It?
Stage-weighted valuation assigns a distinct expected value to each funnel stage so the auction can separate a curious tire-kicker from a buying committee member. You compute the expected value at each stage by multiplying the historical win rate from that stage by the average deal size, then you pass that stage's value to Google the moment the lead reaches it.
A practical table of value-assignment methods helps you pick your lane:
| Method | Effort | Accuracy | When to use |
|---|---|---|---|
| Static value | Low - set one number per conversion | Low - assumes every conversion is equal | Early testing, uniform lead gen with no value variance |
| Stage-weighted value | Medium - map win rates per funnel stage | Medium-high - reflects lead quality at conversion | B2B with a defined pipeline and CRM stages |
| Modeled value | High - build a regression or heuristic model | High - predicts value from firmographics and behavior | Long sales cycles where stage data is sparse |
| Actual revenue | Medium - requires offline import plumbing | Highest - real dollars per conversion | Ecommerce or closed-loop CRM with revenue feed |
Most startups should graduate from static to stage-weighted quickly, then layer modeled values as pipeline data accumulates, and finally close the loop with actual revenue via offline imports.
What Are Conversion Value Rules and What Are They For?
Conversion value rules let you adjust the value Google uses for a conversion based on signals it sees at auction time - device, location, audience, or a custom combination. They are a fast way to tell the algorithm, "a conversion from a enterprise-tier audience is worth 2x my baseline," without changing the value you pass in the tag.
Use them for corrections the conversion tag cannot make. If your CRM shows that mobile leads close at half the rate of desktop leads, apply a 0.5x value adjustment to mobile conversions. If a vertical audience historically produces 3x the revenue, a positive adjustment steers budget there. The caveat: value rules multiply or add to your base value, so they amplify both good and bad base data. They tune a trustworthy base value; they do not substitute for one. Used well, they encode known economic reality into the auction in minutes instead of waiting weeks for the model to infer it.
How Do You Feed Real Values Back with Offline Conversion Imports?
The conversion tag fires at lead time with your best modeled or stage-weighted guess. The real outcome - whether that lead became revenue, and for how much - lives in your CRM, days or months later. Offline conversion imports are the bridge that sends that corrected value back to Google, keyed to the original click or conversion ID.
The workflow is straightforward. At conversion time you store the Google click ID (GCLID) or conversion ID alongside the lead record. When the lead reaches a meaningful stage or closes, you upload an adjusted conversion with the same identifier and the realized value. Google re-attaches the value to the original auction, and the bidding model learns from outcomes instead of guesses. This is how a B2B startup closes the loop without waiting for the model to slowly infer that enterprise leads are worth more. Our offline conversion tracking guide covers the upload mechanics and identifier handling in detail.
Two practical notes. First, imports are most powerful when they correct values upward and downward - if you only ever raise values, the model still cannot tell a $50 lead from a $5,000 one. Second, pair imports with enhanced conversions so that even the leads you never see in the CRM still carry a value signal. Our enhanced conversions guide explains how hashed first-party data improves value attribution when cookies are missing.
How Do You Migrate from Target CPA to a Value Goal Safely?
Flipping from Target CPA to a value goal overnight is the fastest way to crater volume, because the algorithm suddenly sees a different objective and relearns from scratch. A staged ramp preserves the signal while the model adapts.
- Audit your current CPA performance and compute an implied value per conversion. If your Target CPA is $100 and historical lead-to-customer rate is 10% with $3,000 average deal size, your implied value is roughly $300 per lead - that becomes your baseline value.
- Start passing values on every conversion while still on Target CPA. Let Google collect value data for 2 to 4 weeks so the model has a value history before you change strategy.
- Switch to Maximize Conversion Value first, not Target ROAS. This keeps spend flowing while the model learns to optimize for value rather than cost, avoiding the spend-collapse risk of an immediate ROAS floor.
- Run Maximize Conversion Value for 3 to 4 weeks and observe whether high-value segments get more budget. Confirm total conversion value rises without volume dropping.
- Only then introduce a Target ROAS target, set at or slightly below your trailing value-per-spend ratio, and raise it in 10 to 15% increments with 2 to 3 week learning windows between each.
- Keep a conversion-value diagnostic running. If values become stale or imports lag, pause the value strategy and fall back rather than letting the model optimize on bad data.
This ramp keeps the auction stable because each step changes only one variable at a time. The most common failure is skipping step 2 - switching strategy before any value history exists - which forces the model to learn value and strategy simultaneously.
What Breaks Value-Based Bidding?
Value-based bidding fails in predictable ways, and almost all of them are data problems rather than algorithm problems.
All leads valued the same is the classic trap. If every conversion carries a flat $10, the auction cannot distinguish a tire-kicker from a buying committee, so it optimizes for whoever fills forms cheapest - usually low-quality volume. Stale values are the second failure: a value model built on last year's pricing or pipeline shape quietly misallocates budget as the business changes. Low conversion volume is the third - the model needs enough value-carrying conversions (roughly 30 to 50 in 30 days) to learn which auctions predict value, and below that threshold it guesses. Consent and tracking gaps are the fourth: when a meaningful share of conversions lack GCLIDs or fall outside the attribution window, the values you import cannot reattach to the right auction, and the model learns from a biased sample.
The unifying theme: value-based bidding is a data layer dressed up as a bidding feature. Get the values trustworthy - modeled for B2B, stage-weighted where you can, corrected via offline imports - and the strategy does the rest. Get the values wrong and you have automated the spread of bad assumptions at scale.
Layer bid adjustments by device and location on top of value-based bidding to shift spend toward your best segments.
Related reading: if your conversions fire before revenue is known, Google Ads conversion value rules show how to assign a modeled value so value-based bidding has a number to optimize.
Frequently Asked Questions
How Does Value-Based Bidding Use Predicted Conversion Value in the Auction?
The auction predicts the probability a click converts and multiplies it by the monetary value of that conversion to get predicted conversion value per click. It then bids more on clicks with higher predicted value, either spending to capture maximum total value or capping bids to meet a Target ROAS floor. The prediction is only as accurate as the per-conversion values you feed it.
What Is the Difference Between Maximize Conversion Value and Target ROAS?
Maximize Conversion Value spends your budget to capture the highest total conversion value with no efficiency floor, making it the safer on-ramp and the right choice when growth matters more than a strict return. Target ROAS adds a return-on-ad-spend floor so the algorithm will not spend if it cannot clear your target, making it a margin-protection tool once you have enough volume and a known margin floor.
How Do You Assign a Value to a B2B Lead When Revenue Arrives Months Later?
You assign a modeled expected value at conversion time using historical win rates and average deal size, often split by funnel stage so an SQL carries more expected value than a raw form fill. You then correct that value later with an offline conversion import once the lead closes, feeding the realized revenue back to the original click so the model learns from outcomes rather than guesses.
Can Value-Based Bidding Work Without Accurate Conversion Values?
No. Value-based bidding optimizes for the value number attached to each conversion, so if that number is flat, stale, missing, or wrong, the algorithm optimizes for a distorted signal - usually reverting to conversion-count optimization while charging you a premium. Trustworthy values, whether modeled, stage-weighted, or imported from the CRM, are the precondition for the entire strategy.