The math, the process, and the conversion setup behind value-based bidding, plus a calculator that turns your margins and your actuals into a target you can actually hit.
A Target ROAS is a constraint, not a wish.
It tells Google the lowest return you will accept, then the algorithm skips the auctions that would not clear it. Base the target on your actual margins and returns. Everything starts from one line: break-even ROAS = 1 / your gross margin. Price the target above that for profit, near your actuals for volume.
Smart Bidding optimizes toward whatever you tell it to want. Count-based strategies want more conversions. Value-based strategies want more conversion value. The whole game is teaching the machine which conversions are worth more, then deciding how hard to push for return.
An umbrella for strategies that optimize to conversion value, not conversion count. The two members are Maximize Conversion Value and Target ROAS.
Get the most total conversion value within a fixed budget. No return target. This is the entry point and the data-building phase for VBB.
The same engine with a return floor bolted on. You name the return you need, Google bids to hold it. It is a setting on Max Conversion Value, not a separate thing.
Raise the target and Google walks away from cheaper, lower-value auctions. You get a higher return on less spend. Lower it and volume returns. There is no free lunch.
Six formulas carry the entire topic. ROAS is usually shown as a percent (5:1 is 500%) or a ratio (5:1). The calculator below runs all of these for you, but know them cold.
Revenue (or assigned value) returned per dollar spent. $5,000 value on $1,000 spend is 5:1, or 500%.
The floor. At a 40% margin you need 2.5:1 (250%) just to cover product cost and ad cost. Below it, every sale loses money.
Set the net profit margin you want to keep after ad spend. 40% margin, keep 10%: target = 1 / 0.30 = 3.33:1 (333%).
The mirror image of ROAS. If ads can be at most 25% of revenue, that is a 4:1 (400%) target. Useful when you think in spend ratios.
The proxy value of a lead action. A $5,000 deal closing 6% of the time makes that lead worth $300 to the bidding engine.
Bid to profit, not revenue, by passing margin-adjusted value (send profit as the value). Stops the algorithm chasing high-revenue, low-margin junk.
Plug in your numbers. Everything updates as you type. Start with break-even, set a profit-aware target, value your leads, then check whether you have the volume to run a target at all.
The lowest ROAS that does not lose money, from your gross margin.
The ROAS that leaves the profit you want after ad spend.
If you think in "ads can be at most X% of revenue."
Proxy value to assign each conversion action for VBB.
tROAS needs data. This checks your volume against Google's thresholds and maps a stepped path from where you are to where you want to be.
The order matters. Skip the data step and you are bidding to value you never measured. Skip the actuals check and you choke volume on day one.
You cannot bid to value you do not measure. Confirm your primary conversion sends a value greater than zero. Google only counts conversions with a value above zero as eligible for tROAS. No value, no value-based bidding.
Break-even ROAS is 1 / gross margin. This is the
floor. Any target below it is a decision to lose money on the
marginal sale. Calculator A.
Decide the net margin you want to keep after ad spend. Target ROAS
is 1 / (gross margin - profit goal). Or pick a
cost-of-sale ceiling and invert it. Calculators B and C.
Look at achieved ROAS over the last 30 to 90 days. This is the reality your account delivers today. If your math target sits far above it, you have a gap to close gradually, not a number to flip a switch on.
Start about 10 to 20% above current achieved ROAS, not at the aspirational number. The algorithm needs room to operate. An aggressive target out of the gate forces restricted bidding and kills volume.
Expect a one to two week learning period after you set or significantly change the target. Performance wobbles, then settles. Every change resets the clock, so resist the urge to tinker.
Move in roughly 10 to 20% increments, one move at a time, watching what happens to both volume and value. Calculator E maps the steps. Climbing in stages keeps the machine learning instead of stalling.
Read ROAS over a window that excludes the recent conversion-delay period, or recent days look artificially weak because the conversions have not landed yet. If you just started reporting value, wait about four weeks or three conversion cycles before adopting tROAS.
Which strategy belongs at which stage, and what has to be true before you climb. You do not start at Target ROAS. You earn your way there with data.
VBB performs to the quality of the value you feed it. This is where most accounts quietly sabotage themselves: counting the wrong actions, carrying no value, or double counting across platforms.
A conversion only feeds value bidding if it carries a value above zero. A primary action reporting no value is invisible to tROAS. Confirm this before anything else.
In Google Ads, only Primary conversion actions feed bidding. Keep soft proxies like page views and form starts as Secondary so they inform you without distorting value.
Lead gen counts one conversion per click so a single person filling three forms is not three leads. Ecommerce counts every conversion and passes the real transaction value.
Use data-driven attribution, not last click. Set the conversion window to match your sales cycle so credit lands where it belongs and the target reads true.
Ecommerce hands you the value: it is the cart total. Lead gen does not. A form fill has no revenue attached, so you assign a proxy value that reflects what that lead is worth downstream. This is the step that makes VBB work for services, B2B, home, auto, and finance.
| Conversion action | Closes to a deal | Proxy value | Why |
|---|---|---|---|
| Newsletter signup | 1% | $50 | Barely qualified. Tiny value keeps it in the signal without distorting bidding. |
| Qualified lead | 6% | $300 | Real intent. The workhorse action for most lead-gen accounts. |
| Tour or demo booked | 18% | $900 | High intent. Worth bidding up for, three times a raw lead. |
| Application started | 45% | $2,250 | Bottom of funnel. The algorithm should chase these hardest. |
The settings people get wrong most often, in one place.
| Business model | Objective | Start with | Then move to |
|---|---|---|---|
| Ecommerce | Revenue, ROAS | Maximize Conversion Value |
Target ROAS once value data is stable and volume
clears
|
|
Lead-gen B2C (home, auto, finance, services) |
Qualified leads at a return | Maximize Conversions |
Maximize Conversion Value with proxy values, then
Target ROAS plus offline import
|
| B2B / SaaS | Pipeline value | Target CPA on qualified leads |
Target ROAS on imported deal values once the
cycle closes enough volume
|
| New account, no data | Gather signal | Maximize Conversions with a sane budget |
A target only after the learning phase clears and value is firing |
| Campaign type | Documented minimum | For stability | Note |
|---|---|---|---|
| Search and Display | ~15 value conversions / 30 days | 50+ / 30 days | Per campaign, not per account. Below 15, stay on Max Conversion Value. |
| Shopping | ~20 conversions / 45 days | 50+ / 30 days | Threshold is lower, but richer data still wins. |
| Video Action | ~30 conversions / 30 days | Higher | Floodlight optimization must be enabled in DV360. |
| Learning period | 1 to 2 weeks after any target change | Hold steady | Each change resets the clock. Do not tinker mid-learning. |
| Method | You provide | Formula | Example |
|---|---|---|---|
| Break-even | Gross margin | 1 / margin | 40% margin to 250% |
| Profit target | Margin + profit goal | 1 / (margin - goal) | 40% margin, keep 10% to 333% |
| Cost of sale | Max ad spend share | 1 / cost of sale | 25% ceiling to 400% |
Every one of these is common, and every one quietly drains spend or freezes volume. None of them announce themselves.
Official documentation first. Thresholds and feature names change, so confirm against release notes before you rely on a number.