Target CPA and Target ROAS Explained: The Arithmetic Behind Both

The account sells kitchenware. It has run Target CPA at a fixed cost per order for months, and by the one number the strategy is judged on, it is doing beautifully: cost per conversion sits just under target, order count is up, nobody looks twice. Then someone exports the orders and joins them back to the campaigns, and the picture falls apart. Almost every order the campaign bought is a cheap spatula, while the pan sets that actually pay the rent came from somewhere else entirely. That is the moment a team starts searching target CPA vs target ROAS, usually a few months later than they should have.

The old habit of just watching the cost-per-conversion number go green is exactly why this gets missed for so long. Target CPA optimises for how many conversions it can buy, not what those conversions are worth, so it treats a spatula and a pan set as identical wins and quietly drifts toward whichever one converts cheaper. Nobody catches this by reading the weekly report, because the report only shows the metric the strategy was told to chase, not the revenue actually sitting behind it. By the time someone checks manually, real money has already been steered toward the wrong orders for a while.

This post settles the actual decision, not the theory around it. You will see exactly what each strategy optimises for, the arithmetic that connects them, and one number that tells you in advance whether switching to Target ROAS will help or just add noise. By the end, you will know which target fits your order values right now, and what data you need in place before flipping the switch does anything at all.

What Is the Difference Between Target CPA and Target ROAS?

Both are Smart Bidding strategies, both use the same auction-time signals, and both are aiming at an average over a period rather than promising anything about a single auction. The only real difference between them is one term in the bid calculation, and that one term changes everything downstream: what the system chases, what data it demands from you, how noisy the result is, and what a good week even looks like in the report.

Target CPA vs Target ROAS: What Each One Is Actually Optimising

Strip away the interface and both strategies do the same thing on every auction: predict how likely this particular click is to convert, then set a bid that reflects what that outcome is worth. The difference is in the second half.

Under Target CPA, a conversion is worth exactly one conversion. The bid moves with predicted conversion rate and nothing else. A query the model believes converts at four percent gets roughly twice the bid of one it believes converts at two percent, and the fourteen-dollar spatula and the one-hundred-and-eighty-dollar pan set are, as far as the bidder is concerned, the same event. If the spatula converts at three times the rate — and cheap, low-consideration products almost always do — the strategy will correctly, obediently, relentlessly buy spatulas. It is not malfunctioning. It is doing precisely the job you gave it.

Under Target ROAS, the bid moves with predicted conversion rate multiplied by predicted conversion value, divided by the target you set. Value enters the equation. Now a query that converts at one third the rate but carries thirteen times the basket is worth roughly four times as much to bid on, and the system will pay accordingly. That is the entire mechanism. Everything people say about ROAS bidding being "smarter" or "more advanced" reduces to that single multiplication.

Two consequences follow immediately, and both surprise people.

The first is that a target is not a cap. Neither strategy promises that any individual conversion will cost at or below your target CPA, and neither promises that any individual sale will hit your target ROAS. They aim for an average across the campaign over time. Half your conversions costing more than the target is normal and expected. Teams who treat the target as a maximum bid spend their lives confused by the daily numbers.

The second is that Target ROAS deliberately buys fewer conversions. Raising a ROAS target means instructing the system to enter only the auctions where it expects a high return, which mathematically means fewer auctions, less spend, and less total revenue at better efficiency. If you switch from CPA to ROAS and then panic because conversion count dropped, you have misread the trade you just made. Conversion count is the wrong scoreboard for a value strategy — that is the whole point of moving.

The Arithmetic That Links the Two Targets

People treat CPA and ROAS as two philosophies. They are two views of one relationship:

ROAS = average order value ÷ cost per acquisition.

Rearranged: CPA = average order value ÷ ROAS.

This is worth sitting with, because it tells you exactly when the two strategies are interchangeable and when they are not. If every order in your business were worth the same amount, a target CPA would be a target ROAS, expressed in different units, and choosing between them would be a matter of taste. Nothing would be gained by switching.

The moment order values differ, the equivalence breaks, and it breaks in a way that is easy to underestimate. Take an illustrative example: a target CPA of twenty dollars. If the average order is forty dollars, that target is quietly instructing the system to accept a two-times return. If the average order is two hundred and forty dollars, the same twenty-dollar target is accepting a twelve-times return. Same target, same campaign settings, wildly different economics — and the strategy has no idea, because you never told it what an order is worth.

Run that chart in reverse and you get the diagnostic question for your own account. Take your current target CPA, divide your actual average order value by it, and look at the number. Is that the return you would have asked for if someone had put the question to you directly? For a lot of accounts the honest answer is no — the target CPA was set two years ago from a number somebody remembered, and it has been silently setting the account's return ever since.

The Value Variance Test: The One Number That Decides It

Everything above suggests a simple rule: if value varies, use ROAS. That is right, but "varies" needs a definition you can actually test, because every catalogue varies a little and almost none vary uniformly.

Pull ninety days of orders into a spreadsheet — order value only, one row per order — and calculate four things.

1. The median order value. Not the mean. The mean is dragged around by a handful of large orders and is exactly the number that hides the problem.
2. The mean order value. Now compare it to the median. If the mean is meaningfully higher, you have a long tail of large orders, and a conversion-counting strategy is systematically undervaluing that tail.
3. The tenth and ninetieth percentile. These bound the ordinary range of your business, ignoring freak orders at both ends.
4. The share of total revenue coming from the top ten percent of orders. Sort descending, take the top decile, sum their value, divide by total revenue.

Now read them. These are practical rules of thumb from working with accounts, not platform-published thresholds, so treat them as a starting frame rather than a law:

• If the ninetieth percentile is less than roughly twice the tenth percentile, your order values are effectively uniform. Target CPA is fine. Switching to ROAS adds a data dependency and buys you almost nothing.
• If the top ten percent of orders produce more than about half of total revenue, value-based bidding is where the money is hiding, and Target CPA is actively working against you.
• Anything in between is a judgement call, and the tie-breaker is the next section.

The nuance almost everybody misses: variance has to be predictable

Here is the part that gets skipped, and it is the reason a lot of ROAS migrations disappoint. Target ROAS does not need value to vary. It needs value to be predictable from what the system can see before the click.

If your large orders come from a distinguishable pattern — a different product category, a different query set, a different device, a returning-customer list, a different time of day — then value bidding has something to work with, and it will find the pattern faster than you will.

If your large orders are essentially random within the same traffic — the same query, the same landing page, the same audience, and the difference is simply whether that individual shopper happened to add three items to the basket instead of one — then there is no pre-click signal to learn from. Target ROAS will spend the learning period discovering that value is unpredictable, and will settle into behaviour barely distinguishable from conversion bidding, with more variance and more of your patience consumed.

The quick check: take your value spread and cut it by one obvious dimension — product category, or brand versus non-brand queries, or new versus returning. If the average value differs sharply between the slices, the variance is structured and ROAS will find it. If every slice has roughly the same average, the variance is noise, and you should either stay on CPA or split the campaign so that the structure does the work the bidder cannot.

When Target CPA Is the Better Choice

Target CPA is often framed as the beginner option people graduate out of. That framing costs accounts money. There are situations where it is straightforwardly the correct instrument, and staying on it is the sophisticated decision.

Lead generation where every lead is worth roughly the same

A local services business, a B2B enquiry form, a clinic booking, a demo request into a single-product SaaS. If the sales team treats each enquiry as broadly equivalent and the close rate is stable, then value bidding has no additional information to exploit. Target CPA plus a disciplined allowable cost per lead is a complete answer. The upgrade path here is not to switch to ROAS on invented lead values — it is to import real closed outcomes so that the strategy learns which enquiries turn into business.

Thin conversion volume

Ratios are noisier than costs. A cost per action is an average of a quantity that clusters; a return on ad spend is a ratio whose numerator has a long tail, and a single unusually large order can swing a week's reported ROAS enough to make you change something you should not touch. At low conversion counts, the ROAS figure you are reacting to is mostly sampling noise. In that situation, Target CPA is not a compromise — it is the estimator with less variance, and it will get you to a stable state faster.

Values you cannot yet trust

This is the most common and most damaging case. If your conversion values are partially implemented — value passed on desktop but not in the app, revenue including shipping on one path and excluding it on another, some transactions arriving with a value of zero, refunds never sent back as reversals — then Target ROAS is not an upgrade. It is a strategy optimising towards a distorted picture, and it will produce results worse than the CPA setup it replaced. Fix the values first, verify them for a full month, then switch. Never in the other order.

When the business constraint really is a cost ceiling

Some businesses genuinely operate to an allowable acquisition cost rather than a return: a subscription with a fixed monthly price, a membership, a franchise paying per booked appointment. If the finance answer to "what can we pay for a customer" is a flat number and not a ratio, then a flat number is the right target and translating it into a ratio just adds a step where errors can enter.

When there is no value to send at all

Calls, app installs, form fills with no downstream revenue system, offline businesses with no CRM integration yet. You cannot bid on value you do not have. Target CPA is not a fallback here; it is the only honest option, and pretending otherwise by assigning every conversion an identical made-up value is just Target CPA wearing a costume — and a costume that makes your reporting lie.

When Target ROAS Is the Only Sane Choice

The mirror image. These are the cases where Target CPA is quietly costing you money every day it stays on.

A catalogue with a wide price range

The kitchenware account from the opening. Any retailer whose cheapest item and dearest item differ by an order of magnitude is being badly served by conversion counting, because cheap items convert more easily and a conversion-counting strategy will find them and stay there. The larger the spread, the faster the drift, and the drift is invisible in a CPA report that only ever shows green.

Shopping and Performance Max on a mixed feed

When a single campaign covers a whole catalogue, the bidder is choosing between products on every auction. Without value, it has no basis for that choice other than conversion probability, which correlates with low price. Value bidding is what makes a broad feed campaign viable at all. If you are running a mixed feed on conversion bidding, this is usually the single highest-value change available in the account.

Margin that varies by product

A subtle one. Two orders of equal revenue can carry very different profit. If your margin varies materially across the catalogue, revenue-based ROAS will happily scale spend towards your highest-revenue, lowest-margin lines. The fix is to send profit as the conversion value instead of revenue. This works well and is exactly what value rules and feed-level margin data exist for, but be clear-eyed about the consequence: once you send profit, the "ROAS" in the interface is a profit-on-ad-spend ratio, and your target must be rebuilt on that basis. A four-times target on revenue is a completely different instruction from a four-times target on profit. Write down which one your account is using, because six months later nobody will remember. The mechanics of turning margin into a target are covered in the walkthrough of break Even ROAS Calculator: Find Your Profit Target.

New customers worth more than the first order

If a first purchase reliably leads to repeat purchases, the first order value understates what the customer is worth, and a target built on first-order revenue will underbid for acquisition. Two options: assign a new-customer value uplift where the platform supports it, or set a deliberately lower ROAS target for prospecting campaigns than for remarketing, and document the reason so the next person does not "fix" it.

When you are profit-constrained, not volume-constrained

The strategic version of the question. If your bottleneck is cash — every order has to pay for itself now, growth is funded from margin — then the trade Target ROAS offers is exactly the trade you want. If your bottleneck is growth and you have funding or strategic reasons to buy volume at thinner returns, a lower ROAS target or Maximise Conversion Value with a budget cap serves you better. This is a business decision that happens to be expressed in a dropdown, and it should be made by whoever owns the P&L, not by whoever has interface access.

The Data Each Strategy Needs Before It Behaves

Both strategies fail in the same way when starved, but they starve on different things. Target CPA needs conversion count. Target ROAS needs conversion count and value integrity on every single one of them.

Requirement	Target CPA	Target ROAS
Working conversion tracking	Required	Required
Steady conversion volume	Required — enough for a stable average cost	Required, and more of it — ratios are noisier
A value on every conversion	Not used	Required, with no gaps and no zeros
One currency per account	Helpful	Critical — mixed currencies corrupt every ratio
Refunds sent back as reversals	Minor effect	Major — unreversed refunds inflate the target's basis
Clean primary conversion set	Required	Required, and worse when broken
Known conversion lag	Required to judge results	Required, and lag on value is usually longer
Predictable variance in value	Irrelevant	Determines whether the switch pays off at all

The currency row deserves a paragraph of its own, because it is the failure that looks like a bidding problem and is not. If a single account reports in more than one currency, every cost threshold and every ratio in it is being computed across values that are not comparable. There is no partially correct version of this. Either everything is in one currency or the numbers are fiction. We hit this hard enough on our own product that we made the system refuse to proceed rather than convert automatically — a wrong exchange rate does not produce a slightly wrong answer, it produces a confidently wrong one across every threshold at once. Apply the same rule to your bidding: separate accounts, separate targets, no mental conversion.

Setting the First Target Without Guessing

The most expensive mistake in this entire subject is setting the first target at the number you wish the account would deliver. A target set well below anything the account has ever achieved does not force efficiency. It instructs the bidder to withdraw from auctions it does not believe it can win at that price, so impressions collapse, spend collapses, learning never completes, and after two weeks somebody declares that Smart Bidding does not work in this vertical.

The first target is a description of reality, not a goal. Goals come later, one step at a time.

Deriving a first Target CPA

Take the campaign's own history over a window long enough to be stable — usually thirty to ninety days, depending on volume — and long enough ago that the conversion lag has matured. Do not use last week. Compute cost divided by conversions.

Then look at the weekly figures inside that window rather than just the overall average. If the weekly cost per conversion bounces around a lot, take the median week rather than the mean, because the mean is being set by the outlier weeks you would not want to plan around. Set the first target at that number, or slightly above it if the campaign is currently constrained by low bids. Slightly above is not weakness; it buys a clean learning period, and you can walk it down afterwards.

If there is no history at all, work from the business instead: gross profit per customer, multiplied by the fraction of that profit you are willing to spend to acquire them. That produces an allowable acquisition cost. If your allowable cost is far below what similar campaigns in the account achieve, the problem is the offer or the margin, and no bid strategy will solve it.

Deriving a first Target ROAS

Same principle, one extra decision. First, compute the campaign's actual achieved return over a matured window: conversion value divided by cost. That is your starting target — not your ambition.

Second, compute your break-even return so you know where the floor is. On gross margin, break-even ROAS is one divided by the gross margin. An illustrative example: a forty percent gross margin means a break-even of two and a half times, so every dollar of ad spend needs two and a half dollars of revenue simply to leave you no worse off before overheads. Your target has to sit above that floor by enough to cover fixed costs and leave profit.

Now compare the two numbers, because the comparison tells you which problem you actually have:

• Achieved return is comfortably above break-even. Good. Set the first target at achieved, let it stabilise, then decide whether to step it up for margin or step it down for volume.
• Achieved return is near or below break-even. Setting a target at break-even will strangle the campaign, and setting it at achieved locks in a loss. This is not a bidding problem. It is a margin, offer or price problem wearing a bidding costume, and the honest move is to say so rather than to keep adjusting a dropdown.

A worked illustration

All numbers here are invented for the example. Do not carry them into your account.

Input	Value	Where it comes from
Spend, last 60 days (matured)	$18,000	Campaign report, one currency
Conversion value, same window	$63,000	Same report, values verified
Achieved ROAS	3.5x	63,000 ÷ 18,000
Gross margin	40%	Finance, after COGS and shipping
Break-even ROAS	2.5x	1 ÷ 0.40
First target to set	3.5x	Achieved, not aspiration
Room to move	Down to 2.5x for volume, up for margin	The gap between achieved and break-even

Note what this table does not contain: a competitor benchmark, an industry average, or a number somebody heard on a podcast. Those numbers describe other people's cost structures. Yours is the only one that determines what you can afford to pay.

Why Target Changes Stall Learning

Both strategies recalibrate when you change the target. That is not a bug or a punishment; the target is an input to the bid calculation, so changing it changes the model's operating point and the system needs fresh evidence about how the auction responds at the new level.

The failure pattern is universal and it looks like this. Monday: switch to Target ROAS at achieved. Wednesday: spend is down, panic, lower the target. Friday: volume is up but returns look thin, raise it. Following Tuesday: conversions have dropped, lower it again. Six weeks later the account has never had a single stable observation period, and the conclusion reached is that the strategy does not work — when in truth it was never allowed to run.

Three rules prevent this entirely.

One change at a time. If you change the target, do not also change the budget, the structure, the creative or the conversion settings in the same week. You will not be able to attribute the result to anything, and you will have burned the observation period for nothing.

Move in small steps. Around ten to fifteen percent per move is the practical range. The response to a target change is not linear — each step changes which auctions the campaign enters, and a large jump can move it into a completely different competitive set. Small steps let you find the point where volume falls off a cliff before you fall off it.

Set two dates before you touch anything. The date of the change, and the date you are allowed to look. The gap between them is the learning period plus your conversion lag — and for value bidding, the lag on value is often longer than the lag on the conversion itself, because refunds, cancellations and delayed shipping all land later. Anything between those dates is time you have already agreed not to interfere with. Write the dates in a shared calendar so that when someone asks on day four, the answer is a link rather than an argument.

One exception worth knowing: if you have a known short event coming — a sale weekend, a product launch, a period where conversion rate will genuinely be different — the correct tool is a seasonality adjustment, not a target change. It tells the system to expect a temporary shift and then revert, without treating the promo as the new normal. The detailed mechanics of seasonality adjustments and data exclusions sit in the Smart Bidding guide linked at the top of this article.

Portfolio Versus Campaign Level

Both strategies can be set on a single campaign or on a portfolio shared across several. The choice is mostly about data volume against control.

A portfolio pools conversion data across its member campaigns. For an account with several thin campaigns that individually never see enough conversions to hold an average, this can be the difference between a strategy that works and one that never leaves learning. Portfolios also expose settings that campaign-level Smart Bidding generally does not, such as bid limits — check what your interface currently offers before building a plan around it, since these options move.

The costs are real. You lose per-campaign target control, which means every campaign in the portfolio is being held to the same standard whether or not that makes sense. Budget and traffic flow towards whichever campaigns look most efficient, so a strong campaign can quietly subsidise a weak one and you will not see it in the campaign-level report. Diagnosis gets harder in general, because the thing making decisions is no longer the thing you are looking at.

Two rules that hold up in practice. Only pool campaigns that share economics — same margin structure, same customer type, same expectations. And never pool brand with non-brand, in either strategy: brand traffic converts cheaply and returns beautifully, so it will flatter the portfolio average and mask non-brand campaigns that are losing money underneath it. Different jobs, different economics, different strategies.

Migrating Between Them Without Burning a Month

Most of the value in a migration sits in the steps that happen before the dropdown changes. The dropdown itself takes four seconds.

Moving from Target CPA to Target ROAS

Step one: fix values, then wait a full month. Every conversion carries a value, the value means the same thing on every path, one currency, refunds reversed, no zeros hiding in the data. Then let a clean month accumulate so the strategy has honest history to learn from. Switching on day one of clean values gives the model a training set that is mostly the broken era.

Step two: record a baseline outside the platform. Spend, conversions, conversion value, cost per conversion and achieved ROAS for a matured window, written into a document with the date. Interface comparisons shift under you as late conversions arrive; a written baseline does not.

Step three: compute the implied target. Conversion value divided by cost over that matured window. That number is the first target. Resist every instinct to round it up towards what you want.

Step four: switch, and change nothing else. Same budget, same structure, same creative, same conversion settings. If the campaign has been budget-limited under CPA, resolve that as a separate change on a separate date — a strategy that spends to its target combined with a freshly raised budget produces a result nobody can interpret.

Step five: judge it on value, not on count. Conversion count will very likely fall. That is the trade you chose. The questions that matter are whether total conversion value held or grew, whether achieved return moved towards the target, and whether profit improved after margin is applied. If you present conversion count to a stakeholder as the headline after a ROAS migration, you will be asked to switch back, and you will deserve it.

Step six: only then, step the target. One move, ten to fifteen percent, two dates, wait.

Moving from Target ROAS to Target CPA

Less common, and usually a symptom rather than a strategy. Before you do it, check whether the real problem is broken values — if so, moving to CPA hides the breakage instead of fixing it, and you will carry the same corrupt data into every report you build afterwards.

There are legitimate reasons. The business pivoted from ecommerce to lead generation. The catalogue narrowed until price spread disappeared. Conversion volume fell to the point where ratio-based bidding is pure noise. In those cases, migrate the same way in reverse: derive the first target CPA from what the campaign actually achieved during the ROAS period, change nothing else, and give it a full learning period plus lag before judging.

Mistakes That Look Like a Bid Strategy Problem

Micro-conversions sitting in the primary conversion set

Newsletter signups, add-to-carts, video views marked as primary. Under Target CPA the strategy will buy the cheapest thing in that set, which is never the sale. Under Target ROAS it is worse, because those actions typically carry no value or a token one, dragging the average down and making the target unreachable. Clean the primary set before blaming the strategy. This is a fifteen-minute job that fixes more accounts than any bidding change.

Partial values and silent zeros

A checkout that passes value on one payment method and not another. A subscription that records the first payment only. A mobile path where the value variable never populates. Sort your conversions by value and look at how many are zero. If the answer is more than a rounding error, your ROAS target is being computed against a distorted base and every conclusion built on it is wrong.

Comparing before and after on different attribution settings

If the attribution model or conversion window changed anywhere near the migration date, the before and after numbers describe different universes. Check the change history before you draw a conclusion, and if the settings did move, extend the baseline window to a period entirely inside one regime.

Judging on a seven-day slice

Especially damaging for ROAS, where a single large order can lift a week and its absence can sink the next one. If your conversion volume is such that one order can move the weekly figure by a fifth, you cannot read weekly figures. Use a longer window or accept that you are reading noise.

Importing a target from a benchmark

"Four times is the standard for ecommerce" describes somebody else's margin, price point and competitive set. A four-times target is generous for a business on seventy percent margins and fatal for one on twenty-five. The break-even calculation takes two minutes and replaces the entire genre of benchmark advice.

Splitting one campaign in two to run a fair test

Tempting, and it does not work the way people hope. Two campaigns targeting the same queries compete in the same auctions, so each one's results are contaminated by the other's behaviour. If you need a genuine comparison, use the platform's own experiment mechanism, which splits traffic rather than splitting the campaign — and even then, give it a full learning period plus lag.

Changing the target the same week as a price change

A price rise changes conversion rate and order value simultaneously, which is precisely what both strategies are modelling. Do one, wait, then do the other. Otherwise you have two explanations for every result and no way to choose between them.

How Far Manual Work Goes, and Where a Tool Helps

Everything in this article can be done by hand, and for one account with three campaigns it should be. The spreadsheet takes twenty minutes, the break-even calculation takes two, and the discipline of two dates on a calendar costs nothing.

Manual work runs out in a specific and predictable place: not at the analysis, but at the memory. Six accounts across three platforms, each with its own targets, each moved by different people at different times. Three weeks later somebody asks why the target on a campaign is what it is and nobody can answer, because the change lives in a platform log with no reasoning attached. The comparison you carefully baselined is now unreadable, because two other things changed in the same window and nobody wrote them down. This is the real cost of manual bid management, and it is not the hours — it is that decisions become unauditable and you stop being able to learn from your own account.

That is the gap we built Orova Ads to close. It syncs campaigns, ad groups, ads and daily figures from Google Ads, Meta and TikTok into one table, carries 214 optimisation action codes across the three platforms, and splits every one of them into an advisory code where the AI recommends and you act, and an execution code where you have explicitly allowed it to act for you — with three operating modes, advisory being the default. Rules are written in plain language with data variables in them, so "if this campaign's cost per conversion has been above target for seven days, propose a target change" is a sentence rather than a script, and every proposal lands in a history tab with the reasoning and the numbers that produced it, approved or rejected or expired, logged either way. That log is the thing manual management cannot give you: three months later you can see exactly which target moved, when, why, and what happened next.

Frequently Asked Questions

Can I run Target CPA and Target ROAS in the same account?

Yes, and in many accounts you should. Lead generation campaigns on CPA and ecommerce campaigns on ROAS in the same account is a perfectly coherent setup, because they are different businesses with different economics. What you should not do is mix them inside one portfolio, or compare their headline metrics side by side in a report as if they were measuring the same thing.

My conversions dropped after switching to Target ROAS. Did it fail?

Probably not — that is the expected behaviour. A ROAS target instructs the system to skip auctions it does not think will return enough, so fewer conversions at higher value is the design, not a fault. Check total conversion value and achieved return against your written baseline. If value held or grew while spend fell, the switch worked. If value fell proportionally more than spend, then you have a real problem and the first suspect is your conversion values, not the strategy.

Is Target ROAS just Target CPA with extra steps?

Only if every order is worth the same. Then the two are algebraically the same instruction expressed in different units and there is no reason to switch. The moment values differ in a way that can be predicted before the click, ROAS is seeing information CPA is structurally blind to, and the gap between them widens with the spread of your order values.

How long should I wait before judging a target change?

The learning period plus your conversion lag, and both are account-specific. Measure your lag directly: compare what a given week reported at the time against what the same week reports a fortnight later, and the difference tells you how long value takes to arrive. Add that to the learning period and set your review date accordingly. For accounts with low volume this can be a long time, which is itself useful information — if you cannot wait that long, your bottleneck is conversion volume, not bid strategy.

My Target ROAS campaign has stopped spending. What do I do?

Almost always the target is set above anything the campaign has achieved, so the bidder cannot find auctions it believes will meet it and withdraws. Check your written achieved-return baseline. If the target is above it, lower the target to achieved, wait a full cycle, and step it up from there. If the target is at or below achieved and spend still collapsed, look for a change that landed at the same time: budget, conversion settings, attribution, or a structural edit somebody made without telling you.

Should I send revenue or profit as the conversion value?

Profit is better if you can produce it reliably, because it makes the strategy optimise for the thing you actually care about rather than a proxy for it. But do not do it halfway. Once you send profit, your target has to be rebuilt on a profit basis, your reports need relabelling, and everybody looking at the account needs to know which basis is in use. A partial migration where some conversions carry revenue and others carry profit is worse than either pure option, and it is very hard to spot after the fact.

Does the choice between them matter more than the account structure?

No. A well-structured account on a slightly wrong bid strategy beats a badly structured one on the perfect strategy, every time. If your conversion volume is scattered across twenty thin campaigns, neither strategy has enough data to work with and consolidation is the real project. Get the structure and the measurement right, then choose the target type. The dropdown is the last decision, not the first.

What To Do This Week

Pick the campaign carrying the most spend and do not change its bid strategy. Open a spreadsheet instead.

Export ninety days of orders or leads with their values, one row each. Calculate the median, the mean, the tenth and ninetieth percentiles, and the share of revenue from the top ten percent of orders. That takes fifteen minutes and it answers the question this entire article is about — whether the difference between target CPA vs target ROAS is worth anything in your specific business, or whether you have been reading about a problem you do not have.

Then do the two-minute check on your current setup. Divide your average order value by your current target CPA. Whatever number comes out is the return your account has been quietly accepting. If you would not have chosen that number deliberately, you now know what to fix.

And if the spreadsheet says your values are missing, partial or zero on a meaningful share of conversions, stop there. That is this week's entire job. No bid strategy — neither one — can outperform data that is not telling the truth, and every hour spent on the dropdown before the values are clean is an hour spent making a confident decision on a false premise.

Making the Switch Without the Guesswork

Figuring out whether your order values vary enough to justify Target ROAS means pulling conversion data, joining it back to campaigns, checking whether the variance is predictable before the click, and watching the account for weeks afterward to see if the switch actually paid off. Done by hand, that is hours of exporting and cross-checking every time you want to sanity-check a bidding decision, and it is easy to draw the wrong conclusion from too small a sample.

Orova Ads is built to handle that kind of checking automatically, so you can see where your value variance actually sits and whether a bidding switch is likely to help before you commit budget to finding out the hard way. If this is the decision you are sitting with right now, it is worth taking a look.

See Orova Ads