What Happens When You Optimise Google Ads for Revenue, Not Form Fills
The theory is well known by now. Stop telling Google to find cheap form fills, start telling it which leads turned into money, and the algorithm goes looking for buyers instead of browsers.
The theory is right. What most write-ups leave out is the part that decides whether it works: the switch has a hard conversion volume requirement, and most B2B SaaS accounts fail it. Google requires at least 15 conversions in the past 30 days before Target ROAS will run on Search. If you move your primary conversion from form fills to closed deals, you may land well under that and the strategy simply will not function.
So here is what actually happens, in what order, with the traps.
Quick answer
Switching Google Ads from form fills to value-based bidding produces a predictable J-curve:
| Phase | What you see |
|---|---|
| Weeks 1 to 4 | Lead volume drops, cost per lead rises, the algorithm re-enters learning. Nothing looks good |
| Weeks 4 to 12 | Qualification rate climbs, cost per SQL starts falling, cost per lead stays high |
| Month 4 onward | Cost per customer falls and keeps falling as the signal compounds |
Three things decide whether you get through the J-curve:
- Pick a conversion event with enough volume. At least 15 a month, ideally more. For most B2B SaaS that is SQL, not closed won.
- Send gross profit, not revenue. They are different numbers and optimising for the wrong one buys you big unprofitable deals.
- Give it 4 weeks or 1 to 2 conversion cycles, whichever is longer, before judging it. That is Google’s own guidance, and acting earlier is how teams abandon a working change.
Key takeaways
- Target ROAS needs at least 15 conversions in the past 30 days on Search and Shopping. This single requirement rules out optimising on closed deals for most B2B SaaS.
- Revenue and profit are not the same optimisation target. A $120,000 services-heavy deal at 45 percent margin is worth less than two $30,000 deals at 80 percent.
- Use primary and secondary conversion actions to separate what bidding optimises toward from what you merely want to see in reports.
- Conversion value rules can raise the value of conversions from your first-party audience lists, which is the closest Google gets to optimising for company fit.
- The plumbing changed. By April 2026 Google consolidates enhanced conversions for web and leads into one setting, and imports should carry user-provided data plus GCLID, not GCLID alone.
- Expect to wait. Google recommends reporting values for 4 weeks or 1 to 2 conversion cycles before setting a target, and another 1 to 2 cycles after changing it.
Revenue or profit? They are not the same
The most common question on this topic is whether to optimise for revenue or for actual profit, and it matters more in B2B SaaS than in most categories, because margin varies so much by segment.
| Segment | ACV | Gross margin | Value if you send revenue | Value if you send profit |
|---|---|---|---|---|
| Enterprise, services heavy | $120,000 | 45 percent | $120,000 | $54,000 |
| Mid-market, standard | $30,000 | 80 percent | $30,000 | $24,000 |
| SMB self-serve | $6,000 | 90 percent | $6,000 | $5,400 |
Send revenue and you have told Google that the enterprise deal is worth four times the mid-market one. Send gross profit and it is worth 2.25 times. The algorithm will chase the first number as hard as you let it, which in practice means bidding aggressively on the segment that generates the most invoice and the least margin, often with the longest sales cycle and the heaviest implementation load.
Send gross profit. If you can, send contribution margin, which also nets off the cost to serve. The simplest version that is still correct is ACV multiplied by your gross margin percentage.
What value to actually send
Do not invent values. Derive them, using one line of arithmetic:
Value = ACV × gross margin × probability of closing from this stage
That gives you an expected profit for each funnel stage. Worked for a $30,000 ACV at 80 percent gross margin:
| Stage you send | Probability of closing | Value to send |
|---|---|---|
| Raw form fill | 2 percent | $480 |
| MQL | 5 percent | $1,200 |
| SQL | 20 percent | $4,800 |
| Opportunity created | 35 percent | $8,400 |
| Closed won | 100 percent | $24,000 |
Two things this gets you that arbitrary values do not. The ratios are correct, which is what the algorithm actually learns from. And when a stage’s close rate changes, you have a rule for updating the value rather than a guess.
If you serve several segments with genuinely different economics, calculate a value per segment rather than one blended number. That difference is most of the signal.
The volume trap that breaks this
Here is the part that is missing from almost every article on value-based bidding.
Google requires at least 15 conversions in the past 30 days at the conversion tracking level before Target ROAS will run on Search and Shopping campaigns. Display needs 15 with valid values across campaigns, Demand Gen needs 50 in 35 days with 10 in the last 7, and Video Action needs 30 in 30 days.
Now apply that to a realistic B2B SaaS funnel producing 400 form fills a month:
| If your primary conversion is | Monthly volume | Target ROAS viable? |
|---|---|---|
| Form fill | 400 | Yes, but this is the problem you are trying to fix |
| MQL | 100 | Yes |
| SQL | 24 | Yes |
| Closed won | 6 | No, below the threshold |
This is why “optimise for revenue” fails so often in practice. Teams move the primary conversion all the way down to closed won, because that is the number the business cares about, and the strategy quietly cannot operate.
The answer is to optimise on the deepest stage that clears the threshold with room to spare, and carry the value of the stages below it. For most B2B SaaS that is SQL, valued at the expected profit of an SQL. You are still optimising toward revenue; you are just using a predictor of revenue that occurs often enough for the algorithm to learn from.
If even SQL is under 15 a month, you are not ready for value-based bidding. Fix lead volume or run manual and target CPA bidding with tight targeting until you are.
Primary and secondary conversion actions
A related mechanism solves the reporting problem this creates. Google lets you mark each conversion action as primary or secondary. Primary actions are counted in your Conversions column and used by bidding. Secondary actions are recorded for reporting only and do not influence bids.
So the correct setup is not to delete your low-value conversions. It is:
- Primary: the one stage you want bidding to optimise toward, carrying its profit value
- Secondary: everything else, newsletter signups, content downloads, pricing page views, kept for visibility and for understanding the funnel
This is how you stop the algorithm chasing a free guide download while still being able to see how many people downloaded it.
Can Google optimise for company fit, not just form fills?
Partly, and the mechanism is underused. Conversion value rules let you adjust the value of a conversion based on conditions, and Smart Bidding uses the adjusted value in real time for Target ROAS and Maximise conversion value.
You can set rules on:
- Audiences, including your own first-party audience lists
- Geographic locations
- Device types
- No condition, for certain goals such as store visits
The audience condition is the useful one for B2B. Upload your target account list as a customer list, then write a rule that multiplies conversion value for anyone on it. The algorithm now bids more for people from companies you actually want, before anyone has filled in anything.
Each rule supports a primary and a secondary condition, and they apply to Search, Shopping, Display, Travel and Performance Max. Note that some conditions are restricted for advertisers running housing, employment or credit ads.
This is not full ICP targeting, because it depends on match rates on your uploaded list. But it is a genuine way to express company fit inside the bidding system, and it works alongside the offline conversion values rather than instead of them.
The plumbing changed in 2026
If you set this up before 2025, or you are following a guide written then, check this.
Google is consolidating enhanced conversions for web and enhanced conversions for leads into a single on and off setting, and from April 2026 accepts user-provided data through website tags, Data Manager and API connections. Existing users are migrated automatically by then.
The practical change is what you send. The old instruction was to capture the GCLID and upload against it. The current instruction is to send more identifiers:
| Identifier | Status |
|---|---|
| User-provided data, hashed email and phone | Now expected |
| GCLID | Still recommended for accuracy |
| GBRAID and WBRAID | Include where present |
| Order ID | Include, to prevent duplicates |
Google’s stated sequence is to turn on enhanced conversions from offline sources in Settings, configure the conversion actions through Data Manager or the API, upload with the identifiers above, and then after 3 conversion cycles or 4 weeks switch the new enhanced conversion action to primary.
GCLID-only imports still work, but they lose every conversion where the click ID was not captured or was lost between the form and the CRM, which in practice is a meaningful share.
What the timeline actually looks like
| Period | What to do | What to expect |
|---|---|---|
| Week 1 | Audit tracking. Calculate close rates by stage and gross margin. Work out which stage clears 15 conversions a month | Nothing visible |
| Weeks 2 to 3 | Set up offline import with the full identifier set. Assign profit-based values. Set primary and secondary actions | Nothing visible |
| Week 4 onward | Let values report for 4 weeks or 1 to 2 conversion cycles, whichever is longer, before setting a target | Data accumulating |
| Then | Switch to Maximise conversion value, add a Target ROAS once you have a baseline | Volume drops, CPL rises, learning period |
| Next 1 to 2 conversion cycles | Do not touch it | Qualification rate starts improving |
| Month 4 onward | Add conversion value rules, refine values by segment | Cost per customer falling |
The most common failure is impatience in the fourth row. Lead volume falls, someone escalates, and the change is reverted two weeks before it would have paid off. Agree in advance, in writing, what the expected volume drop is and how long you will leave it alone.
A worked illustration
To show the arithmetic rather than to promise a result. Same $20,000 monthly spend throughout:
| Before | After | |
|---|---|---|
| Form fills | 400 | 180 |
| SQLs | 40 | 72 |
| SQL rate | 10 percent | 40 percent |
| Closed deals | 2 | 6 |
| Cost per lead | $50 | $111 |
| Cost per customer | $10,000 | $3,333 |
Cost per lead more than doubled. Cost per customer fell by 67 percent. That is the whole argument in one table, and it is why judging this change on cost per lead will always tell you it failed.
Be honest about the size though. The illustration above shows an SQL rate going from 10 to 40 percent, which is at the very optimistic end. A realistic first-year outcome is a meaningful improvement in qualification rate and a clear fall in cost per customer, not a quadrupling. The direction is reliable. The magnitude depends on how much waste you had to begin with.
When not to do this
- Under 15 conversions a month at every stage. Build volume first.
- No reliable CRM stage data. If your SQL definition changes depending on who you ask, you will send the algorithm noise.
- Sales cycle longer than about 6 months with no mid-funnel signal. You need something that happens sooner than closed won to optimise on.
- You cannot tolerate a volume drop for a quarter. This change costs you lead volume before it pays. If the business cannot absorb that, fix waste and targeting first and come back to bidding later.
Frequently Asked Questions
Q1. What happens when you switch Google Ads from form fills to revenue optimisation?
Lead volume falls and cost per lead rises almost immediately, because the algorithm stops buying the cheapest conversions. Qualification rate then improves over the following one to two months, and cost per customer falls from around month three or four. The pattern is a J-curve, and judging it in the first month will always look like failure.
Q2. Should I optimise Google Ads for revenue or for profit?
Profit. Revenue over-weights large low-margin deals, which in B2B SaaS are often the services-heavy enterprise contracts with the longest cycles. Send ACV multiplied by gross margin, or contribution margin if you can calculate it.
Q3. Should I optimise for form fills or for qualified leads?
Qualified leads, provided you have at least 15 of them a month. That threshold is what decides it, not preference. If SQLs are below 15 a month, optimise on the deepest stage that clears it and use tighter targeting to control quality.
Q4. How many conversions do I need for Target ROAS?
At least 15 in the past 30 days at the conversion tracking level for Search and Shopping. Display needs 15 with valid values across campaigns, Demand Gen 50 in 35 days with 10 in the last 7, and Video Action 30 in 30 days.
Q5. Can Google Ads optimise for company fit rather than just form fills?
Partly. Conversion value rules let you raise the conversion value for people on your first-party audience lists, and Smart Bidding uses that adjusted value in real time. Upload your target account list as a customer list and multiply its value. It depends on match rates, so treat it as a strong signal rather than a filter.
Q6. What values should I assign to each conversion?
Expected profit, calculated as ACV multiplied by gross margin multiplied by the probability of closing from that stage. The ratios between stages matter more than the absolute numbers, and deriving them means you have a rule for updating them when close rates move.
Q7. How do I track reporting conversions without them affecting bidding?
Mark them as secondary conversion actions. Secondary actions appear in reporting but are not used by bidding. Keep one primary action, the stage you want optimised toward, and make everything else secondary.
Q8. Do I still need to capture the GCLID?
Yes, it is still recommended for accuracy, but it is no longer sufficient on its own. Current guidance is to send user-provided data such as hashed email and phone alongside the GCLID, plus GBRAID and WBRAID where present and an order ID to prevent duplicates.
Q9. What is changing with offline conversion imports in 2026?
Google is consolidating enhanced conversions for web and leads into a single setting, and from April 2026 accepts user-provided data through website tags, Data Manager and API connections. Existing accounts are migrated automatically, though setting it up now improves accuracy sooner.
Q10. How long before I see results?
Google recommends letting values report for 4 weeks or 1 to 2 conversion cycles, whichever is longer, before setting a target, then another 1 to 2 cycles after any target change. In practice, expect roughly three months before cost per customer clearly improves, and plan the first month as a loss.
Q11. My sales cycle is nine months. Can I still do this?
Yes, but not on closed won. Optimise on the earliest stage that genuinely predicts revenue, usually SQL or opportunity created, and value it at the expected profit given its close rate. Then track the cohort through to revenue as a reporting exercise rather than as the bidding signal.
Q12. Why do agencies avoid this?
It makes month to month reporting look worse before it looks better, it needs CRM access and technical setup that many do not have, and it exposes performance that cost per lead reporting was hiding. The first month of a correct implementation is indistinguishable from a bad month, which is uncomfortable to present.
Q13. Will my cost per lead go up permanently?
Yes, and that is the intended outcome. You are buying fewer, better leads. The metric that should fall is cost per customer. If cost per lead rises and cost per customer does not fall within about three months, something else is wrong, usually the values or the conversion data quality.
Q14. What if I get this wrong?
The most common errors are optimising on a stage with too little volume, sending revenue instead of profit, and reverting too early. All three are recoverable. Set the values, let it run a full learning period, and change one thing at a time.
Want this set up properly?
We build value-based bidding for B2B SaaS accounts, including the CRM integration, the conversion values and the reporting that makes the J-curve legible to your leadership while it is happening. If you have tried this before and reverted it, that last part is usually why.
Related reading
- Send HubSpot offline conversions to every ad platform
- The ICP scoring system for B2B SaaS paid ads
- The real reason B2B SaaS paid ads attract junk, and the fix on each platform
- The B2B SaaS Google Ads audit checklist
- Are B2B ads getting too expensive in 2026?
- The B2B Google Ads Waste Report 2026
Sources
- Google Ads Help, Target ROAS bidding (conversion volume requirements by campaign type; report values for 4 weeks or 1 to 2 conversion cycles before setting a target): https://support.google.com/google-ads/answer/6268637
- Google Ads Help, About conversion value rules (conditions: audiences including first-party lists, locations, devices; used in real time by Smart Bidding for Target ROAS and Maximise conversion value): https://support.google.com/google-ads/answer/10518330
- Google Ads Help, How to upgrade offline imports (consolidation of enhanced conversions into a single setting from April 2026; user-provided data, GCLID, GBRAID and WBRAID, order ID; swap to primary after 3 conversion cycles or 4 weeks): https://support.google.com/google-ads/answer/15479791
- Conversion values and the worked illustration in this article are calculated from the stated ACV, gross margin and close rates. Substitute your own numbers before setting values in your account.
About the author
Ishan Manchanda is Co-Founder at GrowthSpree, a B2B SaaS marketing agency and Google Partner and HubSpot Solutions Partner rated 4.9 on G2. GrowthSpree manages $60M+ in B2B SaaS ad spend across 300+ accounts, optimising paid media on cost per SQL and pipeline rather than clicks.
