ROAS campaign first-timer: edit old campaigns or start fresh?

brandvoice

I've run two install volume campaigns - one in Tier 2 countries and one in the USA. Now I want to run a ROAS campaign, but I keep reading that ROAS campaigns need data before starting.

I already have data from my install volume campaigns, so I'm confused about next steps. Should I edit the existing campaigns and switch them to ROAS, or create completely new ROAS campaigns?

If I create new campaigns, will they still be able to use the data from the previous install campaigns for optimisation?

Has anyone here successfully run ROAS campaigns? Would really appreciate guidance - I'm worried about wasting budget if I do this wrong.

wordweaver

Oh for the love of god, do NOT touch your existing campaigns. That's a one-way ticket to resetting all your learning and watching your performance tank. Seen it happen too many times.

Start fresh. Build brand new ROAS campaigns from scratch.

Will the platform borrow data from your old install campaigns? Sort of. There's some cross-campaign learning at the account level, but don't count on it doing the heavy lifting for you.

Here's the actual game plan:

Grab your highest-value purchasers from those install campaigns - not people who just downloaded, but the ones who actually opened their wallets
Build a Value-Based Lookalike from that goldmine
Drop that audience into your new ROAS campaign
Set your Target ROAS embarrassingly low to start - like 70-80% of breakeven. You can crank it up later once it's stable
Budget needs to be at least 10-15 times your average CPA per day. No cheaping out
Leave it alone for a full 7 days after launch. No tweaks, no optimisations, no panicking. Let the algorithm do its thing

One more reality check: if you haven't had at least 50 purchase conversions in the last month, you're not ready for a ROAS campaign yet. You simply don't have enough data. Run a standard purchase optimisation campaign first and come back when you've got the numbers.

Good luck - you'll need it.

buggedout

I'd migrate the existing campaigns to preserve historical data and learning curves.

Key things I've found:

Conversion value tracking needs to be airtight. ROAS campaigns get wrecked if your value data is patchy.
Keep US and Tier 2 completely separate - lumping them together kills targeting accuracy.

communityfirst

Great to see someone diving into ROAS! One thing I've learned the hard way is that these campaigns really need enough data to optimise properly. Converting an existing install volume campaign into ROAS can backfire because changing the bid strategy resets the learning phase. When you start a fresh ROAS campaign, Google can use some account-level historical data, but campaign-level learning starts from scratch - so there's only partial benefit, not a full transfer.

ROAS optimisation typically needs at least 15 to 20 revenue events per month to work well. With only installs, the algorithm doesn't have enough signal to optimise toward revenue. The best approach from my experience is to keep your install campaigns running, make sure revenue tracking is properly set up, and once you have enough data, launch a new ROAS campaign with a conservative target.

dailygrind

Honestly, this is one of those traps that catches almost everyone at least once. I've seen it kill ROAS campaigns inside a week because people keep tweaking.

Here's what I've learned from running these at scale (FinTech, so LTV is everything):

Always build fresh campaigns from scratch for ROAS. The install data is useful for seeding audiences, but editing an existing volume campaign completely resets the learning phase. You lose all the optimiser's signal.
Keep your volume campaigns running in parallel. Don't pause them. Let the ROAS campaigns learn in a separate ad set or campaign structure. The volume campaigns will keep your funnel filled while the ROAS ones slowly find the high-LTV users.
Patience is brutal but necessary. Expect 7-10 days of junk performance before you see any directional signal. If you touch the creative, budget, or bid strategy in that window, you're back to square one.
Use a dedicated attribution window. Standard 7-day click or 1-day view will look terrible early on. I'd recommend a multi-touch model (Google Analytics + GSC data to sanity-check) rather than relying on Meta's in-platform numbers.

One more thing: don't try to do this with a tiny budget. If your ROAS campaign can't generate at least 50 conversions per week in the learning phase, you're better off sticking with volume and accepting lower returns until you can scale the budget.

Been there, wasted the ad spend. Live and learn.

24fire

I've watched more than a few founders light their budgets on fire chasing vanity numbers. ROAS campaigns are like sailing-knowing where the wind blows helps, but without a compass you'll end up wrecked. Find someone who's navigated those waters before.

brandvoice

Oh mate, I feel your pain. Running ROAS campaigns when your only revenue is from ads-and no purchase conversions-is a total nightmare. You've got install and first_open from Firebase, linked ad_impression to Google Ads, but it's showing values with zero conversion counts? That's maddening.

Here's the thing: Google Ads treats ad_impression as a secondary conversion action by default when you're running an install volume campaign. That means it won't count toward your primary conversion column-it's just there for reporting. To make it actually count, you'd need to set it as a primary conversion action in Google Ads under "Conversions." But wait-if you do that for an install campaign, it might mess with your optimization goals.

Honestly, the real struggle is that without purchase or in-app revenue events, ROAS campaigns are basically blind. You're trying to optimize for value that doesn't exist in the conversion funnel yet. I've been there-it feels like shouting into the void.

What kind of ad revenue are you tracking? If you can fire a custom event (like ad_revenue or earn_ad_impression) back to Firebase and then to Google Ads as a conversion, that's your best bet. Otherwise, you're stuck using installs as a proxy, and ROAS will be a guessing game.

Hope that helps without making you want to throw your laptop out the window.

wordweaver

I'd rather watch an hour of unskippable ads than try to salvage a ROAS setup that's got ad_impression as a secondary conversion. You've basically told Google "track this, but don't bother bidding on it" - which, spoiler, Google will happily ignore.

Your main problem: ROAS campaigns need a primary conversion with real dollar values attached. Right now you've got impressions with values but no actual conversion count because they're sitting in the secondary pool, just gathering data like a neglected Excel sheet.

What I'd do:

Head to Google Ads > Conversions > snatch that ad_impression event
Switch it from Secondary to Primary - stop treating it like a backup dancer
Make sure "Count = One", not "Every" (unless you want to pay for every single eyeball)

Give it 48-72 hours to reprogram the algorithm's brain. Then build your ROAS campaign fresh - don't try to Frankenstein an existing one.

Thing is, ROAS normally feeds on purchase events, not impressions. If your whole model lives off ad revenue, you'd be smarter to optimise for ad_impression value directly, or use the AdMob + Google Analytics for Firebase hybrid. Standard ROAS bidding will just leave you shouting at a dashboard.

Make sense, or should I draw it on a napkin?