You launched a new ad campaign four days ago. You check it every few hours. Cost per lead is high, conversions are near zero, and the temptation to pause everything and rebuild it from scratch is strong. Before you do that: a rough first week is normal, and it usually means the platform is still gathering data, not that your campaign is broken.
Automated bidding on Google Ads, Meta, and similar platforms runs on this same basic mechanic. When you launch a new campaign or make a significant change to one, the platform doesn't yet know who to show your ad to or how much to bid for the best results. It has to find out by running the ad, watching what happens, and adjusting.
Meta calls this a learning phase: the algorithm collects performance data so it can serve ads to the audiences most likely to take the action you want. Google Ads has a similar learning period for campaigns using automated bidding strategies. During the Google Ads learning phase, an automated bidding campaign gathers performance data across contexts like time of day, location, and targeting, then adapts delivery based on where the ads perform best.
Why the first few days look worse than they are
During this window, the platform is essentially running small experiments. It's testing different audiences, placements, and bid amounts to see which combinations produce results, and it hasn't collected enough data yet to know which ones work for your business. That means costs swing, results are inconsistent, and the numbers you're staring at on day three or four don't reflect what the campaign will look like once it settles.
When you create a new ad set or make a significant change, the Meta ad set enters a learning phase while the system explores which people and placements drive your optimization event, like purchases or leads.
During this period, cost per result and delivery can swing, and once the system has enough data, it exits learning and delivery typically stabilizes.
How long this takes depends on how quickly the platform accumulates enough signal, not just how many days have passed. Agency posts describing Google's process give a fairly wide range: some estimate a week to one month, while others describe it as closer to 1–2 weeks while Google's algorithm collects data to optimize.
For Meta, the repeated benchmark is roughly 50 conversion events per ad set, with Meta suggesting at least a week of data collection to exit the Learning Phase. That 50-conversion figure shows up across nearly every article on the topic, but it traces back to Meta's own help documentation rather than any of the blogs repeating it, so treat it as a rule of thumb rather than a hard number for your specific account.
The exact figure is less important than the constraint behind it: a campaign that isn't generating enough conversions per week will struggle to exit this phase at all, regardless of how many days go by. A campaign with a tiny daily budget chasing an expensive goal may never gather enough data to stabilize, because the volume of results it produces per week never gets high enough for the algorithm to learn from.
Why editing the campaign mid-week backfires
This is where a lot of business owners managing their own ads get stuck: when the first few days look bad, the instinct is to go in and change things. Swap the headline, adjust the budget, tighten the audience, switch the bidding goal. Each of those changes can restart the process you were trying to get through.
The learning phase can restart when you make a change that significantly affects how the algorithm bids, including switching bidding strategies. When that happens, the platform goes back to gathering fresh data instead of building on what it already learned. A campaign that gets edited every two or three days because the numbers look rough never gets the chance to stabilize. It stays in a permanent state of starting over.
Not every change resets the clock. Small changes, like minor bid adjustments or adding a few keywords, usually don't trigger a full learning phase reset. And if you're running Google Ads with manual bidding rather than an automated strategy, manual CPC does not undergo this learning period at all, because there's no algorithm making bidding decisions on your behalf. That's a meaningful distinction if you're trying to figure out why one campaign feels volatile and another doesn't.
How to tell normal volatility from an actual problem
Not every rough first week is just the platform learning. Sometimes the campaign really is broken. Before writing off a bad week as "just the learning phase," check the things that are within your control:
Conversion tracking should be confirmed working before you judge any results. If the platform isn't recording form submissions, calls, or purchases correctly, it has nothing to learn from no matter how much you spend or how long you wait.
Budget should be sized to the goal. A campaign trying to generate leads at a target cost that the daily budget can't realistically support will struggle to gather enough data regardless of how patient you are.
The landing page the ad sends people to needs to load properly and make the offer clear. A learning phase can't fix a page that's broken, slow, or confusing once someone clicks through.
If those three things check out and the numbers still look rough after a week or two, that's a reasonable point to make a real change.
What this means for how you budget and plan
Set the evaluation checkpoint before launch, not after the first few days of results. Decide upfront that you'll give a new campaign a defined window, generally somewhere between one and two weeks, before making any real judgment about whether it's working. Resist the urge to check it obsessively and adjust based on a single bad day. And size the budget so the campaign can realistically produce enough conversions in a week to give the platform something to learn from, rather than spreading a small budget so thin that no campaign ever gets enough signal to stabilize.
Key Takeaways
- A rough first week on a new ad campaign is often the platform's data-gathering period, not a sign the campaign is broken.
- Automated bidding strategies on both Google Ads and Meta need enough conversion volume before results stabilize, and a campaign starved of budget may never gather enough data to settle.
- Making significant edits, like changing the bidding strategy or audience, during this window can restart the process instead of fixing it.
- Before blaming the platform, confirm conversion tracking is working, the budget matches the target cost per result, and the landing page functions properly.
- Set an evaluation checkpoint before launch, generally one to two weeks, rather than judging performance from the first few days.
The judgment call is knowing when to hold steady and when to intervene, so ad spend has time to work instead of getting pulled too soon. If you'd rather have someone who manages this daily handle the setup and the judgment calls, Mindstate Strategy's paid ads services cover exactly that. Reach out if you want a second opinion on a campaign that isn't performing the way you expected.
