Do I let it run for a certain number of days? Or perhaps wait until it gets a certain number of clicks and impressions?

How do I know when to turn off an ad or create a similar ad?

What’s the simple answer for those who are not math geeks?

Simple Answer

Let each ad generate 10 clicks or $10 in spend, whichever comes first. This gives you enough data to tell if the CTR or the CPC is not where it needs to be.

If you can’t get a 1% CTR in the newsfeed, something is wrong with either your targeting or your content.

If you have multiple ads in a campaign, sort them by how you’ve spent over the last 7 days, high to low. Pause (not delete) the low performers, such that you keep 3 ads live at any one time.

So each time you optimize (every couple days), you create a few new ads and kill a few ads. Takes 5 minutes tops.

Sophisticated Answer

If you have your ads set up properly, with the right counts and structure, as defined here, you’ve got your Facebook funnel of Audience, Engagement, and Conversion set up to nurture your customers.

At 3 campaigns of 8 ads each, you’ve got 24 ads to watch– each with a different combination of content and targeting, along with varying business objectives and audience counts.

Within Audience, you’re looking to gain new fans and reach people who don’t know about you yet. You’re comparing your audience ads against only the other audience ads, since the engagement and conversion campaigns have other objectives.

The CTR on your conversion campaigns will be much lower than your engagement campaigns, but the revenue per click is higher for conversion. Your CPF (cost per fan) is lower for your Audience campaign, since you’re explicitly asking Facebook to generate new fans.

The early indicator on ad performance is CTR, since it tells you if there is a mismatch between your content and audience. I like to see a 1% CTR in the newsfeed, though you might be able to get 5-10% if you are in a “social” vertical (entertainment, food, babies, pets, etc). If you’re running RHS (Right Hand Side placements), your CTR will be much lower, but likely generate the same CPC, since the decrease in CTR is offset by the decrease in price for oCPM bidding.

Above you can see 4 ads, of which one ad is getting most of the traffic. The second ad is generating new fans at 20 cents each, while the first ad is at 32 cents. However, the second ad is not necessarily better, as it’s spent only $1.59. Be careful not to declare a winner too early.

We like to see 10 clicks or $10 in spend before making a decision to pause an ad. Anything less and you’re looking at an issue of statistical significance– results that are noise from not enough data. For example, if you got one click on only 2 impressions, that’s a 50% CTR. But it’s noise.

Ads that have a decent CTR can stay alive for the second level of ad pruning– a decent CPC and engagement rate.

Usually CTR and CPC have an inverse relationship. When you bid on oCPM (which you should always do, except in rare exception), Facebook is going to allocate inventory for you automatically based on your chosen business objective. They select who within your target is most likely to engage/convert and they determine the right price to pay for you.

In an engagement campaign, you’re looking for a strong CTR and a low cost per action. Look at the ratio of your total actions vs your total clicks. If that’s high, you know that only is your content interesting enough to generate interest, but that people will also engage.

In the above example, the CTR on the second ad is a whopping 6.7%, while the first ad is still an excellent 3.6%. We have enough data so far to say we should keep both ads alive for now.

Note that the ad that spent more ($8.91 for 42 action) is hitting a larger audience (4,600 people) than the smaller ad (less than 1,000 people). Facebook used to tell us a precise number, so less than 1,000 can be anywhere from 1 to 999 people. Use the data to tell.

The second ad, even with a higher CTR (which normally means a lower CPC), does cost more per click and per action because the price we pay for the inventory is higher– a $10.26 vs $6.95 CPM. Don’t worry about the effective CPM being higher for some ads. Facebook is choosing what members of your audience is most likely to convert, and that can often skew towards more educated or older, which will cost more.

In your conversion campaign, you’re looking at 3 levels of progressive culling of your ads– a decent CTR, a strong engagement rate, and a good CPA (Cost Per Acquisition). Of course, the cost per conversion (also called CPA) is most critical, but the early indicators of CTR and engagement will help you shut off ads before they spend too much.

I like to see 10 conversions per ad before I can make a proper decision on if an ad is working well.

In the example above, we’ve sorted this e-commerce campaign by spend. As is often the case, one or two ads take most of the volume. So we’d focus on just the top 3 ads, which have spent more than $10 in the last week.

Notice that the second ad has a $4.34 cost per conversion, which is a shopping cart checkout. If we have our ROAS calculations set properly (a topic for a separate blog post), we know we can afford a $10 CPA. So this is strong.

The first ad, though it has an unprofitable CPA of $88.85 does have a higher CTR than the other two ads, as well as a lower CPC than the other two ads, as well.

In most cases, especially if this were Google AdWords, we’d cut this ad for being unprofitable. However, this ad might be profitable with some landing page optimization and it may contribute as an “assist” by driving more people into Google searches for our name.

This is a tricky example, since Facebook does drive more Google traffic, enhance the performance of email campaigns, and boost other marketing efforts, in general.

This ad, though in the conversion campaign, might actually belong better in an engagement campaign. Most of the actions are photo views. So it tells us that the creative is strong, but that we might need to trim down the targeting, change bid types, and tweak the landing page.