Our Facebook Ads Business Experiment is almost done, and we’ve come a long way. It’s time to share some of the lessons we’ve learned, and the results of our work.

We’re breaking down our final post into two parts. We’ll share the full results of our experiment in Part 2, but first, we’re looking back at what we learned through our process.

For those who missed the previous posts, here they are:

Our goal was to start a business and use Facebook Ads to find an audience and sell our products. We started our business, King’s County Threads, created a line of shirts and ran multiple ad campaigns.

We’ve talked about the results of our first two campaigns. But looking back on our process as a whole, we realized we made some mistakes that we hadn’t talked about.

These mistakes cost us time and money. But they also taught us some important lessons. And with a process in place, we could quickly learn from our mistakes and make improvements.

Since you’re following along, you can learn from us and avoid them in the first place.

Here are four things we wish we’d done differently, and how the lessons we learned helped us run better campaigns.

Lesson #1: We Were Too Hands-Off

Once you choose the duration for your Facebook Ads campaigns, they should run that whole time. We wanted our campaigns to look similar, so we could measure the impact of changing one or two variables each time.

There are some problems that can prevent you from getting useful information, let alone achieve your goal. We ignored some big red flags in our first two campaigns. We should have stopped them, fixed the issues, and tried again.

We’ve talked about the specifics of our first two “L Train” campaigns at length, but the major issues were:

• In our first campaign, Facebook wasn’t placing our ads where we wanted. They were only delivering to apps on Facebook’s Mobile Network, but we wanted to measure results across different placements.
• In our second campaign, we had a high cost per click (CPC) and low relevance score. That suggested we weren’t reaching our right audience.

Two days into our second campaign we were paying $1.46 for every click on our ad. That was way too high. In comparison, the average CPC for clothing ads is about $0.08.

Our process called for post-campaign analysis, but we were seeing big problems from the start. It was clear that we shouldn’t just look for issues at the end of our campaign but throughout the process.

What We Learned

Monitor your campaigns for major problems while they’re running. Our process won’t work if all we learn from a campaign is “well, let’s not do that again.”

Regular vigilance will save you money, and it doesn’t take up a lot of time. The AdEspresso dashboard has key metrics on how your campaigns are performing so you can quickly see if something’s wrong. Catch major problems early, and you’ll save money to put toward better campaigns.

Lesson #2: Our Timing Was Off

One mistake had nothing to do with the design of our Facebook Ads campaigns, but with their timing. Our first shirt was aimed at people concerned about the L Train shutdown in New York. It’s a big deal, but it’s also not clear when it will actually happen.

The first news about the L Train shutdown was in January. Google trends revealed a lot of interest that month, but it dropped off by the time we ran our experiment, only spiking around major news reports.

We weren’t running our six-week experiment in January. But we could have done more to time our campaigns around major news about the L train shutdown.

We ran our first campaign at the end of March. That happened to be a current low point for L train related news.

Our second campaign ran at a time of higher interest in the L train. But with more forethought, we could have tracked major news stories and timed our campaigns for maximum effect.

What We Learned

Time your campaigns for maximum impact. Try to plan your campaigns around relevant events that will naturally generate buzz. If something unexpected comes up, be prepared to move quickly.

You can also sequence campaigns so they build on each other and become more effective than any one campaign alone.

In 2015, British retailer Shop Direct used this strategy to get 20x returns on their ads on Black Friday. Their first campaign used videos to get peoples attention. A second campaign targeted people who clicked on the videos to drive them to the website. Finally, on Black Friday itself they showed people who’d been to the website ads for sales on specific products they’d been browsing.

Going forward, we’d try to incorporate timing into our strategy. That’s why we created a second shirt to take advantage of the buzz around the Presidential race, the biggest, most sustained news story of the year.

Lesson #3: Our Buyer Personas Needed Work

Facebook lets you find people with insane specificity. But you have to know who you’re looking for.

We’re from New York City, ride the L train, and go out in Brooklyn and Manhattan. We knew what the target audience for our shirt looked like. But we didn’t think about how that demographic would behave on Facebook.

Looking at our first two campaigns, we realized there were problems with our geographic targeting and our interest targeting.

Our first campaign targeted every zip code adjacent to the L train in Manhattan and Brooklyn. But there are hundreds of thousands of people living along that line with a lot of demographic differences. For instance, Manhattan and North Brooklyn are much wealthier, on average, than neighborhoods further into Brooklyn.

Disposable income is a big factor in t-shirt purchases. We should have narrowed our focus to areas where people would be most likely to buy a t-shirt. However, we learned quickly and adjusted for our second campaign, which focused only on Manhattan and North Brooklyn neighborhoods.

We also should have done more research into Facebook interests for our audience before our first campaign.

Our ideal buyer persona was young, rode the L train, and would have a sense of humor that allowed them to convey their transportation-related distress with a t-shirt.

We targeted people who liked popular internet humor sites like Lolcats and College Humor because those sites seemed like a convenient shorthand for a sense of humor. In our second campaign, we targeted people using the same sites, but also tried to incorporate the idea of people who were politically or socially active in New York City.

As we discussed in our last article, this targeting didn’t work well either. We got one sale, but also a low relevance score and generally poor engagement. That translates into a lot of money spent for very few clicks.

What We Learned

You can find the right audience for almost any product on Facebook. There are so many granular, specific interests you can pick from to craft the perfect buyer persona.

But it’s also possible to get overwhelmed with choice and make poor assumptions about your audience interests. We needed to really question our assumptions about our audience interests based on the information we gathered.

We needed to see whether people with these interests were engaging with our ad. If not, we needed to choose new interests to see whether they’d have a bigger impact.

Custom and lookalike audiences are so important because they take the guesswork out of finding an audience. They let you leverage user behavior off Facebook, like when they visit your website or sign up for your newsletter, and turn that information into a Facebook audience of people with similar interests. If you advertise to the sort of people who’ve shown interest in your product, you’re already at a major advantage.

Lesson #4: We Didn’t let Our Products Sell Themselves

“Keep Calm and Don’t Shut Down the L Train” was the first t-shirt we made, but it wasn’t the only one. We created several other t-shirts on our website aimed at our target audience.

We wanted to do a series of small, quick tests to iterate and improve quickly. But we didn’t take advantage of Facebook’s multi-product ads to help us compare results for lots of different products.

We were thinking too narrowly. Although we ran A/B tests, we could have used multiple image ads to test out even more products, potentially increasing sales, and in any event learn more about our audience. We’d come up with a wide range of products, and we should have done more to see whether any of them could really take off in our earliest experiments.

What We Learned

Facebook multi-product ads are perfect for retail brands. More options mean you can potentially reach a larger audience than any one ad alone, and drive more people to your website.

They also allow you to A/B test even more language and images, and to see whether different combinations of products result in more clicks.

We focused on single product ads to keep things simple and consistent across our campaigns. But with a range of different products to choose from, we could have run multiple, parallel campaigns.

Some could feature carousel ads, and some could feature single product ads. It would cost more money, but the more information you have early on, the more you can learn.

Finding a Better audience with the “Bernie Bro” T-Shirt

The whole point of our process was to recognize that mistakes happen, and campaigns take time to bear fruit. Rather than kicking ourselves, we applied these lessons to our next campaigns.

These were built around the “Bernie Bro” t-shirt. In a heated election year, we thought there were people who liked political revolution as much as they liked Of A Revolution. But we couldn’t be sure unless we made sure not to repeat mistakes we already made.
How We Addressed Mistake #1

After two days of our Bernie Bro ads, we noticed that we’d only gotten 4 impressions We stopped our first campaign iteration early, tweaked our campaign, and ran it again.

How We Addressed Mistake #2

When we advertised our “L Train” t-shirt, we didn’t think about timing. But our Bernie shirt was timed to really gain traction. We started advertising during a period when Bernie Sanders won multiple state primaries in a row and was in the news on a daily basis.

How We Addressed Mistake #3

In our first L train campaign, we used proximity to the L train as a proxy to find our right audience. But that doesn’t account for the huge variety in the hundreds of thousands of people who live across Manhattan and Brooklyn.

There were two reasons we were confident about our “Bernie Bro” targeting. First, there was strong evidence that Bernie Sanders was popular with college students, and we knew there was a big market for Bernie swag.

Second, college campuses have a lot less demographic diversity than New York neighborhoods so we thought our ads had a better chance of being shown to the right people.

We were looking specifically for people who were in fraternities and we could target schools where Greek life was a major part of the social scene.

How We (Will) Address Mistake #4

We didn’t run a carousel ad around our Bernie shirt. We felt very confident in our demographic targeting around the presidential race, and wanted to get right down to it and see how our Bernie Bro shirt would perform.

Single ads or multiple-image ads are different weapons in your arsenal. But if we have success with the Bernie Bro shirt, we’ll use carousel ads to try and build on that momentum.

How Did We Do So Far?

We haven’t finished running our latest Facebook Ads Campaign yet, but we’ve seen signs that we’ve learned from some of our mistakes.

We haven’t sold a shirt yet, but we’ve found an audience that is engaging with our ad:

• The ad has 21 likes from people at universities we targeted
• People are commenting on the ad, which is a much more active type of engagement than liking.
• We have one share. If we can increase social sharing even more, it’s basically free advertising.

Our overall metrics are also much better, validating these social numbers. We have a relatively low CPC of $0.64, and a click-through rate of 1.66%.

Since our ad is engaging people, we’re going to keep it running and go through our process to see what can be improved.

But we’re not sitting still, either. We’re inviting people who liked our ad to like the King’s County Threads page too. That way we can create more organic buzz around the rest of our products.

The Learning Process

When you’re just starting out with Facebook Ads, you might make mistakes. Campaigns will take some time to gain traction, and unexpected bumps in the road might mean you’ll simply need to go back to the drawing board.

However, if you stick to your process, you’re going to learn from your mistakes a lot faster, instead of aimlessly trying different quick fix tactics.

In the time it took to run 3 campaigns, we went from barely any engagement to an audience that was regularly liking and sharing our ads. Considering that our company didn’t exist a few weeks ago, and that these people had never heard of it before, that’s not too shabby.

But we’re not resting easy, because we have a goal: sell our t-shirts. In our last installment, we’ll wrap up the rest of what we’ve learned, and give you the run down of how we ran our final Facebook ads campaigns.