• Jess Bachman

Behind the Scenes of a High Pressure Social Ads Campaign Told Through the Award Winning Film Gravity

My God, it's full of stats

There are very few rock stars in the social ads space. Placing and optimizing ads just isn’t sexy. But that doesn’t mean it can’t be incredibly dramatic.

If you were to look beyond the creative at all the little gears turning in a hundred different directions in order to pull off a time sensitive client win, you might think differently about skill and care required to turn a 1% CTR into a 2% CTR. After all, that’s doubling the effectiveness of the entire campaign. A 100% improvement.

Well this post is just that, a behind the scenes look into exactly what went down in order to drive thousands of people to a live stream about the federal budget on a Tuesday morning.


Context: Former Microsoft CEO, Steve Ballmer, has a new organization, USAFacts.org, which produces an annual report about how effective our government is in all sorts of areas. If you were a shareholder of the United States, this report, and live presentation would be an earnings call.

The task: Drive a ton of people to watch Steve Ballmer deliver this report, live.

The solution: Settle in, this is going to be a long one…

Mobile Ad Version #0138

Since this is a social media publication, I won’t get into the reasons behind the creative decisions, but just assume a lot of really talented people came up with the creative direction.

We shot some great video of ‘opinion’ balloons tethered to a giant block of granite (foam) and had dozens of copy variations ready to go.

My team was tasked with delivering that creative execution to the right people in the right way so they would tune into the livestream.

Getting people to a website is easy peasy. Let’s say you have $10k in media, and your cost-per-click is $1. Well then you are looking around around 10k clicks to site. But getting people to a website in a 15 minute window on a Tuesday morning is an entirely different challenge. You just can’t spend $10k on Facebook in 15 minutes without all sorts of inefficiencies killing your budget. Believe me, we tried.

The Best Laid Plans of Social Media

So here is the actual plan we drew up and presented to Steve and crew. It was smart, well thought out, and ultimately, quite useless.

A typical strategic flow chart

It was setup in three phases. We would start with video and static ads that would let the audience know about the existence of the USAFacts organization (awareness). We would drive people to USAFacts.org where they would be pixelled for retargeting (consideration). We also used 10 second video views as an indication that they were familiar with USAFacts.

Event response ad

From there, we would serve Event Response ads to the retargeted audience. If you don’t know what those are, they look like this (left). Basically, an easy one-click way to say you are going to an event on Facebook. This event was going to be live streamed to both Facebook Live and a 3rd party platform on USAFacts.org. The people who clicked interested or going were those who committed themselves to checking it out. Once they had opted in, all we had to do from there was remind them of their commitment the day before, the day of, and shortly before the live stream in order to turn that commitment into action. There was an additional step of looking at all the people who said they were interested and building a look-alike audience to send through the funnel. LAL audiences don’t always work better than manually targeted audiences, but we always run tests with them just to have that data and option. How many people could we deliver for the live stream? There are a lot of assumptions in this plan and I really didn’t want to be pinned to a number, but Steve Ballmer’s people were persistent. It felt like 10,000 was a nice big target to aim for. In the realm of possibility, but not easy either. 10k was the name of the report that USAFacts released so our rally cry was “10k for 10k”. That was how everything was supposed to work.

T-12 days

What on earth does this have to do with the award winning film Gravity? Nothing, and yet… everything. This was actually a very dramatic experience for our team and no one does drama better than Alfonso Cuarón. So the movie clips and creative subtitling are here to represent what this campaign felt like at various stages. If you haven’t seen the movie, consider yourself spoiler warned.

I will be playing the part of George Clooney (Digital Strategy Director), and my colleague at NAIL.social, Rachyl, will be Sandra Bullock (Social Ads Manager).

T-10 Days

It only took two days before the first signs of a problem surfaced. We had a massive spreadsheet (linked below) that outlined all of our assumptions about the CTR and shrink rate of each step and were tracking results against it. We certainly knew there would have to be adjustments along the very low CTR’s from the jump took us by surprise and immediately started to gum up the works. A low CTR could be due to all sorts of factors, like poor performing creative, or poor audience targeting, or even the fact that some industries and subjects are just easier to sell through on social. We knew convincing people to watch a presentation about government data was going to be a challenge, but still, the CTR was out of tolerance.

When you are spending $█,███ a day in media with a rapidly approaching event, there isn’t a lot of time to make adjustments and wait and see. We had to make all sorts of optimizations by partitioning the budget into experiments and then making campaign-wide changes based on the results after only 12 hours of test time. Normally, we run tests for at least a week.

T-8 Days

After multiple optimization rounds, we still couldn’t find a definitive cause for the low CTR. Through testing we ruled out creative or targeting issues and there wasn’t much more we could test for with limited time. We set aside the majority of the budget for the awareness campaign in order to fill the coffers of the retargeting audience, but we only had 10% of website clicks we expected for the first few days. Could it have been optimized further? It didn’t matter. There was no time, and we didn’t want to spend $█,███ per day on something that wasn’t working. Time to make changes.

T-7 Days

Moving large sums of money around in Facebook ads manager is pretty easy: it just takes a few clicks. Scuttling the largest phase of our campaign in the first week: not so easy. You just can’t dump money from one bucket to another and expect everything to work like a finely tuned machine. The whole strategy was based on certain formulas and assumptions which rapidly turned out to be wrong. It’s like training for the 100m dash and then showing up on race day and being handed a vaulting pole.

T-6 Days

There was actually several unknowns in our strategy. We could get people to say they were going to the event on Facebook, but how did we notify them that the event was live? There are about a dozen ways this could be achieved, but only one way that actually worked. This information isn’t googlable. Even Facebook support chat has little answers. The only way to truly see if something works is to test it yourself. For this campaign we created multiple fake events where everyone in the office indicated they were going and we tested the best way to reach them on their phones.

For the record, this is the proper process to trigger a push notification for a live stream:

  1. From the Publishing Tools tab on the client’s Facebook page, click Video Library, then click +Live.

  2. From the dropdown on the live window, select “Share in an Event”.

  3. Type in the event name so it populates, and create the post.

  4. Click Go Live.

  5. As the host of the event, grab the URL of the now-live video, and post the URL in the discussion section of the event page.

  6. Users who have clicked going or interested will receive a Facebook push notification with a one click link to the live video. Simple.

T-5 Days

At this point we had the majority of the budget in the the consideration phase running event response ads. We had our targeting dialed in, or so we thought, and were over-spending thanks to the newly appropriated funds from the awareness phase. The problem was that our carefully crafted audiences were not producing the results we wanted. The cost-per-event response was $19, which was way too high and applying that figure to the remaining budget would have left us severely under target.

Remember that look-a-like audience from the original plan? To our surprise, it was performing very well. The LALs rarely outperform our own targeting, but in this case, we were getting beat by the AI. That was fine with us, there is no ego involved, but we had to move the entire budget to the LAL audience to maximize the return. The problem with LAL audiences is that you never really know who they are. It’s hard to get any learnings out of “X worked for Y audience” because the LAL is just a mass of people chosen by FB because they are more likely to perform a certain task.

It’s our theory that the LAL performed so well here because FB knows the type of people who have clicked going to event responses in the past, and prioritizes those. It was an unusually strong signal that just worked for this type of campaign.

T-4 Days

So now we are firing on all cylinders. The LAL audience is performing well on all metrics and the event responses are starting to really heat up. We have built up a large reservoir of budget from the previous re-alignments and are now ready to start spending about $██,███ per day through the home stretch.

Then, on Friday, we see the entire campaign has stopped spending at a big round number of $5,000.00 for the day. We had daily spending limits in place, but weren’t hitting them, and so big round numbers like this shouldn’t happen, unless it was by chance, and “coincidences” are not something you want to see in your dashboard.

We searched every forum we could to figure out this issue and apparently there is a super-secret spending cap of $5k per day that Facebook places on all ad accounts. They don’t tell you about this ahead of time and the only way to get it removed is through a special support person at Facebook. And they leave at 5pm and don’t work on the weekends. And it was 5:25pm on Friday.

T-1 Day

It’s Monday. One day before the main event. We finally got the cap lifted early that morning. The reservoir of budget from previous phases was now even larger due to the spending cap over the weekend. We had a lot of money to spend in a very short period of time. At this point, it may sound like there was a lot of risk involved. What if the spending cap wasn’t lifted? What if the LAL audiences didn’t work?

While we are risk takers, we also mitigated our risks by effectively hedging the campaign on another platform. Twitter was always in the mix for this campaign, it was just reduced in its budget share because it was less effective. But the share of budget that went to Twitter would fluctuate with the effectiveness of the Facebook campaign. When we hit this mysterious spending cap, we didn’t just sit on our hands all weekend. The Twitter budget doubled in size to compensate. We knew this campaign would involve a lot of assumptions and not much time to sort out the details, so we constantly kept the old blue bird in the mix. All campaigns we run have a plan B, but for this one we had C through Z as well.

T-45 Minutes

NAIL Communications is based in Providence, Rhode Island, and the live event was taking place in Seattle. We had a few of the team members make the trip to ensure that everything went according to plan. Maybe you can guess how that turned out.

Our plan was to start the live stream about 30 minutes early. There are two reasons for this. The first is because we wanted to build up and warm an audience for when Steve came on.

The second reason is more technical. We had discovered through testing that we could start a live stream on Facebook, and then grab the URL of that live stream and use it as a Facebook video ad. This means that we could serve a Facebook ad that was actually running the live stream. As far as we knew, no one had ever done this before, and that red LIVE tag on videos in your feed is very compelling, so we were expecting some great metrics out of this tactic. The problem is that placing Facebook ads is not instantaneous. It takes a few minutes to get them algorithmically approved and up and running. We usually see start times at around 5 to 25 minutes from when the ad is placed. So if we started the live stream right when Steve came on, it might take 25 minutes to get the ads out, at which point the live stream was half over. So this is why we needed that extra time.

  1. Start the live stream.

  2. Get the url.

  3. Place the ad.

  4. Cross fingers and wait for the ad to start running.

That was the plan. There was some miscommunication around who was starting the stream and when. Once Steve Ballmer was on set, everyone’s attention was on him, and not on the technicalities of starting the stream. After some very tense minutes in which our account manager Emily heroically volunteered to jump on camera, the stream finally started with some static images and USAFacts community manager, Marthe, warming up the audience.

T+3 Minutes

Due to the previous delays and contingencies we now had about $██,███ to spend in the next hour. Facebook is great for setting up ad campaigns to run over the next month spending a certain amount per day. Facebook is terrible at trying to do something on a very short time frame. It’s just not set up to do that, internally or externally. One trick we use to force Facebook to behave the way we want is by over-clocking the budget limit. For example, if we have $10k to spend in the next hour, then we might put the adset budget at $30k so it takes off faster. Then we carefully monitor it and manually turn it off before it exceeds our real budget. If you try to overlock the adset budget but not the campaign budget it won’t work. You have to completely cut the brakes.

We had tested this before and it worked fine. Of course when it was time to do it for real, it didn’t. The adsets were approved but they weren’t serving any impressions. We had no idea why. Steve Ballmer was about to come on. The killer tactic we had discovered, and pinned the success of the campaign on… was dead in the water.

This scene is a very accurate depiction of what that was like.

T+15 Minutes

If you have ever had to deal with Facebook Ads support you know what this feels like. We had multiple chat windows open and a phone call into our agency rep at Facebook. After several minutes of explaining what the hell we were trying to do, we were also wasting precious time trying out their suggestions that we already knew wouldn’t work. Their folks run off scripts for the most part, and when you are at the bleeding edge of Facebook Ads tactics, you might as well be speaking a different language.

T+31 Minutes

We tried everything to get these adsets moving to no avail. We were close to performing a blood sacrifice to the almighty Zuckerberg when all the ads just started like nothing was wrong. And by start, I mean they exploded.

There is some slight lag between what is actually spent and what is reported in the ads manager but you would only notice when trying to spend $██,███ in an hour. We would refresh the page and would have just spent another $█,███. At this point, with all cylinders firing, the only thing left to do was grab all the money we had put into Twitter (the fall back) only moments ago, and put it back into Facebook and just let it all ride. There was no time to make any more changes or adjustments. Steve was well into his presentation and it was time to enjoy the show.

T+58 Minutes

When you try to force $██,███ through a very small Facebook window all sorts of strange behavior can be observed. It’s like looking at your campaign on a quantum level. Total-spent numbers go up, then they go down. Time and space lose all meaning at this point. We questioned a lot of what we were seeing. The most important metric was watching the total live viewers on the actual live stream climb higher and higher. That’s how we ultimately knew it was working.

T+91 Minutes

So, with all the set-backs, contingencies, contingencies to contingencies, and and other quantum hardships, did we reach the goal of 10k viewers we set many months ago?

You’re goddamn right we did. (Results in the final video below)

This campaign was certainly a challenge on every front and we learned more about the dark corners of the undocumented Facebook behaviors than we had in the previous several clients or campaigns combined. If there are any TLDR takeaways from this whole post it’s this:

  1. Test, test, test, and ultimately only trust the results that you see from your own testing. Building a strategy or tactic around some comment from a blog post or support forum is asking for failure.

  2. Things will go wrong. Have a back up plan. Then have a back up plan for when that goes wrong too.

Try new things. If you wait for a tactic to become a ‘best practice’ then the effectiveness has already been milked out. The only way to find those practices is to be a practitioner and try things that have never been done before. Thanks for getting to the bottom of this indulgence. Be sure to follow our blog for other out-of-this-world posts on social media and social strategy. If you have specific questions, shoot us an email at coffee@nail.social

11 views0 comments