Goodbye, Facebook and your useless results. [UPDATED]
Over the last couple of years, we've been testing Facebook Ads and AdWords to see which one works better for bringing in new traffic to UNA.IO. Here is what we've learned.
With Facebook, you get incredible "results" that don't matter.
The Facebook test spanned about a year and a half, with now just over a million users "reached" and overall spending of over $4,500 (USD).
Each campaign and ad had a different combination of settings - target audience, format, language style and goal (clicks, engagements, likes, etc.). We also made sure to try different landing pages, such as the UNA Facebook Page, articles at UNA.IO, articles at Boonex.com related to UNA and even articles at Andrewboon.com, which would promote UNA in some way.
In all cases, Audience was selected with interest in social networking software, open source CMSs or some associated topic.
In most cases, Audience was limited to English-speaking countries, such as US, UK, Canada and Australia. Only a few ads also included India, Singapore, Russia and some Europian countries like Germany, Italy, France, etc.
In total we received:
- Almost 20,000 page "likes" for UNA Page at Facebook
- About 15,000 "likes" to various boosted posts.
- About 5,000 mixed "engagements" (likes, clicks, shares, etc.)
- Over 18,000 clicks.
Looks pretty good for a $4500 budget, right? If you think so, you are exactly the kind of customer Facebook strives on.
When we examine the real results, using Google Analytics, here's what we see:
- Total of 1851 sessions generated for UNA.IO.
- Less than 2 pages per session (while average pages per session excluding Facebook is under 5.34).
- Over 66% bounced (while average bounce rate excluding Facebook is under 39% for UNA.IO).
We also had just over 4000 sessions for Boonex.com, with largely similar KPIs.
So, first, no matter how you slice it, at least 10,000 of those clicks reported by Facebook are nowhere to be found. When this question is raised with Facebook Ads support, they routines reply that the discrepancy is due to different traffic counting methods used by Google Analytics and Facebook. Take it as you will, but I have never seen any visitors observed by any other traffic tracker (we used Intercom, KISSmetrics and Mixpanel) that wasn't counted by Google Analytics. I call bullshit on this one.
Anyway, let's talk about ROI. After all, even if there is some -lie- discrepancy involved, it's no drama as long as we profit from the whole thing.
In our analytics, we track revenue, account registrations and downloads. With Facebook ads, in total we received...
New Accounts: 0
Ta-da! A whole lot of "Facebook results" and no results.
Perhaps the worst part is that all those ads required quite a bit of time to set up, monitor and adjust. With 41 campaigns in total, even if each one is taking half-an-hour, that's 20 hours out the window.
Another interesting observation about Facebook "reactions" is that many of them are completely irrelevant. We received comments like "Nice selfie!" or "You look cute." for posts about software that had no image of a person or anything remotely related to comments like that. As if Facebook just randomly thrown in some comments from elsewhere to make the numbers look better.
Google is expensive, but it works.
In contrast, our latest campaign with Google AdWords has a total spent of A$640 (about $510 USD), which generated 604 clicks and 62 conversions (signups and sales). That translates into a conservative estimate of about A$778 return.
These are the results of a very simple display ad pointing to the homepage. It can be further optimised by A/B testing different landing pages, target audiences, keywords and ad types.
In other words, even in most simple form Adwords "results" translated into profit for us.
We will use and observe Google Adwords for a few more months and will share more details on overall performance, tips, tricks and mistakes to avoid.
Clearly, there is no reason for us to continue using Facebook Ads in any shape or form. Our Facebook page continues collecting 10-30 likes every day somehow, which also never bring any meaningful results anyway, so we should ignore that as well.
Besides spending money on Facebook "boosts", we should stop spending time on Facebook page, too. This post will be the last one to be shared on our Facebook page. The page will stay active, but not supported. We keep it only to prevent fake UNA pages from presenting like out official resource (which happened before with Boonex), but we will not support it in any way. It's just a waste of time.
As for Google Adwords, we will continue experimenting and will share more details in coming months.
Why such a difference?
Google is generally used to find things. You usually already have something that you need in mind before using Google, and it just helps you to find it. Therefore, if your product is targeting an established market and fits certain predictable search query, Google is the best way to market it.
Facebook is used to kill time. Much like in the days of television, ads are unsolicited and mostly perceived as annoyance. Facebook Ads team figured, however, that if the ad looks like like a post and if it has some cool image attached, then millions of people that indiscriminately click/share/like everything that catches eye, may actually engage with the ad. They won't buy anything, but they will click something. That click can be sold to a marketer that doesn't have to watch the company bottom line.
Well, maybe (an this is a wild guess), just like with TV ads, if you have a huge budget and your product doesn't fit existing market, then you may successfully use Facebook to create awareness about your offering, especially if it's POP enough to catch a viral sharing wave. For example, you invented drone-socks that fold and move themselves to a drawer. Nobody is going to search for that, so you'd have to run a Facebook campaign. For everything else, just stick to Google.
These are answers and clarification following comments we received so far:
The only way a FaceSucks advert could have worked is if it was targeting web developers on FaceSucks
And we did, as much as we could. When it comes to targeting developers, Facebook won't let you actually target those who develop Facebook apps, or have registered for API keys, for example. You can only target by interests, which we always did, selecting interests, and here's what we had:
Didn't help much and actually I never found any evidence that any of the people who "liked" our posts actually had interest in those categories. Their feeds are usually littered with a random mix shares.
Have you done any kind of trial with Instagram or Pinterest?
Facebook will automatically place your ads on Instagram if you want. And we did try it. For example one of the campaigns had this:
Not a whole lot, but still nearly 2000 clicks from Instagram just for one campaign. Oddly enough, Google Analytics hasn't registered 194 clicks with average visit duration of 28 seconds, which is even worse than Facebook proper (52 seconds). Go figure. Maybe Instagram users have an extremely short attention span?
You said you spent $4,500 on a year and a half campaign, which amounted to over $100 a day. Those are the numbers that don't add up.
I mentioned that we raised an issue with Facebook during a period when the spend was sitting at $100/day. That's when we had 3 boosts running together, each for 3 or 5 days. Facebook boosts are usually short-term. So, our spending of $4500 over a year and a half is because we only came back to Facebook Ads when we had something to promote, which could potentially interest a typical Facebook user. We tried a total of 41 different campaigns so far, including Page promotion, post boosts, direct website promotion, boosts optimised for clicks and boosts optimised for reactions.
My whole point to posting this comment is simply to illustrate that if facebook didn't work for Andrew Boone, fine it didn't work. But that does NOT mean that it was facebook's fault that it didn't work, OR that it won't work for anyone else. The most successful business people in the world KNOW that when something doesn't work, the direction to look in to find the problem is inward. Outside forces almost NEVER have anything at all to do with failure, ever.
My "by default" sentiment is always to look for what I have done wrong and what I could do better. This is the ONLY reason why we've spent so much money and so much timing try to figure out Facebook Ads. As I said, we tried different approaches over different time periods, with different kinds of ads, targeting, locations and placement. Still, I acknowledge that someone else could have done a better job, but considering that we do have reasonably good understanding of web marketing, social networks and web software; plus having tried many different advertising and analytics platforms I am confident that if we failed to benefit from Facebook Ads, then 9 out of 10 people would fail as well.
You see, we didn't do any special "hacks" and our product is proven. We have a long-term record of conversions and know how each visitor performs on average. Using your "tools" analogy, we have tried many different tools, and already know which tools work. It's a matter of simple comparison. Just look at this:
These are referrals from social networks, all to UNA.IO - same site, same product. First, note how Facebook and Instagram have average session duration of under 1 minute, while all others have 4 minutes or more (you can disregard Hackernews, since it was a quick experiment and most clicks are our own).
Now, look at Google organic and paid results:
We only started CPC campaign for UNA recently and had 670 sessions, which resulted in 70 sign-ups so far:
So, more than 10% conversion rate. For whatever reason, when people from Facebook come to exact same page, conversion is ground zero. Oh, we had ONE conversion from Facebook, which happened yesterday! Perhaps someone is trying to prove me wrong. :)
Moreover, I am showing the numbers from those that could actually be registered by Analytics. We tried KissMetrics and MixPanel to see if maybe Google Analytics is making Facebook look bad, but the results are about the same. Facebook, on its end, reports results in order of magnititude better than we see on-site. Here is just one campaign, optimised for clicks, with simple text+image ad-set, linked directly to homepage (as close to AdWords as it gets):
As you can see, we've spent just over $400 on that one. Got over 9000 clicks. We know that our average revenue conversions sit between 50c to $2 per new user at 30day tracking, depending on source quality. So, 9000 clicks should make a good difference, right? Let's see what we got in reality:
So, instead of 9000, we actually had only 854 sessions for this campaign. 747 users. That's 53c per user acquisition cost already (Adwords is about $1). Next, session duration is 42 seconds, while Google CPC is over 3 minutes. And not a single sign up.
All in all, I don't think that we should entirely blame it on our own mistakes. Back in early 2000s I made a huge mistake and bought into some "$100 for million visits" email marketing scam for one of our sites, which performed very similarly. Huge numbers reported by their system, 10 times less reported on our end, no sales, skewed graphs and wasted money.
If you tried to change brake pads with a tool like Facebook Ads, you'd find that it attaches to the nuts just fine, rotates them, shows a fancy gauge saying "pads are now changed 350 'mega-times'", but as soon as you take it off you see that nothing happened. Then you pay for it. :)
#facebook #adwords #marketing #advertizing