I spent $50 on Twitter Ads so You Don't Have to23 Aug 2017
If you use twitter, you’ve likely seen those tweets with the little *promoted* tag on them. Twitter has always been a huge source of traffic to my tech articles, so I wondered if a promoted tweet or two would be just as successful. To find I went down the twitter advertising rabbit hole for the first time. Keep reading to find out how well my promoted tweets performed.
I publish a tech article once a week. In fact you’re reading one right now! I recently had two very successful posts:
These two articles represent tens of thousands of page views without any kind of external paid promotion. My philosophy towards writing and promoting things is to only do it if I think people will benefit from them. These two posts are some of the best things I’ve written and based on feedback, people in tech found them genuinely useful, so they should be ideal candidates for promoted tweets, right?
We’ll start with my first promoted tweet:
One query can wreck your database performance. Find out how I found this silent killer and sped up my apphttps://t.co/KZ6WHUXWJO— Richard Schneeman (@schneems) July 18, 2017
I got the idea to promote this tweet while I was clicking the “engagements” tab. There’s a little “promote this tweet” button right there. This next screenshot is not the tweet I promoted, but it will give you an idea of the interface:
You see all the stats for your tweet: link clicks, retweets etc. When you click the “promote this tweet” it opens up to a slider to let you pick your spend, and estimates the “impressions” or how many people will see the tweet. And the “engagements” or how many people will click the link, retweet, like, etc. More money means more eyeballs, means more engagements.
Any views generated by retweets or likes are not charged to the Ad campaign, so the idea is that if you promote something really useful, it will have more legs.
How much did I spend? How well did it work?
I mostly care about link clicks because I want people to actually read through the article. My ultimate measure of success is people who subscribe to my mailing list but to get that far you have to click the link.
From my original tweet, I got 21,987 “organic” impressions. Meaning that almost 22K people saw the tweet on their timeline. Of those, I got 317 link clicks. Meaning I had a click through rate of:
317/21987 * 100 # => 1.4 %
While it doesn’t seem great, it is pretty decent. My “referral” numbers for my blog show that I had about 1,328 visitors referred from
t.co which is twitter’s link shortener. I’m guessing that if someone already read my post and then they see my tweet, maybe it encourages them to write their own tweet about it (instead of directly engaging with mine).
When I promoted the campaign, twitter didn’t try to estimate the number of link clicks, instead it estimated the number of total “engagements”. Unfortunately, I didn’t screenshot that page so I don’t have exact numbers. The total engagement percentage though was around 4%, I think which mimicked my organic numbers (pretty sure they just re-use your organic numbers).
I threw down $25 of my own hard earned cash and waited until my numbers started rolling in.
Sidenote: I don’t make money from my blog, you’ll notice there’s not any Ads or anything, I do write about my employer and my side projects, but I’m not directly incentivized by either of them at this time for my writing.
The first thing that happened is my promoted tweet was placed in some kind of a quality check queue. I think they had to manually make sure I wasn’t some phishing campaign before being able to “trust” my account. That took about a day. After that I started getting prompted impressions. They kept coming until my $25 was up.
At the end of the experiment I got 4,236 impressions that resulted in a whopping 16 link clicks. My click through rate was:
16/4236.0 * 100 #=> 0.37 %
Ouch. My promoted tweet did nearly 4x worse than my organic tweet in getting link clicks. I’m paying nearly $1.5 PER PERSON to click that silly link. Granted there were other “engagements” but they were pretty miserable:
- 20 promoted “detail expands” (when someone clicks the tweet but not necessarily the link).
- 1 promoted “like”
- 0 promoted retweets
- 4 promoted profile clicks
- 0 promoted replies
- 0 follows resulting from promotion
Had I gotten the same engagement rate of my organic tweets, I would have expected at least 59 clicks which would have been about $0.42 cost per click. Still not a very financially viable model to drive traffic to a free blog, but hey, it would have been better than my actual rate.
Would I ever “quick promote” a tweet again? Keep in mind this is one of my BEST tweets, and one of my BEST articles. I’m assuming that means this would be one of the best cases of tweet promotion. Based on these numbers: No, I wouldn’t use quick promote on a tweet again.
I had some theories though. Maybe twitter didn’t do a good job of picking out who to show my tweet to. My writing is heavily biased towards Ruby programmers, and maybe they didn’t figure that out and instead dumped my tweet on 4 thousand randos. Maybe, by choosing my own targeting, I could do better.
That’s what I did. Luckily it didn’t take long for another article to get traction (I have no idea when I write them which will do REALLY well or won’t). Here’s the tweet:
What happened when I compared the "slowest" server in Ruby to NGINX? The results surprised even me https://t.co/kGjLc8ZbBV— Richard Schneeman (@schneems) August 1, 2017
Now I did have the benefit of retweets on this one from Matz and DHH, so my organic numbers are way higher. I got 61,329 “impressions” and 1,831 organic link clicks from this tweet. According to twitter analytics, it puts me again around the 4% engagement rate.
I decided another $25 wasn’t going to break the bank, so I manually promoted the tweet instead of a “quick” promote. In this process you build up a “campaign”. I targeted people in the US because that’s where the majority of my organic page views come from. I picked the “computer science” topic. And then there’s a section where you can select other twitter accounts. The idea is that your content would be a good fit for followers of those accounts. I picked a handful of my favorite Rubyists with high follower numbers and that was that. I scheduled the campaign to go out.
This time, there was no waiting validation period, though there was a strange billing issue where they couldn’t bill my credit card, but could bill the exact same card if I added it again. I opened a support ticket that got forwarded to twitter devs and never heard from them, so who knows.
What were the grand results?
Previously my $25 bought me about 4,236 impressions. With this targeted tweet it bought me 2,700 impressions. Of those, 44 people clicked on my link, bringing my click through rate to:
44/2700.0 * 100 #=> 1.6 %
This is much closer to what I had expected to see on the “quick promote” tweet in terms of ratio. On the surface of things it looks like going the extra “mile” (i.e. taking about 5 minutes longer to build a campaign) gave me about 4x the campaign performance improvement. If I was going to advertise on twitter I would definitely go this route.
Also worth noting I got some other engagements:
- 8 promoted likes
- 1 promoted retweet
- No net new promoted followers
These are better than the “quick” option, still it’s a little disappointing that when you target people your money doesn’t go as far. For the same dollar amount, I got 1,536 fewer impressions. This is 36% fewer impressions for the same cost. On one hand, my dollar per link click was WAY better, so this was a better per dollar deal.
On the other hand, as a user of twitter, it looks like there’s not much of an incentive for them to show me good Ads. Basically, they are just as happy showing me an Ad not related to my interests at all, as they are something I really care about. They’re only interested in targeting to my interests when it’s something the advertiser can pay for.
I understand the economics of Ad buys, and that advertisers are supposed to spend more money for more specific criteria, but it also seems that if you HAVE to show an Ad to 1,500 Ruby programmers maybe it makes more sense to show them something they care about than a random energy drink Ad (or whatever). Are there really that many companies explicitly targeting the same group of Rubists to the point of targeted Ad saturation? Obviously I don’t work at twitter, so I can’t answer these things.
Would I advertise on twitter again? Maybe if I had an actual service to sell. For example, if I were writing a book and stood to actually recoup some of my costs.
Assuming 1% of people who clicked through bought a hypothetical book. I would need 100 clicks per book sale and at the current rate of $0.46 per click it would cost me $42 dollars. Assuming I kept $50 from each book sale, my net profit from advertising would be $8. Which isn’t a loss. I also don’t know how the quality and cost for click throughs compares to other advertising sources.
Twitter also offers different kinds of advertising that I haven’t looked into, for the hypothetical book example, they have a “website clicks or conversions” Ad type. Someone will have to write about that because my wallet is feeling light.
That’s it to my great twitter Ad buy experiment. If you liked this article please sign up for my mailing list. Also feel free to “organically” promote this on twitter, cause I’m all out of money 😜.