HN Explorer

How to burn $20 on Twitter with $10 real CPC

61 points by hayksaakian ↗ HN
I heard twitter advertising was pretty good, so I decided to give it a shot.

Why I try new advertising channels, I typically spend $20 to be fair.

I also run a personal link shorten-er, and decided to use that to track my CTR.

In about 30 minutes 20$ bought me 31 engagements (all clicks in this case) on a promoted tweet.

At ~$0.65 ecpe, this sounds like an awesome deal. Obviously the quality of the traffic won't be as good as search traffic.

However, I made sure my ad was at least geographically targeted.

Reality:

Of the 31 reported engagements, my link shorten-er showed 10 clicks.

2 were from self proclaimed bots (Mozilla/5.0 (compatible; TweetmemeBot/3.0; +http://tweetmeme.com/))

6 others were also from bots, they had these user agents:

Ruby

Kimengi/nineconnections.com

MetaURI API/2.0 +metauri.com

JS-Kit URL Resolver, http://js-kit.com/

(twice) UnwindFetchor/1.0 (+http://www.gnip.com/)

At the end of the day, only two other clicks came from realistic useragent/ips that could be construed as a real person.

Real Cost:

$10 CPC

conclusion:

Totally not worth it. Twitter needs to get their shit straight and resolve their blatant click fraud OR provide the real data behind those engagements (useragents, precise IP based geolocation, referrers.)

17 comments

[ 3.4 ms ] story [ 44.7 ms ] thread
gue5t
if you want you can paypal me 50 cents and I will check out literally any link for you. cut out the middle man, keep disrupting and stuff.
ruswick
I'll do it for 25 cents.
yarou
I'll do it for free. In fact, I will pay YOU for the privilege of helping you out.
brianbreslin
when their self serve platform first came out, i did a few promoted tweets, but the fact you can't pick a specific tweet to promote left me promoting things that weren't time-relevant anymore. Also my CPC ended up being like $4, and the "followers" I gained were all bots on the experiment. Facebook definitely has their act together better on this front.
alex_doom
Have you tried promoted posts on Facebook? I've been curious how well that works. I'm skeptical of it's effectiveness.
zacmartin
On large pages I've had massive issues with Promoted Posts - flooded with spammy comments that suggested all additional impressions were fake.

Facebook advised Promoted Posts were more effective on smaller pages (eg. <1,000 fans).

hayksaakian
I've tried Facebook ads proper: decent CTRs and good targeting, but I had a poor landing page for facebook users so eventually underwhelming conversions.
the_watcher
Facebook unpublished page posts have been really successful for me once you get the targeting down. I've had success with lookalike audiences, as well as with targeting one precise interest + one partner category.
Dystopian
I had to mention this just because we're talking about ads etc...

Facebook iOS install ads - ridiculously good conversions on installs, and then active users. Better than any other ad platform that I've tested with.

mansigandhi
Yes, but also ridiculously expensive CPI!
photorized
One could speculate that if you increase the spend, and traffic - the percentage of bots would actually be much lower.
hayksaakian
I agree completely. But it still doesn't account for all the reported clicks/engagement that never hit my servers.
Ologn
I tested Twitter ads for the first time recently.

I liked the demographic information they gave me - where the engagements came from geographically, as well as information on gender and interests. We can use that information with regards to Twitter, as well as independent of Twitter.

I know our campaign (probably) hit real people because we got a reply and five re-tweets. Our campaigns totalled to $225.

SpenserJ
I feel like your return on investment is sorely lacking. A good app or post, and you can easily pull off the same results for free.
gojomo
Twitter and downstream services at times both add and strip URL redirects. (That's probably what the 'unwind' gnip process is doing: following redirects to find the real destination.)

Are you sure all clicks from all clients would have visited your shortener, as opposed to skipping it using metadata about its final destination?

(It might be worth testing with a unique, 200-returning terminal URL.)

hayksaakian
You have a good point. Let me check the actual server logs.

Shoot; the destination is a static site served with a rack app and Google Analytics does not have data yet (too old for real time, too new for normal). I'll have to check back later.

dgunn
I would say you have it right here. Link shorteners really aren't a great way to track visitors to a link. You really need something like Google analytics or count how often your server actually served a resource.