Only getting conversions after ads first approved, any idea why?

I have made a very weird observations of my adwords PPC activity, and I'm hoping some of you experts can shed some light on it.

When I first made an ad for adwords, I thought I was in heaven. At an average of about $2 a click, I got about 10 clicks my first day, and 2 of them became conversions (form contacts soliciting me doing work for them.) The next day about as many hits...no conversions. The next 2 weeks I actually upped my cost per day to about double that each day...and zero conversions for that whole 2 weeks.

So I thought day 1 must have been a fluke, so I made a new landing page, edited my ad to point to the new URL, and once it was approved that very day I got about 10 clicks and 3 conversions! The next day about as many hits...1 conversion. Then another 2 weeks where for some of it I left it the same expenditure, and some of it spent even more per day...no more conversions.

At this point I had a theory: so I modified and tiny aspect of the ad (inconsequential change really) and BAM, again conversions the day that ad was approved! Then none after that.

Does anyone have any clue why newly approved ads may be converting amazingly well for me (20%-30% every time) yet a few days after approval I can pay for hundreds of hits without a conversion? What is going on?

 

Ten clicks is too small a sample. Look at a longer period of time, a hundred clicks or even a few hundred. Then you'll get a better idea of how well your page really converts in the long run. So, what you experienced is a fluke. There's also other data to look at such as your positioning the first day (I'll bet it was higher that first day than other days). But there's not much you can tell with so little data, it's statistically insignificant as they say, plus you also changed ads and most important, the page. Everything is diluted and you can't really make any conclusion or forecast.

 

Maybe I wasn't clear enough, but when I said 10 clicks I was referring to the first day the first time I posted the ad. I then proceeded to talk about roughly 6 weeks worth of time and many many clicks where the exact same pattern repeated itself 3 different times. The pattern exists, the reason for it is what I'm curious about.

Basically in a more abridged form: there's been a few hundred clicks over the course of about a 6 week period, and 10 conversions. Every single one of those conversions coming on the day of one of 3 different ad/ad change approvals, or the day after if it was a later in the day approval. There were changes in page and ad...but not in the pattern, it's repeating regardless. The first time it could be coincidence, the second enormous coincidence, the 3rd...starting to not look like coincidence.

The only possibility I can think of is the google's performance optimization option of "Focus on Clicks". That possibly after each approval it shows ads without any optimization in place (because no data yet), and as it gets data about clicks it starts optimizing the campaign toward getting the most clicks per dollar spent in my daily budget, and that optimization may have have the ironic effect that the changes that lead to more clicks per dollars spent tend to be clicks that won't convert for my niche. But I don't have quite an intimate enough knowledge of the inner adwords system to know for sure if that's a possibility.

 

I understood what you meant, although you are clearer now and providing more information but not quite enough. What about that positioning? I still say it's coincidence. You are still talking about only a handful of clicks per day on average.

As for your theory, doesn't hold water to me. The simple fact is, even without data, the system makes a pretty good guess as to your ad quality. It's not 100% perfect but I'd say it's 95%. I believe this "optimization" you speak of is quality score. It's not the system that optimizes, it's you. The system is designed to pretty much take care of itself.

True that it does need and works best once it has more data to work with. And to do so it may bend the rules a bit at times. It could show your ads in a higher position just to get data and refine your quality score calculation. That's why a more thorough analysis is needed. Strange it would do so every time but it could happen. It has happened to me many times: a sale in the first few clicks but none in many dozens of clicks after that.

If you are so sure it will happen every time, create another ad. Make it a crappy one and see what happens. Another test would be to unpause one of your previous ads. This test will be invalid if you deleted those ads and have to recreate them. You've already invalidated the test by changing the landing page as well as the ad as I understand it (you created a different ad with the different landing page or did you only change the destination URL?).

 

Ten clicks is too small a sample.

 

Yes, where 1 sale falls within a stream of many clicks can be easily chalked up to coincidence. But 3 or 4 in the first 10 or so clicks, then hundreds of clicks without any...and this exact pattern repeating multiple times. Quite different.

No...I validated it by changing the landing page and ad. By changing those variables and observing zero change in the observable pattern, that is evidence that the pattern is not being created by the landing page or the ad, but rather another factor. Ideas for hypotheses of what that other factor may be (so that I can then set up experiments to test those possibilities as variables to either rule them out or confirm them) are what I'm seeking.

Thanks! That's the kind of hypothesis I'm looking for. With this type of hypothesis I can set up a test with a variable and see if this is the effect.

 

Not sure if you're trying to be funny or if you can't read...

 

When you want to split test, you need at least 30 CONVERSION to make a decision.

 

Not exactly exact, but communicates my message clearly enough.

 

What are the values of the X and Y axis? Is each data point per day? If each grid is one click (as opposed to 5 or 10 or more), then this is just coincidence. You also are not providing positioning information. I'm assuming you are not changing anything else such as bids.

Like I said, coincidence, or something else we don't know yet, something that graph doesn't tell us. I'll agree, it's a strange coincidence. But that's what coincidences are and there's no explanation, it just happens. Like two US presidents dying on the same day which happens to be the country's independence day. Like a Canadian prime minister having two sons born on Christmas day. Like my son and daughter born on the 21st of March and May. No divine intervention, just improbable (but not impossible) events. There's so many Adwords advertisers, I'm sure you can find many such events or if you look hard enough, other types of patterns.

By the way, you did not validate the test by changing the landing page and ad. You just don't have enough data right now to say that one page converts better than the other in the long term. If you changed more than the destination URL of the ad, right there you invalidate the test as the ads are different. I can show you lots of data of ads that are almost the same save but for one word or comma, and with very different results in CTR and conversions.

 

You can use website optimizer (now integrated in Google Analytics) to A/B test your landing pages and see if the new one's are statistically outperforming the old ones.

 

Wow, not the thread subject at all. But gold star for you for giving me more data on my new experiment with this hypothesis: "nobody on DP comprehensively reads the thread before posting."

 

Yes, and I didn't validate that cats have wet noses either, luckily that's not what I said I validated. Nothing of this thread so far has been about determining whether landing page A or B was better, so why would I have been talking about it there? I said that by changing those factors multiple times and observing no change in the overarching pattern, that the pattern was likely not being caused by those factors. Focus!

BTW, just did it a 4th time and am seeing similar results so far yet again. So help me hypothesize so that I have something to test with an experiment...or please just leave the thread alone!

 

Ten clicks is nothing.

You have to have at least 35 click to make sure.

 

This might help: www dot learntoreadfree dot com