Why the meta description is so important.

[directmarketingadvice]

Unlike the beautifully crafted unique descriptions that I use based on vast experience of selling to TGBP and being pretty well aware of what Mr&Mrs Jones digs..

Utter nonsense.

For a start, "experience" of what? At best, all you have is experience of guessing and, because you've never split-tested, you've never had meaningful feedback from those guesses.

And, without feedback, you can't know how well you're doing.

(You don't even know what clickthrough rate you're getting.)

Gene Schwartz - one of the greatest copywriters that ever lived - said he couldn't guess which version would win any better than around 50%.

Of course, he was talking about two versions written by Gene Schwartz, not one written by him and another written by an illiterate hillbilly.

When I put up those split-tests I did and you guys can never guess the winner better than 50%, that's two versions written by someone with experience of testing.

What you guys are calling "good" meta descriptions are ones that have some benefits. That's as sophisticated as it gets.

You've no idea if they're the best benefits, or in the best order, or with the right tone, the right verbs, the right adjectives... none of that stuff.

As a result, you've no idea if you're throwing away 20% or 50% (or more) of your potential visitors.

And, given you can't guess the winners of my split-tests better than a chimp throwing a dart at a dartboard, we can safely assume that, accross your pages, you'll have a lot of underperforming meta descriptions.

Steve

 

[awebapart.com]

Here we go again!

At best, all you have is experience of guessing and, because you've never split-tested, you've never had meaningful feedback from those guesses.

Unless your site home page or other pages are in the fortunate position of being reindexed by google on a very regular basis, e.g. like the UKBF home page being reindexed say every 10 minutes, it is impossible to realtime parallel split test different versions of meta descriptions around the same time, to give meaningful feedback. Anything else, not testing the real thing of testing different versions of actual natural search result description tested over the same period, would be testing other things and as such not giving true meaningful feedback.

Gene Schwartz - one of the greatest copywriters that ever lived - said he couldn't guess which version would win any better than around 50%.

Of course, he was talking about two versions written by Gene Schwartz, not one written by him and another written by an illiterate hillbilly.

So Gene cannot work out what is better without testing? Then how did Gene narrow down to just two versions of copy, out of the hundreds or thousands of possible variations that Gene could have possibly written?

Gene didn't test all of those hundreds or thousands of variations. Gene used skill, experience, and common sense to rule out the other versions, apparently breaking the rules that you keep saying, i.e. you must test everything.

You cannot test everything, otherwise you will spend all your time testing everything and doing nothing.

Even Gene doesn't test everything. If Gene is choosing 2 versions to test, and ruling out all the other hundreds or thousands of possible versions without testing, then Gene is testing such a small percentage of variations, that Gene is testing virtually nothing.

 

[Mr Niche]

I also agree with Title tags and Meta discriptions.

I was wondering though......what happens if you have very similar or the same meta discriptions for different pages?

Regards

Mr Niche

 

[Top Hat]

How do you split test meta descriptions?

Do you have to test it with adwords?

 

[awebapart.com]

How do you split test meta descriptions?

Do you have to test it with adwords?

You can only accurately parallel split test meta descriptions around the same time if you have both a lot of search engine visitors and pages that are reindexed by google many times each day (so your changes to meta descriptions during the test are picked up regularly). Most of the businesses on this forum do not have that luxury.

Testing anything else, like adwords descriptions rather than natural results descriptions, is testing something different and as such is not giving true meaningful feedback.

Unless your site home page or other pages are in the fortunate position of being reindexed by google on a very regular basis, e.g. like the UKBF home page being reindexed say every 10 minutes, it is impossible to realtime parallel split test different versions of meta descriptions around the same time, to give meaningful feedback. Anything else, not testing the real thing of testing different versions of actual natural search result description tested over the same period, would be testing other things and as such not giving true meaningful feedback.

 

[directmarketingadvice]

apparently breaking the rules that you keep saying, i.e. you must test everything.

When did I say that?

It seems you're creating a straw man argument. Though, bravo for being able to knock it down.

Steve

 

[OldWelshGuy]

How do you split test meta descriptions?

Do you have to test it with adwords?

That is pretty much what you do (in chunks)

 

[bdw]

IMO talk of split testing makes people sound as though they know what they are talking about but in reality proper scientific, statistical split testing on stuff like this by little guys like us is virtually impossible.

Apart from having the required knowledge and resources, designing split (or multivariate) testing takes an incredible amount of time. Even the best of split tests have an element of subjectivity about them.

There are few if any self employed people working from home who have the time and resources to set tests like this up.

Oh and by the way! You have to also find clients who can (a) understand what you are rattling on about and (b) be willing to pay for it.

 

[MrTempleDene]

I was wondering though......what happens if you have very similar or the same meta discriptions for different pages?

It's generally accepted that you are better having custom descriptions for each page.

 

[OldWelshGuy]

IMO talk of split testing makes people sound as though they know what they are talking about but in reality proper scientific, statistical split testing on stuff like this by little guys like us is virtually impossible.

Apart from having the required knowledge and resources, designing split (or multivariate) testing takes an incredible amount of time. Even the best of split tests have an element of subjectivity about them.

There are few if any self employed people working from home who have the time and resources to set tests like this up.

Oh and by the way! You have to also find clients who can (a) understand what you are rattling on about and (b) be willing to pay for it.

I disagree with this as I have found the opposite to be true. ALL you are doing is trying to find which of the two performs better. We are not talking about rocket science here, we are talking about a simple 'which one gets most clicks' scenario.

And as for "talk of split testing makes people sound as though they know what they are talking about " could it be that those talking about split testing are doing so because they DO know what they are talking about?

I don't see what is so hard to grasp when we are talking about a simple split test of 'which performs best'.

 

Here we go again!

Unless your site home page or other pages are in the fortunate position of being reindexed by google on a very regular basis, e.g. like the UKBF home page being reindexed say every 10 minutes, it is impossible to realtime parallel split test different versions of meta descriptions around the same time, to give meaningful feedback. Anything else, not testing the real thing of testing different versions of actual natural search result description tested over the same period, would be testing other things and as such not giving true meaningful feedback.


So Gene cannot work out what is better without testing? Then how did Gene narrow down to just two versions of copy, out of the hundreds or thousands of possible variations that Gene could have possibly written?

Gene didn't test all of those hundreds or thousands of variations. Gene used skill, experience, and common sense to rule out the other versions, apparently breaking the rules that you keep saying, i.e. you must test everything.

You cannot test everything, otherwise you will spend all your time testing everything and doing nothing.

Even Gene doesn't test everything. If Gene is choosing 2 versions to test, and ruling out all the other hundreds or thousands of possible versions without testing, then Gene is testing such a small percentage of variations, that Gene is testing virtually nothing.

Thank you Paul you put it so much better than I can.

Earl

 

[awebapart.com]

apparently breaking the rules that you keep saying, i.e. you must test everything.

When did I say that?

I very much doubt whether you would explicitly say that, as it is totally impractical, but it is what I am logically inferring that you saying from other things you are saying about the reasoning for split testing and the bad things it rules out (guesswork/personal opinion).

On this thread, and many many other threads, your recurring theme is that guesswork/personal opinion is not good, and split testing should be used to rule out this guesswork/personal opinion...

You mean "guesswork"?

I suspect all you're doing is writing one that appeals to yourself and assuming everyone is identical to you.

For a start, "experience" of what? At best, all you have is experience of guessing and, because you've never split-tested, you've never had meaningful feedback from those guesses.

I am saying that the only way you can rule out guesswork/personal opinion is if you split-test everything. If you are narrowing down, without testing, the hundreds or thousands of possible variations to say 2 versions, before you start testing, you are still using personal opinion and guesswork, and as such the split test is not ruling out guesswork or personal opinion at all. The only way guesswork or personal opinion can be ruled out is if everything is tested. Obviously this is impractical and I certainly wouldn't recommend it.

So my logical conclusion, is that split testing does not rule out guesswork or personal opinion, because guesswork or personal opinion is still used to narrow down the versions in the first place. As such it cannot be said that split-testing is good and guesswork is bad, because split testing still includes guesswork. Likewise split-testing cannot be sold on the idea that it removes guesswork.

I do not have a problem with guesswork or personal opinion. Those words might sound derogatory, but more positive words I would use are skill, experience, judgement, common sense, market awareness etc.

 

[awebapart.com]

I do not have a problem with guesswork or personal opinion. Those words might sound derogatory, but more positive words I would use are skill, experience, judgement, common sense, market awareness etc.

I should add that I do not have a problem with split-testing either. By all means use it if it is practical to do so, if it can return meaningful results, if you are not sure which versions to use, as part of an ongoing process of continuous improvement, etc. Just don't use it and think that it is removing, or is opposite to, guesswork and personal opinion, because it still includes guesswork and personal opinion.

 

[bdw]

I don't see what is so hard to grasp when we are talking about a simple split test of 'which performs best'.

It's not at all hard to grasp.

As Awebapart has said the problem lies in the choice of what to split test and this discussion evolved into whether the title or description was the most important element in getting people to click into website. How do you split test that?

 

this discussion evolved into whether the title or description was the most important element in getting people to click into website. How do you split test that?

Not quite my point was that one should use the meta description for enticing people to click a site as the title has a seperate function.IMHO

Earl

 

[directmarketingadvice]

And as for "talk of split testing makes people sound as though they know what they are talking about " could it be that those talking about split testing are doing so because they DO know what they are talking about?

Not only that, but how does someone know anything without feedback?

How could an SEO know what works in SEO if they'd never seen the impact of their changes to rankings? (Or, if not rankings, traffic.)

People who split-test learn what actually works. And, if we see the same thing winning split-test after split-test on different sites, we are able to put together an idea of what is "best practice".

But even within that "best practice", there are things you can't know until you actually test them. For example, saying "lcd" v saying "digital", quoting a price or quoting the savings ("£299" v £50 off") or what order benefits should be presented in.

Steve

 

[OldWelshGuy]

It's not at all hard to grasp.

As Awebapart has said the problem lies in the choice of what to split test and this discussion evolved into whether the title or description was the most important element in getting people to click into website. How do you split test that?

Better sidestep than Gareth Edwards

Your post was about split testing and you made a rather bold statement that kind of said that people who are suggesting split testing don't know what they are talking about ("as if they know what they are talking about") That is pretty confrontational isn't it. Yet when challenged, you completely change tack.

Experience tells you pretty much what you need to know to get close, then you split test to refine.

 

[bdw]

How could an SEO know what works in SEO if they'd never seen the impact of their changes to rankings? (Or, if not rankings, traffic.)

You don't need split testing to measure the effects of changes made for ranking purposes.

Listen, I am not knocking split testing. It is a proven method of analysis but what I said was...

There are few if any self employed people working from home who have the time and resources to set tests like this up. Oh and by the way! You have to also find clients who can (a) understand what you are rattling on about and (b) be willing to pay for it.

AFAIAC that applies to me.

 

[directmarketingadvice]

I very much doubt whether you would explicitly say that, as it is totally impractical, but it is what I am logically inferring that you saying from other things

So...

The fact I've never said it + the fact you wouldn't expect me to say it + the fact that it's "totally impractical" = I "keep saying" it.

That's your rationale, is it?

Let me suggest something: being unable to test everything is an incredibly feeble reason to test nothing.

I am saying that the only way you can rule out guesswork/personal opinion is if you split-test everything. If you are narrowing down, without testing, the hundreds or thousands of possible variations to say 2 versions

You seem to be assuming that, once your split-test is over, you never split-test that thing again. Again, that's something I've never said, so for that reason, I guess it must be something I "keep saying".

You seem to be hung up on my use of the word "guesswork". Let me quote what I said:

At best, all you have is experience of guessing and, because you've never split-tested, you've never had meaningful feedback from those guesses.

The key words there are "all" and "never".

I'd also suggest you read post #37, which is what I was responding to.

Steve

 

[directmarketingadvice]

You don't need split testing to measure the effects of changes made for ranking purposes.

No, but you need feedback. That was my point.

Steve

 

[bdw]

(OWG) That is pretty confrontational isn't it.

It was not intended as confrontational and if anyone took it that way then I apologise but we know that by suggesting this it is clear that you are once again just trying to stir things up. In your position you should be doing the opposite.

 

Now that it seems that Awebapart has destroyed the myth that split testing will produce the ultimate .

And can at best only produce the better response from several variations.

The best copy ( not the ultimate as that may never be known ) would seem to be produced by the individuals experience and creativity.IMHO

Earl

 

[Dawg]

Now that it seems that Awebapart has destroyed the myth that split testing will produce the ultimate .

And can at best only produce the better response from several variations.

The best copy ( not the ultimate as that may never be known ) would seem to be produced by the individuals experience and creativity.IMHO

Earl

Quite agree. Cognitive fluency kicks ass.

 

Quite agree. Cognitive fluency kicks ass.

Stop agreeing with me Dawg.

You is making me nervous.:|

Earl

 

[scottdsmith]

Totally agree with this - I have been able to dramatically increase a clients web traffic almost instantly with improving their title and meta description by increasing click conversions. It still amazes me the amount of businesses that spend a small fortune on the website build and marketing but don't give enough attention to this area, which is after all, your "advert" you're showing off in the SERPS.

 

[OldWelshGuy]

It was not intended as confrontational and if anyone took it that way then I apologise but we know that by suggesting this it is clear that you are once again just trying to stir things up. In your position you should be doing the opposite.

Who is we? Are you now the official spokesman for the forum

You stated that people are posting as if they know what they are talking about. I stated that was conforntational. No more, no less.

So cmon, who is the 'we' fess up. name names who is in the BDW gang (or is it multiple personality disorder?) oh, and how big is your dad? My dad is 6'1" is your dad bigger?

 

[bdw]

He's at it again folks.

 

Who is we? Are you now the official spokesman for the forum

You stated that people are posting as if they know what they are talking about. I stated that was conforntational. No more, no less.

So cmon, who is the 'we' fess up. name names who is in the BDW gang (or is it multiple personality disorder?) oh, and how big is your dad? My dad is 6'1" is your dad bigger?

I'm in his gang ,all ways thought the old split test was a bit iffy ,young Paul has exposed it for what it is .A test of 2 guess's and not the unadaulterated genius that me and my gang come up with.

Artistic creativity at its highest level will always beat the geeks slaving over there fact and figures.

When me and Turing were inventing the computer we did artistic scribbles on a blackboard.

Shame he beat me to it.

P.S and my old man was one of the few Jewish pilots in WW2 and if he was alive you would not want to mess with him.

Earl

 

[KM-Tiger]

Artistic creativity at its highest level will always beat the geeks slaving over there fact and figures.

You're missing a trick there.

The real winners combine artistic creativity with the geeks facts and figures.

 

You're missing a trick there.

The real winners combine artistic creativity with the geeks facts and figures.

I know that.:|

But you got to allow me a bit of artistic license.

Earl