Note: Although this application note is written for Pay Per Click Campaigns, the principles can be applied to any form of advertising such as email campaigns, mailshots and banners.

The first stage of any advertising testing is to come up with different adverts.  With PPC adverts this is fairly straightforward as you only have two (or maybe three) different elements to test:

The advert title

The advert body.

Although the body can be broken down into two possible elements – either two lines as you do with Google Adwords or the hook (the section that confirms interest) and the call to action (getting the recipient to do something).

So for each different version of the advert we will need a different campaign setting up on AdTraka. For each of these you will get a different tag (_t=). Simply add the tag to the destination url for the advert.  It does help at this stage to all your campaign something descriptive.

Now go to events, and view the telephone conversions by campaign – this will show how many telephone conversions you have per campaign, and how many impressions that advert had.

This where many people make a basic mistake. We need to know if we are seeing a result that is significant. If you have had 5 impressions of each advert and one had 1 conversion and the other had none – is this significant.  Well no it isn’t.  You need to make sure that each.

So let’s calculate statistical significance for our results.

OK we need to make sure we have done enough test to make the numbers start hanging together – it depends on budget, time constraints etc, but you should have at least 10 conversions for each test (30 is better).  This will depend on your website and circumstances.

I will present to you a rough and ready method for making sure your results are significant – if you want a more statistically correct way of doing it, I will present this in a later article, but for most people this will sufi (and will save going cross eyes over the stats).

If we have two tests one has converted 19 times and another 17 times.

Total the number of telephone conversions together. (so in our case this comes to 36)

Calculate the difference between the telephone conversions (in our case 19 -17 = 2)

Take the square root of the total (36) = 6.

If this is greater than the difference then it’s not significant.

So in our case we would need to do further testing.

I would like to reiterate, this is not statistically the best method to use, but it’s a darn site easier than the more accurate methods, and in more cases will do the job.

So there you go, you now know how to test different adverts with telephone conversion tracking.

However, if you do have a significant proportion on your sale coming via incoming calls then you should definitely be considering tracking your calls down to keyword level.

It is often quoted that Pareto said 80% of causes come from 20% of effects.

Or, in online marketing terms, 80% of your turnover comes from 20% of your online advertising spend.

The question is, which 20%?

Call tracking will allow you to identify that 20%, with a degree of accuracy that may surprise you.

And that 20% saving isn’t just for Christmas, oh no, it’s for life.   After you have identified the keyphrases that sell and finished using your chosen call tracking system, visitors won’t suddenly hop to a new keyphrase, so you’ll keep on making those savings.

Can I use call tracking for anything more?

That depends on the keyword level call tracking system you choose to use. Certainly the more advanced systems can tell you quite a bit more (our own included), such as

• The last page visitors saw before calling you.
• How many times they came on to your site before calling you.
• The organic and paid phrase they used to get to your site.
• Their IP address so you can physically locate their location.

Will it be worthwhile?

Well that depends on a lot, but the rule of thumb tends to be, if you are spending over £1,000 per month on pay per click, then you will probably find call tracking worthwhile, particularly if that is over large amount of keyphrases.

So if your business receives incoming sales call, is spending more than £1,000 per month and would like to reduce your PPC costs, you should really be talking to us about keyword level call tracking.

OK, to set the destination numbers separately simply go to the wrapper for each number (the <span class=‘adtrakaTelephone’ bit)  and add a line within the initial span tag that reads :

destination = ‘xxxxx’;

where xxxx is the destination number you want for that number.

So, if we had two offices, one in Southampton and one in Bournemouth.

For the appearance of the Southampton number we would set the wrapper to:

<span class=‘adtrakaTelephone’ destination=‘0238123456’>

and for Bournemouth it would be

<span class=‘adtrakaTelephone’ destination=‘01202123456’>

and that’s it!  AdTraka picks up on the different destination numbers and displays different numbers for each of them.

But I want geographic numbers appearing for the relevant sites

Geographic numbers have a higher conversion rate than freephone in their own area, so it is wise to use geographic numbers if your target market place is in that area.

To do this with AdTraka we would simply set a prefix variable with in the span.   So, for our Southampton example, to make sure only Southampton numbers appear, we would use:

<span class=‘adtrakaTelephone destination=‘0238123456’ prefix=‘0238’>

This means AdTraka will only display Southampton geographic numbers at this location.

And there you go, AdTraka is now configured to track telephone conversions to as many destinations as you need.

So how do you identify visitors that have come through an email campaign?

Step 1: Set up a campaign for your email campaign.  If you are testing several different versions of emails, then you will need several different campaigns.   Once you have set up these campaigns, you will have your campaign id’s.

Simply append your campaign id’s to any links heading towards your website, this will identify this visitor as having come from this email, and so it will show up in your analytics reports.

Wait a minute!

What if someone calls you from the email without going to the website – what then?   When setting up your campaign you will be given a default number.  This number is allocated to the email campaign alone, so any calls coming through it will be treated as if they had been generated by that email.

Just add this number as your phone number in your emails, and then that’s it.

You can now see the calls generated by your emails right there in Google Analytics.

What does this mean?

AdTraka can select to deliver phone numbers to web visitors what have come via a tagged link (i.e. one that you have set up, such as PPC). Everyone else will see the default number only.

So if you have tagged the destination URLs for your PPC campaign with your campaign tag, visitors from these campaigns will have dynamic numbers delivered.   However, visitors coming to your website from other sources will only see your normal number.

How do I Implement PPC Optimization?

You need to add two lines to your header script:

ppc_only =1;

any_other_settings:1;

and that’s it!    AdTraka will now only deliver numbers to tagged visitors.

Please note that this means you won’t see any of your organic traffic showing up on your call reports, or if you have missed a tag on your destination URLs, these also won’t show.

From the data you get now – you can truly optimize your PPC campaign for telephone calls.

On the internet, just as on the High Street, it’s not the footfall that counts, but the profit from each visitor.

This why keyword research should figure as an ongoing activity in your Internet Marketing Strategy.

So, where do we start.  OK, well lets forget about SEO at the moment (gasp, horror,  I know but we don’t know what we want to focus on).

We start with Pay per click, we need to know what keyphrases generate buyers and not just visits.

So after setting up our pay per click campaign we monitor what is actually generating sales (or telephone enquiries if you’re using a call tracking system).

Measuring Costs

We are all in business, so it is a good idea to understand what we mean by costs. Cost is the cost per sale (or telephone enquiry).  Simply take the number of sales generated by a keyphrases over a period, divide by the Pay per click cost for that keyword, hey presto, your cost per sale.

When is cost too high?

This will depend on your business model, but the measure we use is 20% of average sale value.   In other words, 20% or our turnover we budget for going back in to online marketing.

Analysing the data.

Once you’ve got some useful data you can start making some decisions.  You will have four groups in your keywords:

1. Low Conversion, High Cost
2. Low Conversion, Low Cost
3. Low Cost, High Conversion
4. High Cost, High Conversion

OK so what do we do with the groups.

Groups 1 & 2 generally get ignored and dropped.  We make a record of it for future reference, but we really have no interest in them.

Group 3.  These are of the most interest to us.  We maintain them in a PPC campaign, whilst we focus our SEO efforts on them.   You will generally find that these are the easier of the two last groups to get Organic ratings for (not always but most of the time).

Group 4.  We drop from our PPC campaigns, but do include them in our SEO efforts, but secondary to group 3.

There you have it, 2 useable groups of keywords, that should maintain your sales without breaking the bank.

Ask the Caller

You could just ask the caller what they searched for.   Some will remember, some won’t, the other problem is that if they have been on your site for any length of time, they are more likely to remember words from your pages, rather than what they actually searched for.

This can work:

• If you double check everything with your analytics package (did anyone actually come to your website today with that search term?).
• You analyse the information statistically to remove as much incorrect information as possible.

Using Data Proxies

By using more readily available data, such as bounce rate or time on site, you can have a (good?) guess at what is converting into phone calls and what isn’t.   You will never truly know where you are right and where you are wrong, unless you see a huge dip in telephone enquiries once you drop a keyphrase.

At best this is good guesswork.  If you do want to use this method then please use statistics heavily to make sure you are confident of what you are doing, and track everything heavily afterwards, to make sure of your decision.

Multivariate Testing

This is quite an innovative solution we have tried ourselves, but it became more expensive and slower than call tracking, so we stopped using it.

The theory is quite simple, switch keywords on and off in accordance with a multivariate testing table (Taguchi seems to be the most popular at the moment).  Monitor the telephone enquiries you get, tie these in with the experiments and hey presto, you have some stats. Of course you need to analyse these, but there is plenty of information on that on the internet.

Control of this is quite difficult, and you will have to be very regimented about it, as if you attribute a call to the wrong test, it can be very difficult to rectify the data.  Be also aware that not everyone calls straight away – it may be several hours before a caller will phone you.

Hopefully thats covered most of the alternatives to a keyword level call tracking system, if you want to give any a go do let us know how you get on.