SEM Tips To Attract More Customers Today

by Hannah Dudley
18 January 2016


Search drives modern marketing organisations. There’s a reason $63.1 billion of the $537 billion per year spent on digital advertising goes to paid search. It works. Paid search targets potential customers at their most primed moment in the purchase funnel.

However, many marketers just treat search as another media channel and don’t leverage its true strengths. In this article you’ll learn why the best SEM pros treat campaigns like scientific experiments and how you can do the same to attract more customers today.

Get your lab coat on

Every paid search ad you run gives us the opportunity to make your marketing more effective and teach you something new about your customers.

How? By running experiments on different versions of ad copy and landing pages.

To run an SEM experiment you need to:

  • Make a hypothesis
  • Determine the result
  • Further analyse

Let’s apply the above to a hypothetical new shop, candles.com.au, Australia’s premier candles eCommerce site.

Our first task in this SEM experiment? To understand what drives our customers buy our product.

Work that landing page

According to Hubspot, increasing your number of landing pages can deliver at least a 55% increase in leads.

This is because landing pages are where the rubber meets the road, it’s where the magic happens. But it’s no good driving loads of traffic to your landing page if it doesn’t actually convert.

After some deliberation we come up with a worthy hypothesis for our candle site – our customers will be more likely to buy our candles if we feature a cat in the hero image.

To test the hypothesis we’ll show half of our SEM traffic our standard candle image landing page (group A) and we’ll show the other half our new cat image landing page (group B).

The imaginative name for this type of experiment is the “AB test.” If group B is significantly more likely to buy a candle, our hypothesis is right, if not – back to the drawing board. In the next section we’ll go into how we determine if a result is significant.





AB tests are important as they help you determine whether or not a change is statistically significant – that is, whether or not they are actually making a significant impact on your product.

Crunch your results

Let’s imagine we stop our experiment after running it for just a few hours . We’ve had a total of five people visit our landing pages. In each group, one visitor buys a candle. Group A (without cat) has a conversion rate of 50% and Group B (with cat) has a conversion rate of 33%. So, are our results conclusive? Not at all!

Group Visitors Bought candle Conversion rate
A 2 1 50%
B 3 1 33%

We’ve only had five people visit our landing pages. Another handful of people could completely change the results. In other words, we have no idea if our results are a result of the difference in landing page design or just dumb luck.

To overcome this issue, statisticians have come up with the idea of statistical significance, which determines how confident we be that our results are… significant. To determine if our results are significant we can use one of many free tools available such as get data driven by KissMetrics.

significance test

After plugging in our sample size and results we see that we can be 65% certain that Group A is better. Given we generally want to be 95% certain, we realise that we need to continue collecting data.

How many conversions per page do we need to prove our results are of statistical significance?

According to marketo, between 25 to 50 conversions are required before you can be somewhat confident in a landing page’s reported conversion rate. Likewise, you need 25 to 50 conversions per test version to be somewhat confident in your test results.

After running the tests for a little longer a new pattern emerges. The landing page with our cat image has a conversion rate ten times greater than the landing page without a cat.

Group Visitors Bought candle Conversion rate
A 1200 10 0.83%
B 1200 100 8.3%

We enter our experiment’s details into our calculator and find that there is a 100% certainty that our result is significant. As is normally the case on the internet, cats win.

significance test

In summary – targeting two distinct types of buyer is critical because, in order to drive conversion, the customer needs to feel like the page’s content is talking to them within the context of their path to purchase.

By showing different types of landing pages to different types of customers, we can determine who our target audience should be. Conversely, we can also identify who isn’t our target audience and hide our ads from them.

Get smarter

Running a successful experiment tells us much, but to get the full story we usually need to do some deeper analysis of our results data.

In order to have a better understanding of the effectiveness of our cat experiment, we analyse the results by the original keyword used to get to our site.

Group Keyword Visited Bought Conversion rate
A (no cat) cute candles 1,000 0 0%
A (no cat) buy candle 200 10 5%
B (cat) cute candles 1,000 100 10%
B (cat) buy candle 200 0 0%

After analysing our conversions buy keyword we see a clear trend:

  • People who searched “cute candles” only purchased if they saw a cat image
  • People who searched “buy candle” only purchased if they didn’t see a cat.

This new insight gives us two potential options which will optimise our current campaign and potentially impact our wider strategies going forward. We can:

  • Include cat images in all our landing pages and stop targeting the “buy candle” keyword
  • Show cat images for visitors that search for “cute candles” and no cat image for visitors that search for “buy candle.”

After just one experiment we’ve got two great options for optimising our campaign drastically. We’ve also got an insight into our customer’s that could drive our wider strategy  – should we change our URL to catcandles.com.au?!

It’s important to find out what your customers are doing before and after visiting your site. This is known by search experts as ‘upstream’ (online presence before visiting your site) and ‘downstream’ traffic (online presence after visiting your site). Tools like Hitwise can be used to help with this

Upstream traffic will help you identify a whole bunch of sources (forums, website, social or other affiliates) that brings traffic to your website. This would help candles.com.au identify potential partners and affiliates to forge stronger relationships with, ultimately driving more sales.

Downstream traffic tells you exactly where customers are going after they visit your website. Candles.com.au were shocked to find out that most customers visited a key competitor immediately after visiting our website. But by creating unique ad copy, and a value proposition, we were able to put a system in place to minimise this.

And this is just the beginning...

According to eConsultancy, the proportion of companies selecting paid search as the having the biggest impact on their brand has risen 4% since the previous year. And most marketers consider SEM one of the most effective digital channels for generating a high volume of leads.  

Once everything is in place, the key is to continually track, fine-tune and better your performance. Your job is to establish an amazing customer journey, in every way. In other words, make it easy for your customers to buy from you! Simply do more of what does work, and less of what doesn’t.

Ready to optimise your own search strategy? Learn more about Salmat's search offering here or call us on 1300 725 628.


About the author
Hannah Dudley
Campaign Manager - Contact and Managed Operations

Hannah Dudley is the Campaign Manager for Contact and Managed Operations at Salmat. With experience working for businesses of all sizes in Australia, UK and USA, Hannah is driven by the power of customer and marketing insights. Hannah is passionate about helping businesses understand and reach their customers in an effective way, resulting in business growth.