Article

8min read

Revenue Per Visitor: Definition, Formula & Best Practices

Revenue Per Visitor, or RPV, is a highly effective way to measure how your online sales are performing. With so many metrics available to businesses, it is one of the most comprehensive available, covering the myriad of blind spots that are inherent in many other data metrics.

What Is Revenue Per Visitor?

Revenue Per Visitor is a way to accurately assess average revenue per visitor to your website.

The calculation is made by dividing the total income by the number of visitors during a specific time period. For example, if your income for January to March is $20,000, during which time you attracted 5,000 visitors, then your RPV would be $4.

RPV works differently from Conversion Rate (CVR), although it is a similar metric using some of the same data. As we will see, CVR is not as accurate as RPV, and can lead a business down something of a blind alley and affect both website design and marketing campaigns.

Why Use Unique Visitors?

Revenue Per Visitor does not use the total number of visitors to your site in its calculations. Instead, it counts only unique visitors, with each individual counting as just one visitor. In other words, visitors, not visits. This is because almost all first-time visitors to a site, more than 99%, will not make a purchase. Typically they might want to compare prices or offers on another site, or perhaps they want to mull things over before making a purchase, particularly if it’s an expensive one.

This can skew the results considerably and provide falsely negative data on how much a company will earn per visitor. On the other hand, if you are gaining a lot of recurring buyers, you may wish to document this trend and use total buyers instead. This might depend on what it is you are selling, as well as whether you want to track individual habits as well as revenue.

Why Use Revenue Per Visitor?

RPV is a much more thorough and insightful metric for those who wish to measure online sales than Conversion Rate (CVR).

In fact, many of the metrics used in e-commerce have enormous blind spots that can lead you to make poor decisions made with accurate, but incomplete, data.

Put simply, conversion rate tracks the percentage of visitors to your site that makes a purchase. If the product or products you are selling are of one price, this would paint a complete picture of how your business is performing. Most e-commerce platforms, however, are selling products of different value. For example, a conversion rate of 3% for a fidget spinner worth $2 is very poor, but for a sofa worth $1,000 it is highly desirable. So what is needed is a metric that takes into account AOV – Average Order Value.

The average order value is the average revenue made per visitor. The aforementioned fidget spinner has an AOV of just 6p, but the sofa has an AOV of $30. So why not use this metric to calculate how much a business earns per visitor? Well just because conversion rate comes with an inherent blind spot, doesn’t mean it’s a metric without merit. If the only data you have is AOV, how are you to know if your website is performing to its full potential? Also, a low conversion rate directly affects your website’s ranking. The lower the ranking, the less likely customers are going to find it in the first place. The answer? Combine both metrics. This is how RPV came into being.

How To Improve Revenue Per Visitor

There are many methods and ideas that help improve a business’s average revenue per customer, but each tip should not automatically be considered a one size fits all deal. In fact, a lower RPV is not always a sign of a poorly performing website. For example, high traffic that garners low sales, and therefore a low RPV, can lead to higher sales over an extended period of time. It is therefore often good practice to keep every metric in proportion to other data.

Having said that, there is a little doubt about the power and insight that revenue per visitor can provide. These include:

  • Upselling
  • Recommendations
  • Reward programs
  • Basket reminders
  • live chat
  • Optimization

Upselling

A tried and tested method of commerce since humans began trading, upselling is the method of suggesting upgrades or add-ons to the original purchase. For example, a tech salesperson might suggest that spending an extra $50 would result in a faster processor, meaning the laptop will avoid becoming obsolete in the near future. Or perhaps the next model up, which is only $70 more, has a free upgrade for the latest operating system.

Upselling is a balance, one where it is important to make the customer feel as if they are getting something of worth, or perhaps that it would be foolish not to when spending that little bit more. According to a study from Predictive Intent, upselling increases sales by over 4% for e-commerce businesses, and is twenty times more likely to be effective than non-complimentary recommendations.

Recommendations

Despite what was said in the previous paragraph, recommendations can be a highly effective way to increase revenue. By simply suggesting other items, there is little doubt that basket totals increase. By how much will depend on the effectiveness of the recommendation engine. This can be tricky for small businesses with little or no data on their customers, the more information you have, the better recommendations work, but some common sense can work wonders here.

For example, complimentary suggestions, those related to the purchase, will likely result in higher purchase totals. If someone buys a laptop, suggesting a wireless mouse or external hard drive will likely be yield better results than stereo equipment.

Reward programs

It is hard enough finding customers in the first place, when you have their attention, creating an experience and incentives for them to return is essential. Reward programs are perfectly designed to achieve this.

Much like supermarket reward points, these systems encourage customers to return to an e-commerce website, rather than a rival business, by making it worthwhile for them to do so. Reward points are just one method of encouraging loyalty, however. Exclusive offers for returning customers have proven to be highly effective when implemented in the right way. This can be particularly useful during the Christmas period when businesses are most likely to encounter new customers.

Basket Reminders

Revenue per visit can be greatly reduced by those abandoning their purchase halfway through the process. Some estimates have the rate of abandonment at almost 70%. There are many reasons why this might happen, and some, such as the process taking too long, can be dealt with by redesigning parts of the website. Whatever the reason, it is possible to turn some of those near misses into hits.

Simple apps and email campaigns can target those easily distracted customers into big spenders, and such solutions are highly cost-effective. What’s more, setting up such apps and campaigns take the minimum of effort and can lead to loyal customers that make regular purchases.

Live Chat

Depending on the size of a business, setting up a live chat feature can increase revenue in some surprising ways. Firstly, customers are more trustful of a website that has an easy to use customer service platform, and live chat is the most convenient online source.

Secondly, particularly if it’s a major purchase, many customers seek reassurance or extra information about a product. Such reassurance not only makes it more likely a sale will occur, but that returns and unsatisfactory experiences can be greatly reduced.

There is also the opportunity for additional revenue from upselling or special offers to be presented to the customer, one to one. Don’t forget as well to A/B test your live chat solution. You may be surprised!

Optimization

Website optimization is key to best practices in e-commerce and is the most thorough and data-driven aspect of understanding how well a website is performing. This might include heat maps, where businesses can see which part of a page’s content has been engaged with most or testing different versions of a website to ascertain which setups work best. This is known as A/B testing.

Optimization provides data beyond simple metrics and allows a business to make sense of the data at a much more profound level, putting into context what might otherwise be cold, hard numbers that lack context.

Calculating Year Over Year Revenue

Site revenue is pretty straightforward to calculate over the course of a single year. For example, in 2014 Business A’s revenue was $200, in 2015 it was $250. Subtract the $200 from the $250, leaving $50. Then divide the increased total by the original figure from 2014 (50 divided by 200), equalling 0.25. Lastly multiply that by 100, giving you the figure of 25% growth.

For year on year calculations you will need to use Excel, so as you might imagine, it isn’t so straightforward. This method works for any growth calculation beyond one year.

Year Revenue
2014 $200
2015 $250
2016  $350

To calculate overall growth, from 2014 to 2016, simply use the formula above, but the calculating year on year requires three steps. First using the year ending figures for 2016, divide it by the yearly figures for 2014 (350 divided by 200 = 1.75).

Subscribe to
our Newsletter

bloc Newsletter EN

We will process and store your personal data to send you communications as described in our  Privacy Policy.

Article

8min read

Client and Server-Side A/B Testing – The Best of Both Worlds

We’re enriching our conversion rate optimization platform with a server-side A/B testing solution. What is server-side A/B testing, you ask? It’s the subject of an announcement of ours that will make anybody who’s passionate about experimentation pretty excited…because it means they can now test any hypothesis on any device.

No matter if you want to test visual modifications suggested by your marketing team or advanced modifications tied to your back office that are essential in the decision-making process of your product team, we’ve got the right tool for you.

What’s the difference between A/B testing client-side, and A/B testing server-side?

Client-side A/B testing tools help you create variations of your pages by changing the content sent by your server to internet users in the web browser. So, all the magic happens at the level of the web browser (called ‘client’ in the IT world), thanks to JavaScript. Your server is never called, and never intervenes in this process: it still sends the same content to the internet user.

Server-side A/B testing tools, on the other hand, offload all of this work from the web browser. In this case, it’s your server that takes on the task of randomly sending the internet user a modified version.

4 reasons to A/B test, server-side

Running an A/B test server-side has many advantages.

1. Dedicated to the needs of your product team

Client-side A/B testing is often limited to surface-level modifications. These refer to visual aspects, like the page’s organization, adding or deleting of blocks of content or modifying text. If you’re interested in deeper-level modifications related to your back office – for example, reorganizing your purchase funnel, or the results of your search or product sorting algorithm – it’s a bit more complicated.

With server-side testing, you have a lot more options to work with, since you can modify all aspects of your site, whether front-end or back-end. With server-side testing, you have a lot more options to work with, since you can modify all aspects of your site, whether front-end or back-end.

All of this is possible because you remain in control of the content sent by your server to your website visitors. Your product team will be overjoyed since they’ll gain an enormous amount of flexibility. They can now test all kinds of features and benefit from a truly data-driven approach, to make better decisions. The price of this increased flexibility is the fact that server-side testing requires your IT team to get involved in order to implement modifications. We’ll get back to this later.

Your product team will be overjoyed to test all kinds of features

2. Better performance

Poor performance – loading time or the flickering effect – is often the first criticism made about client-side A/B testing solutions.

In the most extreme cases, some sites only add the JavaScript tag to the footer of the page to avoid any potential impact on their technical performance. This policy automatically means excluding using any client-side A/B testing tools, since a ‘footer’ tag is often synonymous with flickering effect.

When using a server-side A/B testing tool, you don’t have any JavaScript tag to insert on your pages, and you’re in control of any potential performance bottlenecks. You also remain responsible for your company’s security policy and the adherence to internal technical procedure.

3. Adapted to your business’s rules

In some cases, your A/B test might be limited to design-related modifications, but you have to deal with profession-specific constraints that make it difficult to interpret a classic A/B test.

For example, an e-commerce merchant might understandably wish to take into account canceled orders in their results, or else exclude highly unusual orders which skew their stats (the notion of outliers).

With a client-side A/B test, a conversion is counted as soon as it occurs on the web browser side when the purchase confirmation page loads or a transaction event type is triggered. With a server-side A/B test, you remain in complete control of what is taken into account, and you can, for example, exclude in real time certain conversions or register others after the fact, by batch. You can also optimize for more long-term goals like customer lifetime value (LTV).

4. New omnichannel opportunities

Server-side A/B testing is inseparably linked to omni-channel and multi-devices strategies.

With a client-side solution – which relies on JavaScript and cookies – your playing field is limited to devices that have a web browser, whether it’s on desktop, tablet or mobile. It’s therefore impossible to A/B test on native mobile apps (iOS or Android) or on connected objects, those that already exist and those still yet to come.

On the other hand, with a server-side solution, as soon as you can match up the identity of a consumer, whatever the device used, you can deploy A/B tests or omnichannel personalization campaigns as part of a unified client journey. Your playing field just got a lot bigger 🙂 and the opportunities are numerous. Think connected objects, TV apps, chatbots, beacons, digital stores…

Use cases for server-side A/B testing

Now, you’re probably wondering what you can concretely test with a server-side solution that you couldn’t test with a client-side tool?

Download our presentation: “10 Examples of Server-side Tests That You Can’t do With a Client-side Solution”

Included are tests for sign up forms, tests for order funnels, tests for research algorithms, feature tests…

How can you put in place a server-side A/B test?

To put a server-side A/B test in place, you’ll need to use our API. We’ve described below in general terms how it works. For more information, you can contact our support team, who can give you the complete technical documentation.

When an internet user lands on your site, the first step is to call our API to get a unique visitor ID from AB Tasty, which you then store (ex: cookie, session storage). If a visitor already has an ID from another visit, you’ll use this one instead.

On pages where a test needs to be triggered, you’ll then call our API passing in parameters the visitor ID mentioned above and the ID of the test in question. This test ID is accessible from our interface when you create the test.

As a response to your API request, AB Tasty sends the variation ID to be displayed. Your server then needs to build its response based on this variation ID. Lastly, you need to inform our data servers as soon as a conversion takes place, by calling the API with the visitor ID, and data relevant to the conversion, like its type (action tracking, transaction, custom event…) and/or its value.

Don’t hesitate to use our expertise to analyze and optimize your test results thanks to our dynamic traffic allocation algorithms, which tackle the so-called ‘multi-armed bandit’ issue.

As you’ve seen, putting in place a server-side A/B test absolutely requires involvement from your tech team and a change in your work routine.

While client-side A/B testing is often managed and centralized by your marketing team, server-side A/B testing is decentralized at the product team or project level. While client-side A/B testing is often managed and centralized by your marketing team, server-side A/B testing is decentralized at the product team or project level.

Should you stop using client-side A/B tests?

The answer is no. Client and server-side A/B testing aren’t contradictory, they’re complementary. The highest performing businesses use both in tandem according to their needs and the teams involved.

  • Client-side A/B testing is easy to start using, and ideal for marketing teams that want to stay autonomous and not involve their head of IT. The keyword here is AGILITY. You can quickly test a lot of ideas.
  • Server-side A/B testing is more oriented towards product teams, whose needs involve more business rules and which are tightly linked to product features. The keyword here is FLEXIBILITY.

By offering you the best of both worlds, AB Tasty become an indispensable partner for all of your testing and data-driven, decision-making needs.

Don’t hesitate to get in touch to discuss your testing projects – even the craziest ones!