Scientific Web Personalisation
Web Personalisation and the Scientific Method
Digital marketers have a wide range of tools and approaches to choose from when optimising customer experience programmes. Web Personalisation is one marketing approach that is generally poorly understood when compared with higher profile techniques like direct marketing and programmatic media buying. However, this approach (also known as ‘test & target’) can achieve significant returns if used appropriately and deployed well.
Web Personalisation enables the delivery of tailored content to a predefined audience based upon specific criteria. This allows marketers to:
Display product-specific banners on websites to visitors that have either visited that product’s webpage or have shown an interest in it offline
Trial proposed changes to a website to only a proportion of visitors in order to determine whether or not the change should be made for everyone. This avoids the cost of making changes that provide no tangible benefit
Optimise website journeys for visitors based on their previous behaviour. For example, not displaying ‘sign up’ calls-to-action to visitors that have previously logged in
Change the content of a website to reflect the country or time zone in which a visitor is located
n3 Hub has seen properly deployed Web Personalisation programmes deliver considerable benefits to marketers through:
Higher conversion rates on lead generation websites
Greater customer acquisition
Increased customer engagement with content
Added return visits
Longer time on site
Lower bounce rates
The pitfalls of poor preparation
Poorly designed Web Personalisation programmes can cause more problems than they solve. They can end up delivering nothing more than additional technical complexity to organisations that attempt to implement them without preparation.
Technical complexity and poor performance can be caused by a number of factors including: incompatibility between the Web Personalisation tool and the website; poor rendering times of personalised content; or the need for extensive development support to create content.
One of the main causes of a failed Web Personalisation deployment, however, is completely non-technical: it is a failure to grasp the scientific nature of the channel. Indeed, it is usually because a ‘traditional’ marketing approach rather than a scientific marketing approach to delivering the programme has been used.
A traditional marketing approach goes through the following stages:
A business objective is generated, either through insight or direct order, then presented to marketing to generate campaigns to meet the desired outcome
Marketing creates concepts that it works through with stakeholders until acceptance is gained
The concept is turned into creative that is placed into the most appropriate channels with as much targeting as possible to get it in front of the best audience
The campaign’s outcomes are measured but usually without the ability to drive performance data directly into the marketing decision-making process
This traditional approach is not effective in the context of Web Personalisation, however, because it is able to be measured very precisely and benefits from a ‘test & learn’ framework.
What often happens in a traditional marketing-based approach to Web Personalisation is that sweeping changes are made to websites through the Web Personalisation tool - for example, the swapping out of an entire page with another completely different one or making changes to multiple pages or content blocks at the same time. The downfall of this approach is that it isn’t frequent or iterative. It can take a long time for the content to be built in the tool and there are major issues with measuring effectiveness - as it becomes virtually impossible to tell what worked and what didn’t.
Digital marketing, and Web Personalisation in particular, is much better suited to a scientific approach, based upon a ‘test & learn’ hypothesis.
A scientific marketing approach has the following stages:
Data analysis, usually using a Web Analytics tool. This can identify areas of a website that are not converting well, or potential reasons for a drop-off in conversions (e.g. poor user experience)
Developing a hypothesis based on this analysis that states that if X is changed then it is likely that positive outcome Y (e.g. an uplift in conversion) will result
Testing the hypothesis using the Web Personalisation tool to make targeted changes to the website. The goal should be to achieve a statistically significant result based upon a sample of the site’s visitors. A percentage of the audience should see the modified version of the website while the rest should see the existing content. The usual rule of thumb is assigning 50% for the modified version and 50% for the control version
Review of the results of the test, again usually using a Web Analytics tool or the Web Personalisation platform itself, to determine whether the test delivered a statistically significant result or not
Drawing conclusions from the test results that usually recommend the changes are made live or that the hypothesis is amended and the test repeated
Most Web Personalisation tools have good statistical analysis and A/B testing capabilities and this scientific approach ensures these capabilities are leveraged to their fullest.
When a scientific approach is taken, more time will be spent on analysis (to both build hypotheses and measure results) than will be spent on building the tests themselves. Data and its correct and disciplined analysis is vital to the success of any Web Personalisation programme.
A common mistake made by marketers is to measure success by the number of tests that are running at any given time. This is akin to measuring the success of an email marketing campaign by the number of emails sent. The main problem with a volume-based approach, however, is that it becomes impossible to tell which tests are successful and which aren’t. There is also the practical issue of how to avoid clashing tests (i.e. tests fighting for the same visitor). A better measure of success is the business outcomes that it generates, which should not stop when the test does.
By focusing on incremental measurable tests underpinned by a scientific marketing approach, Web Personalisation can become a valuable part of any customer experience programme. In reality, nearly every digital marketing channel can benefit from taking a scientific, data-driven approach and in an increasingly data-driven world marketers need to act more scientifically every day.
Where a Customer Data Platform fits in
A Customer Data Platform (CDP) contributes to a Scientific Marketing based Web Personalisation program in the following ways. It:
Provides first party data audiences to drive Web Personalisation programmes beyond that which is possible with only web analytics data
Creates a single marketing customer view that is augmented with a variety of data to enable better hypothesis generation and results analysis
Creates an ongoing attribution model from both online and offline sources to allow the value of implemented tests to be continuously monitored
Enables marketers to build omni-channel strategies and target the customer on their preferred channel