For many web-based projects, there are two kinds of data:
- Quantitative data– expresses the what, who, where and when details in numbers
- Qualitative data – explains the why or how in non-numerical form
Majority of web analytics tools, including Google Analytics, offer quantitative data, but most of them don’t explain the why, e.g. why visitors are taking Action A and not B. Qualitative data is necessary in order to give these answers. It is arguably more important than quantitative research, since it provides perspective which can be used to inform direction.
However, good data driven design models are those that bring both aspects together. In addition, it is necessary to develop a clear understanding of the interrelation between both data forms, which will enable creative and R&D team members to better communicate on key aspects affecting the organization’s web performance.
Specificity and experimentation
Whether data is numerical or non-numerical, it provides better value if it has been empirically derived, i.e. gathered from experimentation or observation. Empirical data provides answers to specific questions, which in turn makes decision-making easier. However, typical Big Data that is empirically derived usually won’t have a direct action specified – think about all your website metrics or website performance measurement data.
In evaluating aggregated Big Data (e.g. app downloads or total site traffic), it is difficult to isolate variables, which in turn complicates drawing hypotheses. This is especially true in a web-based environment where there are many variables. More so, different portions within a website have different individual goals that aggregate to fulfill business objectives. However, proper optimization of design is only possible where this Big Data is condensed to make sense in terms of both business and individual objectives.
Take the following example:
Good data-driven design in action
Consider a hypothetical problem where a content-oriented website is trying to find out how to maintain high engagement levels with their audience. An online research site or an e-zine has approached you as a web designer requesting that you make design changes that will improve user engagement. Where would you begin?
The first step naturally would be logging into the site’s analytics base to derive their data on visitor stay, bounce and exit rates. You will have mountains of data depending on how far back you go, as well as how many pages the site has. Next, you’ll sort all this data according to pages and timelines, and discover that there are three pages with much higher exit and bounce rates than the rest of the site – you have your quantitative data now.
You will then analyze these three pages and discover that there’s a prominent link to a different website in one of the pages, i.e. you intentionally direct visitors to a different site. This doesn’t raise too much concern, since that page has been designed as an exit point. The other pages, however, don’t have any visible reason for the high exit and bounce rates.
Why, then, are visitors leaving so often? Only qualitative data can answer that. Since the best form of data is one that is empirically derived, it’s a great time to set up user testing. Having scientifically narrowed down your focus to the two pages, testing is practically viable.
In addition, you will be able to find out whether any design changes you make are effective, now that you have very specific and empirical metrics (bounce and exit rates) to quantify your goal/objective (raise audience engagement).
Ideas For Using Your Big Data Successfully
As a web designer, your Big Data usage in decision making related to website design elements and content should be directed by the following:
- You and your analytics team and/or tools should have common ground. Train your teams to fully appreciate the implication of any metrics you’re interested in so that they will know exactly how to condense mountains of data to provide useful information for you.
- Always insist on using both quantitative and qualitative data, even when people don’t understand the significance of the latter.
- Always go for specific, empirically-derived data over high-level metrics. You will be better served by data that points to / answers very specific questions in design so that you can determine efficacy of specific corrective actions.
- Do not measure success in blanketing terms. In a web environment, there are many variables, which means that even returning visitors (following the above example) may have varying needs compared with new visitors. In addition, new visitors from email marketing efforts may have different needs and expectations compared with new visitors from social media. Consider how the goals of individual pages, subdirectories or even design elements may differ.
These ideas offer the best-case scenarios for using Big Data to inform the design process, i.e. quantitative data to identify inefficiencies and benchmarking them, and qualitative user testing to outline cause and test improvements.
Case Study: P&G
As far as Big Data is concerned, few companies have more data stores than the multi-billion dollar corporate Procter & Gamble. In addition, few companies are as far advanced in their use of Big Data analytics for product and website design decisions, which is probably why the company enjoys over 1 billion annual visitors in its 1,500+ websites.
The company uses its huge stores of big data to create consumer-oriented marketing campaigns on a global scale. They know where consumers are engaged and why; which consumers are loyal and which ones have potential for loyalty. Their conceptual marketing frameworks are built around the customer, and there are mechanisms to monitor consumer engagement at each point, from in-store to online to media advertising.
It’s easy to understand then why the brand has more than 50 global leading brands, 25 of which enjoy annual sales turnovers exceeding $ 1 billion.
If there is only one lesson to be learnt from this article on data-driven design, it would be that design layout ideas, just like any other marketing campaign aspect, should be carefully thought out and researched prior to application for any campaign. It’s not enough to just place elements on pages and see results; you must know why those results are coming in, which is the only way you will be able to meaningfully impact future campaigns.