Agile web development and the wisdom of crowds

by Mechaferret on November 23rd, 2009

In my previous post, I discussed how hard it can be to do the feedback loop for agile development well when the product you are developing is a general-audience web application.

There is one way, however, in which web applications offers a distinct advantage over other types of applications for agile development: measurement. Measurement is a key part of the feedback cycle for lean processes, but finding the right metrics to optimize value for your customers can be challenging. Web applications, however, can give you incredible amounts of data about what your users actually do: how they reach your site, what they do when they get there, whether they continue an activity or drop off. Everything they do is done through the web application, so everything they do can be retained for further study.

Given this wealth of data, it becomes much easier than in other forms of development to form and test hypotheses about what your users want. You can do “A/B” testing, randomly offering them different implementations of the same feature, and seeing which one does better in achieving your desired goal. Or you can do simple analytics: change a feature, and see how much your desired goal changes. You can also easily collect application-specific data, e.g., which products seem to get purchased together, what attributes of a customer make them likely to order the high-end version of a product, the relationship between followup contact with a customer and repeat purchases.

You get to use your entire user base to estimate the value of a given feature for your desired target audience. Because these users are part of your target audience and because of the effect of the “wisdom of crowds”, their collective evaluation will be far better than the a priori evaluation of your product team.

Of course, I’m glossing over the difficulty of figuring out in sufficient detail what your goals are and how to measure them. But often the goal is obvious and the results of the basic data can be compelling. We recently experienced this on a web application I’m currently working on. We were having a debate about exactly how much information to request on an initial inquiry form. Some of the team wanted to request some information on the form that would be of use for a possible future product (but which also could be collected elsewhere), while others wanted to keep the form’s size to a minimum. The application on production at the time of the debate was collecting the data. In the next release, we stopped collecting the data, and the inquiry completion rate immediately doubled and stayed there. After that, there was no more need for debate about collecting the extra data in the form.

From Methods

Leave a Reply

Note: XHTML is allowed. Your email address will never be published.

Subscribe to this comment feed via RSS