The objective is to create a robustly tested classification / navigation based on top tasks data. To do one level of a classification for a large, complex website can take about 8 weeks. Prices vary from €15,000 to €25,000. We can also train you on the method, which will significantly reduce costs.
The following steps are involved in Top Tasks Classification:
- Take the top tasks (those that have received the first 50% of the vote) and other critical tasks, which might include:
- Top tasks of a key sub-category or demographic
- Tasks that have a very high management priority
- Create a list of tasks that is ideally no more than 20 and with a maximum of 25.
- Get 50-100 customers to sort these tasks into their preferred classifications.
- Identify classification patterns that emerge and then create a hypothetical Level 1 classification.
- Create between 10 and 15 task questions and then ask 30-50 customers where they would click on the hypothetical classification if they were trying to solve these tasks.
- The target is to achieve an 80-90% success rate.
- Usually, after the first round we have a success rate in the region of 60%. We discuss the results with you, agree changes we need to make to the classification and test again.
- Typically, it takes three rounds of testing to achieve an 80-90% success rate.
- Repeat for lower levels in the classification / navigation.
Sorting Top Tasks into classes
The first step is getting customers to sort the top tasks into classes. (We use Optimal Workshop for the sorting and voting, and have designed our own software to do the analysis.) The following diagram shows how it works.
Customers drag tasks from the left column into the center, create class groups and then name these groups. Once we get about 50 people to do this, class grouping trends begin to emerge.
The following chart shows the type of data we can get. Here, 74 people sorted top tasks for a health website. We can see, for example, that 50 of the 74 grouped “Disease prevention” and “Disease causes” together. This sort of analysis reveals the mental model of the customer and from that we can create a hypothetical classification that we will then test with tasks.
The following chart gives a different view on how people are grouping tasks. Here we can see that “Can I treat this myself”; “Check Symptoms” and “What to do based on symptoms” are tightly grouped in the customer’s mind. “Everyday minor ailments” is reasonably associated with these tasks, and “Basic facts about conditions / diseases” is more loosely associated.
Testing the top tasks classes
The next step is to test these classes with task questions based on the top tasks. The objective is to get an 80-90% first click success rate. In other words, when customers’ have a top task we want them to click on the right top level link / class 80-90%+ of the time. That’s how we judge that the classification is intuitive and useful. We will usually create about 10 task questions to test.
The following table shows the hypothetical Level 1 classification that resulted from the customer sorting process.
Example tasks that we will now test against this hypothetical classification include the following:
- You have a pain in your stomach, severe headache and your left arm feels numb. What could that be a symptom of?
- Find a doctor in your area
- Get an overview of acupuncture
We get 30-50 people to select which class they would click on in order to try and complete such tasks. Below is a sample set of results from one of the tasks.
The hypothesis was that people should click on “Everyday Health” for this task. However, only 56% of people did so. 44% of people clicked on “Health Care Services” so clearly we have a problem here that we need to solve. Testing like this reveals navigation issues that must be addressed.
The following task was for an intranet of a large technology company.
Within this company a spare part would be found in Services. However, when we tested this task—and other service-type tasks—a significant percentage clicked on Products. There was clearly a confusion between what was a product and what was a service. Once the company saw this data, they decided to merge the two classes into a new class called “Products & Services.” Interestingly, we found the exact same type of confusion between products and services in other technology companies we tested.
Another interesting class was “Local Services”. This is what we call a “dirty magnet”. No matter what type of task we tested, invariably a significant number of people clicked on “Local Services.” This class name was so vague and all-embracing its drew clicks from everywhere like a dirty magnet. In the next round, we changed the name.
We have found that typically three iterations of testing are required to get an 80-90% success level for your top tasks. You can then take this method and design out Level 2 and lower classifications.