We have have built a complete web analytics application (internal) for tracking Yahoo! products. Being the only designer for the product I got the chance to work on every BIT of it; from each and every component to every pixel. I can tell you its quite an experience handling such a big product on our own - that too when you are not only the interaction designer but also user researcher, product designer and visual designer. I'm not boosting but 'cribbing' :(
With so many flows and interconnections it become overwhelming. During design we fix these complex flows for most ideal conditions. The design works flawlessly; but then came the surprises - ERROR handling.
The product behavior was built such that the settings can be transferred. To give you an example, you are looking for mail data; then you put up certain segmentation settings like show PVs from India with age group 10-15 with users as male. Now you click on another metric like time spent - the rest of the setting of age, country & gender is transferred. The behavior is to change only the setting which user chooses to change.
This on paper looks flawless work well iin concept but then comes the conditions - in some cases it doesnt work. So this behavior works for most cases (lets say 90%) but in rest cases it throws up error.
Designers’ dilemma
Should I change the behavior to make sure system gives less errors (or conditions of no data found) or to provide a way which give most value?
Well, I choose option 2. Most use cases will work perfectly and give tremendous value, as the user settings are seamlessly transferred. I am providing value for 90% cases (this number is rough assumption) and spoiling the experience for rest 10%. But I guess its worth the risk; as it seems to be working :)
But then I had to figure out ways to make sure those 10% cases are also taken care of. This may not sound to be perfect, but we 'redirect' users to conditions that give data - of course with user consent. We show them the options of redirect.
That’s the best we could do for now.
But one thing I realized. Error conditions can some times spoil the party!
Your perfect design can be derailed with these fringe cases. But that what makes a complete design isn't it?
Friday, September 11, 2009
Pain of handling 'Error condition' & designers' dilemma
Monday, September 07, 2009
Too soon a conclusion
There is something that i have observed around me that has motivated me to write about it.
Data analytics is not very natural to designers; its not a part of design training. I see people around me claiming data without understanding them. Some key points to look out for -
Number is just a "number"
Number is just a number; it doesn't convey anything without other supporting information. To give you and example - So if you say my site got 1000 Page Views the question is "What does it mean?". Nothing unless you say it got a 10% increase. So others know if things are going up or down.
Understand the Scale of the number
Secondly, understand the scale 10% of 100 is 10; 10% of 100000 is 10000. Understand this sense of scale in the number.
Number in Context
Don't make conclusions with one metric. Look around and see the other behavior. Eg. Site A is having a 10% rise in PVs but the UUs is going down by 20%. What does it mean? less users are giving you more PVs. But why are you getting less users? This can't be explained by data alone, which takes us to the next point.
Support Metric with other methods.
The data can tell you PVs are going down by "why" is a question that lies beyond data analytics.
There is a lot that data can tell you but you need to triangulate the data with other methods like doing user study (quantitative methods). There is lot that might have gone in the offline world that has influenced the performance - eg you might have run an Ad or might have done marketing. Or some event that has triggered a behavior. So keep that into account.
Where is your traffic coming in and what are the clicking or going is what primarily you might want to look at in the beginning. Next you might want to slice and dice data to look into detail - eg. you might want to segment the traffic based on age, gender, country etc.
These are some basic pointers for designers to get started on analytics. Keep in mind 'dont jump to conclusion too soon; you might land up with the wrong analysis or may only see just one aspect of the problem and may miss out other obvious stuff.'
Data analytics is not very natural to designers; its not a part of design training. I see people around me claiming data without understanding them. Some key points to look out for -
Number is just a "number"
Number is just a number; it doesn't convey anything without other supporting information. To give you and example - So if you say my site got 1000 Page Views the question is "What does it mean?". Nothing unless you say it got a 10% increase. So others know if things are going up or down.
Understand the Scale of the number
Secondly, understand the scale 10% of 100 is 10; 10% of 100000 is 10000. Understand this sense of scale in the number.
Number in Context
Don't make conclusions with one metric. Look around and see the other behavior. Eg. Site A is having a 10% rise in PVs but the UUs is going down by 20%. What does it mean? less users are giving you more PVs. But why are you getting less users? This can't be explained by data alone, which takes us to the next point.
Support Metric with other methods.
The data can tell you PVs are going down by "why" is a question that lies beyond data analytics.
There is a lot that data can tell you but you need to triangulate the data with other methods like doing user study (quantitative methods). There is lot that might have gone in the offline world that has influenced the performance - eg you might have run an Ad or might have done marketing. Or some event that has triggered a behavior. So keep that into account.
Where is your traffic coming in and what are the clicking or going is what primarily you might want to look at in the beginning. Next you might want to slice and dice data to look into detail - eg. you might want to segment the traffic based on age, gender, country etc.
These are some basic pointers for designers to get started on analytics. Keep in mind 'dont jump to conclusion too soon; you might land up with the wrong analysis or may only see just one aspect of the problem and may miss out other obvious stuff.'
Subscribe to:
Posts (Atom)
