A key tenet of UX design is that every ele­ment should incite a spe­cif­ic and desired behav­ior. What actions do most busi­ness intel­li­gence dash­boards incite? None. Sure you can drill down, but what good does that do? It cer­tain­ly doesn’t help the user rec­og­nize or solve a prob­lem.

Ahh, the ubiq­ui­tous dash­board, her­ald­ed as the great sooth­say­er of mod­ern busi­ness. But is it real­ly? Pret­ty much every busi­ness intel­li­gence dash­board I’ve seen has failed to serve its most basic and pri­ma­ry func­tion: to iden­ti­fy prob­lems and help cor­rect them. How many times have you seen a dash­board that told you where a prob­lem was or what to do about it? Prob­a­bly none. Most dash­boards show lit­tle more than sta­t­ic infor­ma­tion that tells you only that it’s work­ing. A good dash­board should do two things: iden­ti­fy trends and high­light excep­tions. But a great dash­board should also incite appro­pri­ate actions that solve a prob­lem.

Data To Action Continuum

The point of a dash­board is to help you iden­ti­fy an anom­aly and take appro­pri­ate actions. There­fore, a dash­board needs to pro­vide more than just data. It has to pro­vide action­able insights. There’s a lin­ear con­tin­u­um that ranges from data to action.

Data

Data is sim­ply a num­ber. This is the least action­able end of the con­tin­u­um. Data does not incite any action. For instance, your dash­board might show you the num­ber 12, but what does that tell you to do? This sim­ple lev­el is com­plete­ly depen­dent upon indi­vid­ual and high­ly vari­able skill lev­els of the many users. Suc­cess is as vari­able as your users’ skill lev­els.

Information

Infor­ma­tion is data with con­text, such as rel­a­tive­ness or scale. This pro­vides a lit­tle more help, but still can­not incite a spe­cif­ic behav­ior that will improve your site. This is the extent of most dash­board dis­plays and it does­n’t reduce the depen­den­cy on user skill to deter­mine what actions to take. For instance, a dash­board might show that your con­ver­sion rate is track­ing at 12 per­cent in step 3 of your 7 step con­ver­sion fun­nel. That doesn’t tell you any­thing about what to do to change that, though. More­over it does­n’t even tell you if it’s a prob­lem. Maybe 12 per­cent is actu­al­ly accept­able. But how would you know?

Insight

Insight is pro­duced by adding in addi­tion­al infor­ma­tion to help iden­ti­fy when some­thing is track­ing nor­mal­ly or not. This might include fore­casts based on cur­rent trends, indus­try aver­ages, etc. This gets you clos­er to action­able state, but still relies on the afore­men­tioned, high­ly vari­able, indi­vid­ual skills. For instance, if your dash­board dis­plays user aban­don­ment rates along your con­ver­sion flow, how do you know if those rates are a prob­lem? Con­sid­er adding com­par­i­son lines that rep­re­sent your goal and indus­try aver­ages. Then you can eas­i­ly com­pare your con­ver­sion rates to deter­mine if they are excep­tion­al or accept­able. A dash­board could even com­pare a pre­vi­ous design to a new one to deter­mine whether the new design was more effec­tive.

Action

The real goal of any data analy­sis is to iden­ti­fy a prob­lem and affect a solu­tion. Charts and graphs pro­vide sta­t­ic infor­ma­tion, but don’t sug­gest cor­rec­tive actions. This end of the con­tin­u­um requires more work to design, but ulti­mate­ly results in a dash­board that incites appro­pri­ate user behav­iors. Part of the design effort includes ana­lyz­ing poten­tial prob­lems and iden­ti­fy­ing actions that actu­al­ly cor­rect them. Exam­ple: We helped design a dash­board for phar­ma sales reps. The orig­i­nal design was pages and pages of num­bers – data. It wasn’t help­ing the reps sell any bet­ter, most­ly because sales reps are bet­ter at sell­ing than ana­lyz­ing reams of data. The desired goal of each rep was to meet their sales quo­tas in order to make their bonus­es. So we tracked pro­ject­ed sales ver­sus desired out­comes and high­light­ed if they were not going to make their bonus. We also iden­ti­fied a for­mu­la that could pre­dict sales per­for­mance based on cus­tomer vis­its and if a rep was track­ing too low, it would sug­gest that the rep vis­it spe­cif­ic cus­tomers. This design replaced pages of use­less data that required hours of analy­sis with a sin­gle page dash­board that took only sec­onds to iden­ti­fy anom­alies and cor­rec­tive actions. This dash­board design remains the indus­try stan­dard for the phar­ma­ceu­ti­cal sales domain.

Example Dashboard

Here’s a com­mon type of dash­board graph depict­ing sales per­for­mance: Look­ing at this dash­board, how does it answer the fol­low­ing basic ques­tions:

• What’s the prob­lem?
• What’s the cause of the prob­lem?
• What can be done about it?
• What action does this sug­gest?

Design Backwards

Typ­i­cal­ly, dash­boards are mere­ly a man­i­fes­ta­tion of the exist­ing data. The think­ing is, if we have the data, let’s dis­play it, with lit­tle regard to what you will do once you see it. But all you end up with is a visu­al­iza­tion of exist­ing data. A bet­ter approach is to start by iden­ti­fy­ing the desired out­come, such as con­ver­sion fun­nel per­for­mance and objec­tives. Then work your way back to deter­mine what data you’ll need to visu­al­ize that. Since trends or excep­tions are key indi­ca­tors that incite appro­pri­ate actions, the sys­tem will need to under­stand what are anom­alous trends or data. A com­mon way to pro­vide this is to define thresh­olds. When a data point cross­es a thresh­old, it is deemed excep­tion­al. There’s usu­al­ly a log­i­cal rea­son why it’s a defin­able thresh­old, and you can usu­al­ly relate that thresh­old to a rea­son and a poten­tial cor­rec­tive action. Thus, the more suc­cess­ful approach is to:

• Deter­mine what goals you are try­ing to achieve.
• Iden­ti­fy what indi­ca­tors you want to see regard­ing those goals.
• Define the thresh­olds that dic­tate an excep­tion.
• Deter­mine what actions you would take to cor­rect the excep­tion.