The analytics challenges of tomorrow…

Have vendors and plat­forms made us lazy? The status quo for most of the industry is tools designed to sell to us on how, through simply installing their solu­tions, we’ll discover insight in our data which will make our busi­nesses more success­ful.

However, we all know that it’s not that simple. There are large gaps between the processes of collect­ing data, apply­ing busi­ness context, discov­er­ing insight, and produ­cing impact­ful recom­mend­a­tions; and it’s our respons­ib­il­ity as analysts to bridge those gaps.

Frequently then, our job is to build, explore, pivot, contex­tu­al­ise, and to discover valu­able insight for our busi­nesses. So we spend our time hunt­ing for trends, react­ing to chan­ging data points, and find­ing nuggets of value in pivot tables and segments. We discover nuance and intel­li­gence, which we marry with our under­stand­ing of the organ­isa­tions we work in, inter­pret, trans­late, and share in the form of insight and action.

We’re daring explorers and adven­tur­ers, in the height of the digital renais­sance.

Tomor­row, everything changes

The threshold of cost, effort, and complex­ity to tag, collect, store, process and integ­rate will continue to reduce, and busi­nesses will increas­ingly connect their systems into data ware­houses. The level of complex­ity and scope of ‘big data’ that this will create in even small organ­isa­tions will be incom­pre­hens­ibly chal­len­ging to analyse.

And so, we’re about to hit a tipping point. It’ll become, compar­at­ively, woefully inef­fi­cient for a human analyst to explore data in the way we currently do. The idea of open­ing up your analyt­ics pack­age and going hunt­ing for insight will seem laugh­ably naive, in retro­spect, when you consider the capab­il­it­ies of machine learn­ing processes which will analyse our entire data sets to discover trends we’d never see, and make recom­mend­a­tions we’d never consider.

We’re not quite there yet, though

Today, systems like Google Analyt­ics’ Intel­li­gent Alerts are still gener­ally limited to alert­ing you on stat­ist­ic­ally signi­fic­ant outliers; things like “Your pageviews-per-session for visit­ors from Italy on mobile devices was up 500% on Tues­day!”. If I don’t market in Italy, or if pageviews aren’t an impact­ful metric for my busi­ness, this might not be insight, and argu­ably, even useful inform­a­tion. The system lacks the busi­ness context it needs in order to make useful recom­mend­a­tions.

The next gener­a­tion of track­ing solu­tions, then, will begin to enable us to define busi­ness scen­arios or criteria and our preferred actions. We’ll build rules like “if it’s rain­ing whilst the consumer is brows­ing our holi­day website, we should increase the propensity to show them summer holi­day offers“, or “if sales volumes are decreas­ing, prior­it­ise the volume of sales over direct revenue”. The systems will have access to the relev­ant data and will identify oppor­tun­it­ies based on our criteria, as well as trig­ger­ing direct action in connec­ted systems. They’ll also learn from these scen­arios and become increas­ingly self-managing when it comes to identi­fy­ing desir­able outcomes and present­ing relev­ant insight and oppor­tun­it­ies.

We’ll become teach­ers, tutors, and ment­ors, managing and under­stand­ing the data sets and inputs, and letting the machines do the heavy lift­ing. As this revolu­tion unfolds, we’ll find ourselves becom­ing curat­ors of systems and data sets, and the amount of time we spend doing actual analysis will reduce.

And so, the analysts of tomor­row will spend today think­ing about how they define success scen­arios and criteria. They’ll be consid­er­ing how they codify busi­ness object­ives and commer­cial context into rules, and plan­ning out how they manage, get access to and connect the data sets they’ll need.

First published on Inked on 13th Septem­ber, 2016