Self-service BI is often an oxymoron evidenced in the dashed dreams of disappointed business users. 

Business users who, tired of waiting for central IT to reconfigure complex semantic layers, used revenue budgets to license small projects with Qlik and Tableau. Attracted by the promise of self-service BI, data discovery, In quick time they present beautiful visualisations in impressive dashboards and everyone believes the new insights.

Then other business leaders license similar products and get different results and they start arguing about "who has the right data?"  Which highlights the weakness of many "self-service BI" claims.

Business, operational and front-line users rarely have the time or skills to source relevant data never mind prepare it. Even if they could, they are prone to the mistake of mistaking cause and effect. Drawing false conclusions from incomplete, variable quality data. 

Business leaders throughout an enterprise are not known for shyness nor for giving up a pet idea just because analytics contradicts it.

Then when they try and scale this across business units and enterprises it just compounds the issues.

Data issue is key and curated data is one solution. I tend to the strategy that data and IT experts ensure that data is relevant, timely and accurate and then unleash self-service users to explore. Then you may avoid  "Failing to see desired ROI from analytics & BI?"

Unless this is tackled self-service BI is a misnomer if not a lie