Blog article

Accuracy: Why I Hate That Term

Posted on
July 31, 2019
In
Read article
Accuracy: Why I Hate That Term

Every other week we have a lunch and learn at Recurve. Topics have ranged from the zen of surfing to D&D strategy. Last week, McGee Young took things in a more serious direction, leading a Socratic discussion that delves into, and debunks, many of the key terms of art in the M&V industry. 

While we typically keep our lunch and learns in-house, we felt that in this case the topic is relevant to the broader industry—and represents a discussion and debate that needs to happen.

What is accuracy?  What is precision?  What is the true value we in the energy efficiency industry are pursuing, and how would we know it if we saw it?

These seemingly rhetorical questions are big issues when one is calculating baselines for measurement of savings. In order to figure out the value of an intervention, we need to know what would have happened if we didn’t actually make a change in the building.  But we did change the building… so what exactly is the true value, and if we don’t have a true value, how do we know who is right?

M&V as we know it often mistakes accuracy for precision and a good R2 with truth, when in fact the true value of what is done to a building may vary greatly depending on the use case.

‍This discussion has relevancy anywhere that building level M&V or a program level EM&V report is being conducted, but it is even more important in states like California that are rapidly moving toward pay-for-performance based approaches that leverage Normalized Metered Energy Consumption (NMEC). In particular, we must take into account the current role and cost of subjectivity in our calculations. Even if adding new variables can achieve a better R2 value, doing so results in bespoke answers that can’t be predicted or aggregated—and therefore lead to greater uncertainty.

For these reasons, we don't think such tests are themselves sufficient to determine how models will perform in the real world (even though the Recurve Platform performs well on modeled testing).

Food for thought everyone!  

We hope this is the start of a fundamental discussion that is overdue. 

Want to find out more about how a new approach to M&V can improve programs, set the stage for pay-for-performance and turn efficiency into a procurable resource? Contact us.

Loading...
Support Center
Contact Us