Showing results for 
Search instead for 
Did you mean: 
New Member

Interpreting Key Influencer Visual

Hi all,


Hoping someone can help me interpret the Key Influencer visual below, created in Power BI Desktop (Nov. 2019 release):

I'm using a Kickstarter dataset (~350k rows), to determine key drivers for project success. When I analyze project_state, explained by subcategory, it shows "Literary Journals" as the top influencer at 2.77x.
There are two things that confuse me about this:
  1. There are many other influential subcategories with a higher volume of projects AND a larger proportion of successful projects (like "Festivals" for instance, which is listed at only 1.89x)
  2. Based on Microsoft documentation, the 2.77x factor should be calculated as the ratio of % Successful for Literary Journals (48.33%) compared to the average (36.06%), which is clearly not the case here.


I understand that this is driven by an underlying regression model and that statistical significance likely plays some role here, but I just can't wrap my head around these results...



Resident Rockstar
Resident Rockstar

Well, the way I look at this is the following (maybe I am wrong, and if so please correct me). I don't know the math behind this, but this is what I learned about statements as 'is x times more likely to have y in case of z) Statistically, it normalizes other variables around it so that it truly can be compared based on the specific variable you are looking at (in this case, subcategory). 

Think of it as this example, a table of festivals:


If you don't look at the whole dataset (all columns), you might draw statements like 'You have much more visitors when it rains'. Obviously this is wrong, these festivals (C and E) might be just very large festivals that even with rain have a higher number of visitors. So you take into account the Revenue and correct for that. Now a better conclusion will be "You have more revenue per visitor when it is sunny", which suddenly makes sense. But you can't say that by looking at absolute numbers or subsets of the data.

So, when you say 'with a higher volume of projects AND a larger proportion of successful projects', that doesn't mean yet that this give a higher chance on success, if other factors combined (and combined over all other subcategories) are playing a role.

I hope this makes sense, it's late where I am 😉


Kind regards



If this answered your question, please mark it as the Solution. This also helps others to find what they are looking for.

Keep those thumbs up coming! 🙂

Did I answer your question? Mark my post as a solution!

Proud to be a Super User!

Key Influencer visual do balanced down-sampling  to 10K, i.e., if you have 130K TRUE target and 140K FALSE target, it will only take 5K TRUE and 5K False due to the performance concern for interactive visual and platform limitation (cannot afford to train too large input dataset with ML models).


They adjust the lift with the positiveRatio = subsetTrue/totalTrue, and negativeRatio. Yet it is still likely to have the discrepency on lift because sampling is just a subset of the total population. 


In particular for sparse features, e.g., your subcategory may have a lot of distinct categries, and Literary Journals is one of them with small portion of input, the sampling may not work that well.  

Helpful resources


Experience what’s next for Power BI

Join us for an in-depth look at the new Power BI features and capabilities at the free Microsoft Business Applications Launch Event.


Power BI Women

Join our monthly meetings and learning sessions.



We are excited to announce the Power BI Super Users!


The largest Power BI virtual conference

100+ sessions, 100+ speakers, Product managers, MVPs, and experts. All about Power BI. Attend online or watch the recordings.

Top Solution Authors
Top Kudoed Authors