What Goes Into A BrandGuard Score?
36 Different models make up a brand score
What does it mean to say something is “on brand?” It turns out to be a pretty challenging question if you are trying to teach a machine to figure that out.
The interesting thing about branding is - machines can’t really get better than humans because the humans define what the brand means. Training BrandGuard models is about showing the models what we think is “on” brand or “off” brand and letting them learn.
There are 4 main categories of models that BrandGuard uses.
SFW Models - These models check content for nudity, violence, and other socially inappropriate material. These models are mostly open source but are fine tuned per customer because, for example, Victoria’s Secret and Ann Taylor may have different ideas about what is safe for work.
On Brand Models - These models are deep learning models trained on tens of millions of branded assets and fine tuned with a few dozen examples of a particular brand. The way they work is illustrated here.
Style Conformance Models - These models are built off your style guide. We ingest it and build models for brand tone and voice, logo, font, color, spacing, etc. These models are more classical machine learning models that check for clear violations of style guide rules.
Regulatory and Compliance Models - This is the fastest growing segment of BrandGuard models. We currently check for trademark and copyright violations in your content but, are engaged with customers in industries with other marketing regulations. Stay tuned for models that score content on compliance with rules for consumer finance, healthcare, pharma, and more.
These models are run on every piece of content produced by your team, and then run through a decision function which learns how to weight each model based on what your team considers most important. We learn all of this from you telling us what is on-brand, and what isn’t.
Brand consistency has been shown to increase revenue significantly. If you want to gain the benefits of brand consistency, BrandGuard can help. Please contact sales@brandguard.ai to see a demo.