No idea. From the Guardian article:
So it seems there was actually a fairly major methodological problem with that particular study - if you ask a load of people who are statistically much more likely than average to be Labour voters what they think of Labour, it doesn't tell you much. Maybe Zahawi didn't want the company he co-founded to have its reputation damaged by a poor prediction even though he no longer worked there?
Conversely, this Curtis guy is now apparently working for Opinium, which is a rival polling company, so if he left YouGov on bad terms then he may be exaggerating what happened out of sour grapes. [Edit: although even he's saying it wasn't a deliberate attempt to harm Labour's chances.]
Maybe, or maybe they're just trying to cover their arses. They offer no explanation as to the methodological problems.
Fair point re: Curtis, but it could just as easily work against him. His new employer may now worry he might kick up a fuss about them at some point too.