EDIT: Accidentally posted while still typing and reading, aaah, this is unfinished.

EDIT2: Okay, this is as done as I will do it, I also looked at the clock, and I won`t be awake for long now anyway.

EDIT3: Okay, this one is the last one, I really have to get to bed, because I am also noticing how diving into this is not good for my health. But turns out you can disregard the stuff I wrote below except for the last sentence, and me still thinking the data is more ambivalent narrative-wise. But while I still maintain the language was confusing, I finally noticed an unambivalent line from the survey: "Of the following issues, choose any that played a role in your [vote for presidential candidate/decision to not to vote for president]. Check all that apply." So, yes, this was indeed also non-voters.

I admit, now that I explicitly checked, that is also how I would interpret (from the PDF):

This survey is based on 604 interviews conducted by YouGov on the internet of registered voters who voted for Joe Biden in 2020 and not Kamala Harris in 2024.

But it is also just ambivalent enough to create questions when combined with the language of the article: Since both the study and the article seem to be by the same institute, I doubt it's a miscommunication error. The language of the article is repeatedly so specific.

For Biden 2020 Voters Who Cast A Ballot For Someone Besides Harris

29% of voters nationally who voted for Biden in 2020 and cast a ballot for someone besides Kamala Harris

When Biden 2020 voters cast a ballot for someone besides Harris in 2024

And then there are questions in the PDF like:

Next, think back to how you voted in 2024. You will see issues that some say may have impacted their vote. For each of those, please say how you feel about that issue.

That seem to indicate that this indeed only targeted people that did vote.

So, colour me genuinely confused, it seems like such a specific and deliberate usage of language. And I have to admit, it feels weird to me, especially considering the IMEU has an interest in making Gaza the most important topic. Note that the same numbers of the survey could also be used to support different narratives, like: 68% said abortion access was important to them and influenced how they voted in 2024, vs 27% saying the same about violence in Gaza. Or Question 12 vs 13, showing that on a policy difference exclusively on Gaza, the people surveyed would still predominately support the Democrat, and only 8% mention not voting if the Democrat supports Israel unconditionally. So, this also does not fit the narrative neatly.

But if this does indeed represent non-voters as well, and one third of those truly did not vote because of Gaza, yes, that is indeed a large enough group to swing close results in battleground states.