The German federal constitutional court ruled, in 2009, that elections had to have a public nature. EasyVote, a promising hybrid electronic voting system for conducting elections with complex voting rules and huge ballots, meets this requirement. Two assumptions need to hold, however. The first is that voters will verify the human-readable part of the EasyVote ballot and detect discrepancies. Secondly, that electoral officials will act to verify that the human-readable part of the ballot is identical to the machine-readable part, and that they, too, will detect discrepancies. The first assumption was tested in prior work, so in this paper we examine the viability of the second assumption. We developed an EasyVote tallying component and conducted a user study to determine whether electoral officials would detect discrepancies. The results of our user study show that our volunteer electoral officials did not detect all of the differences, which challenges the validity of the second assumption. Based on these findings we proceeded to propose two alternative designs of the EasyVote ballot: (1) In contrast to the original EasyVote ballot, the human-readable part highlights only the voter's direct selections in orange, i.e. votes that are automatically distributed by selecting a party are not highlighted; (2) The second alternative includes only the voter's direct selections and highlights them in orange. Both alternatives reduce the number of required manual comparisons and should consequently increase the number of discrepancies detected by election officials. We evaluated both alternatives in an online survey with respect to ease of verification and understandability of the cast vote, i.e. verifying that the human-readable part contained the voter's selections and understanding the impact (distribution of votes) of the corresponding selections. The results of the online survey show that both alternatives are significantly better than the original EasyVote ballot with respect to ease of verification and understandability. Furthermore, the first alternative is significantly better than the second with respect to understandability of the cast vote, and no significant difference was found between the alternatives with respect to ease of verification of the cast vote.