We recommend these minimum procedures for every county using optical scan voting equipment, to ensure accountability in its voting process.

Emergency Measures to Protect the 2004 Vote Count These measures are a crucial protection for this election. They will help to ensure that votes are counted correctly and will protect counties from potential legal challenges to the election.
a.	All votes for federal offices should be counted by hand, in public view, at the polling place before they are transmitted or transported to a central facility.
b.	All precinct totals, whether manually or electronically tallied, will be posted at the polling place.
c.	Ballots may be counted and accumulated electronically, in addition to the hand-count. However, in any case where a discrepancy exists, the hand-count will be considered the official result.
Election Transparency Election observers: All election processes -- including testing, tabulating, and auditing -- should be open to public viewing by both party members and members of the public. All observers must wear a name tag which is visible at all times and identifies them and their employer.
Software and Ballot Definition Data Assurance
a.	Ballot programming: All ballot data should be prepared by county officials, not by the vendor. Here's why.
b.	Pre-election testing: Test ballots for the Logic and Accuracy tests should be prepared by the county, not by the vendor. Logic and Accuracy tests should be conducted on all ballot styles. There should be a sufficient quantity of test ballots of each ballot style to test multiple different combinations of votes, overvotes, undervotes, and blank ballots. Poorly completed ballots, like those real voters will submit, should also be included. Observers of the Logic and Accuracy tests should be allowed to add test ballots to the test deck. Pre-election testing should be a public process! This means that the details and rationale of the tests must be disclosed, the testers should make themselves available for questioning prior to and after each testing session, representatives of the parties and campaigns must be invited, and an effort must be made to make space for additional members of the public who may wish to observe. This requires that testing be conducted in facilities that offer both adequate viewing areas and some degree of security.
c.	Central tabulator security: The central tabulation computer should never be used alone by anyone. It should be password-protected such that two people are required to gain access to it. Vendor employees should not touch the tabulator during a live election, nor should any vendor-supplied programming card or update be used on the machine during the course of the election and the count.
d.	Memory card failures: If any memory cards fail during the election, the ballots in the relevant precinct should all be scanned again using a working memory card.
Testing as the Polls are Opened Observers, both partisan observers and members of the public, must be able to observe all polling place procedures, including the procedures for opening the polls.
a.	Machine setup: HAVA requires that all voting systems offer voters the opportunity to change a vote that has been incorrectly recorded. All precinct-count machines should be set up to reject or warn of improperly completed ballots.
b.	Configuration confirmation: Before opening the polls, every vote tabulation system should be checked to ensure that it is still configured for the correct election, including the correct precinct, ballot style, and other applicable details. The machine's configuration is usually displayed or printed when the system is powered up.
c.	Zero verification: Verify that the ballot box is empty. Print a zeros report to verify that the ballot tabulation system has all zeros.
Results Comparisons
a.	Posting results: The results from each polling place should be printed and posted before they are transferred to the county for central tabulation. Copies should be made available to the press and posted at the polling place where they are available for public viewing.
b.	Double-checking results: Printed election results from each polling place should be compared to the corresponding results from the central vote count and audited to make sure these numbers match.
Turnout Comparisons
a.	Comparing voters and ballots: The number of voters who sign into the poll book at each polling place should be compared with the number of votes cast at each polling place and posted along with the results.
b.	Physical comparison: This audit should be made and recorded in a physical, tangible, and human-verified poll book rather than an electronic or self-verifying poll book.
Absentee Ballot Auditing
a.	Tracking ballots: The number of absentee ballots received from the U.S. Post Office should be tracked and made available for public viewing.
b.	Comparing ballots-received with ballots-counted: This number should be compared with the number of absentee ballots counted. Any discrepancy should be investigated.
Ballot Auditing Equipment testing does not displace the need for outcome auditing. Testing is necessary but not sufficient for a well-run election. Testing is never perfect, as it can overlook certain factors or interactions that may be easier to detect in hindsight. Systems interact with each other in unpredictable ways, often impossible to detect in a reasonable battery of tests...Outcome auditing can confirm the validity of testing. Voting, Vote Capture & Vote Counting Symposium. "Electronic Voting Best Practices Summary," page 23. June 2004. Kennedy School of Government, Harvard University. These procedures are recommended in addition to the emergency measures defined above. Since a recount of ballots of one ballot style may not detect ballot programming errors in another ballot style, the machines and the ballot programming should be audited as follows:
a.	Audit by manual count and comparison: A manual count of all races on the ballots, including absentee ballots, in 5% of the precincts selected randomly with even distribution across districts. This audit is comparable to auditing financial records by inspecting 5% of them. In fact, a complete audit would entail auditing one precinct of every ballot style, which would entail a much larger percentage in most elections. While this small percentage (5%) is not sufficient to audit every ballot style, it is likely to expose any gross errors in the results. Note: California law currently requires a full manual recount of 1% of the precincts in each county (or one precinct, whichever is greater). In addition, it requires a manual recount, in one precinct, for each other race not included in that 1% recount. This method is a reasonable alternative.
b.	Handling discrepancies: A difference of four votes in a single precinct or 2 votes on the same race in 2 or more precincts constitutes a discrepancy. Procedures and guidelines must be developed to deal with discrepancies. We recommend the following minimum procedures: (a) The manual recount should be the official tally. (b) The questionable ballots should be scanned through a different machine to see if the results match the manual recount. (b1) If they do match, the first machine is probably malfunctioning. To double-check, a manual audit should be conducted of the questionable races in three additional precincts (if that office or issue appears on the ballot in other precincts). (b2) If they do not match, all votes for the questionable races should be counted by hand for all precincts, and these manual counts should constitute the official tally. In addition, an immediate investigation should be launched to determine why the L&A testing did not detect the error, and steps should be taken to ensure that future tests are more robust.