10 — Operational Practice
This section is about the people side of Actual Vote. The previous sections describe what the system is (Section 4), what’s legal (Section 6), how to access poll tapes (Section 7), and how an analysis works (Section 8). This section describes who actually does the work, how they organize themselves, what their best practices are, how they coordinate with America Counts, and what to do when things go wrong.
The structure is built around the recognition that there is no single “Actual Vote user” workflow. Different people use the tool in very different ways. A lone citizen recording a single poll tape on election night has different needs than an organized video-recording team covering an entire county, which in turn differs from a full Audit Committee running the whole pipeline, and from a paid worker under contract. The best practices for one configuration don’t necessarily transfer to another.
The structure also recognizes several cross-cutting concerns that apply to all users regardless of configuration: safety and privacy, troubleshooting, communication with AC, the rhythm of the electoral cycle, coordination protocols, and lessons learned from prior cycles.
A reader planning their own Actual Vote work should start with Section 10.9 (The Electoral Cycle) to understand the timing, then read the subsection for their user category, then read Section 10.6 (Safety and Privacy) and Section 10.7 (Troubleshooting) before going into the field.
10.1 — Lone User Operations
What a lone user is
A lone user is a single person submitting one or a few recordings, typically without coordination with others. This is the most common user configuration and the one Actual Vote was originally designed for. The “Actual Vote Hero” framing on the website describes this path: download the app, practice with the practice poll tape, go to a polling place on election night, record the posted tapes, submit, and you’re done.
Typical scope
A lone user typically records at one polling place or a small number of nearby ones — the polling place where they vote, a couple of others on the way home, maybe a few more if they’re motivated. The yield is modest: one to five transcribable submissions per user per election. But modest individual yield aggregates into meaningful coverage when many lone users participate across a jurisdiction.
What the lone user experience looks like
The simplest version: On election night, after voting, the user goes back to their polling place (or stays nearby) and waits for the tapes to be posted. This can take anywhere from thirty minutes to several hours after polls close, depending on the jurisdiction and the polling place. Once the tapes are posted, the user opens the Actual Vote app, records the tape by slowly panning the phone camera from top to bottom, confirms the upload, and leaves. The entire active recording takes two to five minutes per tape.
The practice poll tape — a downloadable PDF that shows what a real poll tape looks like — is the recommended warm-up. Recording the practice tape at home gives the user a feel for framing, speed, and focus before the stakes are real. The app includes a practice mode for this purpose.
What support lone users get
Lone users can contact AC at any time by email (j.flatley@democracycounts.org or info@americacounts.us). During election week, AC monitors email closely and responds as quickly as capacity allows. Lone users can also check the My Actual Vote Videos tab in the app to see whether their submissions have been received and processed.
AC does not actively coordinate with most lone users — the lone user path is designed to work without coordination. But users who reach out in advance get the benefit of AC’s knowledge about their jurisdiction: whether tapes are likely to be posted, what to expect at local polling places, and any known issues.
Realistic expectations
Not every lone user submission will be transcribable. Some recordings will be too blurry, too dark, or from angles that make the numbers illegible. Some polling places won’t post tapes. Some users will encounter obstacles — locked doors, removed tapes, confrontations — and won’t be able to record at all. AC’s pipeline is designed to work with whatever submissions arrive; the user should not feel that a failed recording attempt was wasted effort. Even a context submission documenting a missing tape or an uncooperative polling place produces useful information.
The most important thing a lone user can do is submit whatever they have. A partially legible tape is better than no tape. A recording of where the tape should have been posted but wasn’t is useful context. AC’s vetting and transcription pipeline handles the quality sorting; the user’s job is to capture what’s available and send it in.
10.2 — Video-Recording Team Operations
What a video-recording team is
A video-recording team is an organized group whose purpose is to video-record poll tapes and submit them to America Counts. The team takes on the field work — the recording itself, the logistics of getting to polling places, the on-the-ground coordination — and hands the submissions off to AC for the analysis pipeline. The team’s job ends when the submissions are uploaded and confirmed; AC’s job begins from there.
Teams can range widely in scale and structure. A loose team of friends might be two or three people covering their neighborhood polling places together with informal text-message coordination. A local civic group — a neighborhood association, voter protection group, or election integrity organization — might organize a few volunteers to cover a city or part of a county. A coordinator-led team has a single coordinator who recruits and schedules unconnected volunteers (sometimes paid, sometimes not) to cover a larger geographic area. A multi-team effort has several teams working in different parts of a state in the same election, with cross-team coordination handled by a state-level lead.
The key feature of all of these is that the team handles the recording and AC handles the analysis. This division is deliberate. Centralized analysis is more consistent (the same methodology is applied every time), benefits from accumulated jurisdiction-specific expertise, and carries AC’s reputation rather than the team’s. For most groups, the video-recording team model is more sustainable than the Audit Committee model (Section 10.3), which requires the full analysis workload.
How a team forms
Teams form through partner organizations (Scrutineers, Common Cause, League of Women Voters, local voter protection coalitions), around a candidate or campaign concerned about vote reporting in their race, spontaneously around a local concern about election integrity, or through AC outreach when AC identifies a jurisdiction that would benefit from coverage.
In all cases, the early-stage work is similar: identify the people, identify the goal (which election, which jurisdiction, which polling places), identify the leadership (who’s coordinating), and make initial contact with AC.
Geographic coverage planning
For teams with multiple people, coverage planning is the most important pre-election work.
Identifying the target polling places. Get the list of polling places in the team’s target area from the county elections office or its website. The list includes polling place names, addresses, and the precincts that vote there.
Prioritizing. A team usually cannot cover every polling place in a county. Prioritization criteria include coverage redundancy (two members covering the same place provides backup), geographic clustering (a member who can cover several nearby places in sequence is more efficient), high-stakes contests (contested precincts are higher priority), historical issues (places that have had problems before warrant re-verification), and AC’s guidance about which precincts matter most for the analysis.
Routing. Plan the order in which each team member will visit their assigned places. Consider drive time between locations, expected closing times, and typical posting delays. The RouteXL tool has been used to optimize routes for teams covering many locations. The GA2024 debrief identified route assignment as one of the most time-consuming pre-election tasks for coordinator-led teams — a coordinator assigning ten-plus locations to each of several paid workers using Google Maps took at least thirty minutes per worker and was error-prone. AC is working toward centralized tooling to make this easier.
Backup assignments. Each team member should have a backup plan if their primary assignment doesn’t work out — the polling place doesn’t post, the member can’t make it, the tape is illegible. Overlap between members’ assignments is a feature, not a bug: it provides redundancy.
Communication plan. Agree in advance on the communication channel for election night — group text, Signal, Slack, phone calls — and test it before election day.
Team logistics
Beyond coverage planning: equipment (each member needs a phone with the app installed, a backup recording method, a charged battery, and ideally a portable charger and flashlight), vehicles and drivers (essential in suburban and rural areas), meeting points (where the team will gather before, during, or after the recording effort), food and rest (election night recording often happens after a full day and can extend into the early morning), and identification (personal ID in case anyone asks who you are and what you’re doing).
What to do on election night
The protocol for recording night: start early (be at the first polling place before closing time), wait for posting (don’t enter restricted areas), find the tapes (typically on a wall, window, or door visible from public space), approach from a legal vantage point (public sidewalk or accessible area), record carefully (frame the tape clearly, move the camera slowly so each value is captured in focus, cover the full tape from header to footer), confirm the upload, move to the next polling place, and communicate with the team and AC if anything unexpected happens.
What to do after election night
The team’s work continues for a few days after election night. Each team member should confirm their submissions appear in the system. The team should be available to answer follow-up questions from AC during vetting and transcription. The team should document the experience — what worked, what didn’t, what was surprising — in a short debrief. And the team should stay available for context queries during the investigation of apparent discrepancies, since AC may need to ask about specific conditions at specific polling places.
A worked example: Wisconsin Supreme Court 2023
The Wisconsin Supreme Court 2023 election effort illustrates how a video-recording team coordinates with AC end-to-end. A team of volunteers identified target polling places in advance during the pre-election phase, coordinated with AC on scope, jurisdiction-specific details, and communication channels, then recorded on election night using Method 1 (publicly posted tapes after polls close). The resulting submissions fed directly into AC’s Wisconsin analysis. The Wisconsin effort demonstrated several things: that the team model works for races where a single statewide contest drives high voter and volunteer interest, that advance coordination with AC pays off in higher-quality submissions, and that Method 1 recording by organized teams can produce meaningful coverage of a state in a single election night.
Coordination with America Counts
Coordinating with AC before the active phase is one of the highest-leverage things a team can do. Teams that coordinate well in advance produce more useful submissions and have fewer operational problems on election night.
The recommended protocol: in the off-year, email Jason Flatley to introduce the team and get on AC’s radar. A few weeks before the election, align with Jason on which election and jurisdictions the team will cover, which access methods, the team’s capacity, AC’s pipeline capacity, communication channels for election week, specific contests or precincts of interest, known issues with the jurisdiction, and backup plans. During election week, keep communication open with AC — if something goes wrong, AC should hear about it quickly. After the election, debrief with AC on what worked and what should change.
10.3 — Audit Committee Operations
What an Audit Committee is
An Audit Committee is a formally organized local group that does the entire Comparison Analysis pipeline on its own — video-recording, transcription, comparison, investigation, and reporting. Where a video-recording team’s job ends when submissions are uploaded, an Audit Committee continues through the full pipeline. The committee typically has its own coordination structure, its own analyst or analysts, and its own publication channels, though it may also publish through AC.
The canonical example is Citizens Audit Broward (CAB) in Broward County, Florida. In 2020, CAB organized a full pipeline analysis of two elections using Method 3 (in-office viewing at the Broward County Supervisor of Elections office). CAB members did the recording, the transcription, the comparison, and produced findings. After their first full-pipeline effort, CAB transitioned to operating more as a video-recording team, mobilizing volunteers who recorded poll tapes and submitted them to AC for central analysis. This transition is informative: even committed Audit Committees often find that the full-pipeline workload is hard to sustain and that the video-recording team model is more sustainable for ongoing engagement.
What it takes to run an Audit Committee
The Audit Committee model requires significantly more from its members than the video-recording team model. The requirements: a core group with sustained engagement over the duration of an analysis (typically several weeks from election to final report); at least one person with the technical skills for the comparison analysis (building comparison logic requires familiarity with the jurisdiction’s reporting practices, data manipulation skills, and the patience to investigate apparent discrepancies thoroughly — this is the highest-skill role and often the bottleneck); time for the work (a full analysis takes days to weeks of cumulative effort, especially during the before-certification window when the work is on a deadline); local knowledge (members who know the polling places, the elections office, the local political environment, and the relevant officials); a clear scope (what will be audited and what won’t); and a communication channel with AC (even fully independent committees benefit from AC’s accumulated knowledge and tools).
When to operate as an Audit Committee vs. as a video-recording team
For most groups, AC’s recommendation is to operate as a video-recording team. Centralized analysis is more consistent, AC has accumulated jurisdiction-specific expertise, the video-recording team model is more sustainable, and findings published under AC’s name carry AC’s institutional reputation. The cases where Audit Committee operation makes sense include: the jurisdiction is unfamiliar to AC and the committee can build the knowledge faster locally; the committee has specific expertise (legal, technical, journalistic) that AC doesn’t; the committee wants to publish findings under its own name for local credibility reasons; the committee plans a different methodological approach; or the committee is in a jurisdiction AC isn’t planning to analyze.
In any of these cases, AC welcomes coordination and is happy to share tools, methodology, and accumulated knowledge. AC wants more vote reporting audits to happen, regardless of who runs them.
How to coordinate with AC
Even an independent Audit Committee benefits from coordinating with AC. Early in the planning phase, contact Jason Flatley to describe the committee’s scope and goals, ask about AC’s experience with the relevant jurisdiction, ask whether AC has tools or knowledge that would be useful, and discuss methodology alignment. During the active phase, share progress with AC. After the analysis is complete, debrief with AC so the committee’s experience contributes to AC’s understanding of how distributed analysis can work. For publication, the committee can publish independently, jointly with AC, or through AC — the choice depends on the committee’s goals. Joint or AC-published work goes through AC’s editorial review.
10.4 — Coordination with Elections Offices
Why this matters
Coordination with elections offices — Supervisors of Elections, county Boards of Elections, county clerks, municipal clerks, depending on the jurisdiction — is one of the most important and most delicate aspects of operational practice. Elections offices are the source of the official results that AC compares against. They are the first point of contact when AC finds an apparent discrepancy. They are the gatekeepers for Case 3 access (in-office viewing, scans, physical copies). And their cooperation or lack thereof substantially affects the quality and completeness of an analysis.
How to approach offices
The framing matters. AC’s recommended approach is non-adversarial: “we want to confirm your good work.” The goal of an independent audit is not to catch the elections office doing something wrong — it is to provide independent evidence about whether the vote reporting was done correctly. Most of the time, the audit will confirm that it was. The elections office should understand that AC’s work is in their interest: a clean audit result is the strongest possible endorsement of the office’s professionalism.
The tone should be professional, specific, and respectful. “I’m with America Counts, a non-profit non-partisan organization that conducts independent audits of vote reporting. We video-record poll tapes and compare the values against the official results. In almost all cases, we confirm that the reporting was done correctly. We’d like to work with your office on the upcoming election.”
What to request
Depending on the access method, a team or committee may need to request: posted tapes (confirmation that tapes will be posted and where), results formats (how the office publishes official results — CSV, Excel, HTML, PDF — and whether machine-readable formats are available), reporting structure documentation (how results aggregate from precincts to county totals, how vote centers are handled, how early voting and absentee ballots are reported), and in-office access (for Case 3, the ability to view and record tapes at the office after the election).
Handling refusals and resistance
Not all offices are receptive. Some respond promptly and cooperatively. Others don’t respond at all. Some are actively resistant — suspicious of outside scrutiny, defensive about their procedures, or worried about the political implications of cooperating with an independent audit.
When an office refuses: don’t escalate at the point of refusal. Thank them, document the refusal, and report to AC. AC has experience with resistant offices and can advise on next steps, which may include a more senior approach, a public records request, or a decision to focus resources elsewhere. In some cases, the refusal itself is a finding worth documenting.
AC’s experience includes the full range. The Henderson County, North Carolina Board of Elections responded to AC’s inquiry the same day, provided a detailed explanation of a discrepancy at the Bat Cave precinct, and shared additional documentation. At the other end of the spectrum, some offices have been unresponsive despite multiple attempts. Both responses are documented and both become part of the public record of how election administration operates in those jurisdictions.
State and jurisdiction variations
The mechanics of coordination vary enormously by jurisdiction. In Wisconsin, municipal clerks handle elections on election day and county clerks handle the canvass — so the user may need to coordinate with the municipal clerk for election-night access and the county clerk for post-election access. In Milwaukee, tapes are printed at closing time but there is no law requiring them to be posted; the three copies are earmarked for different branches of government. A member of the public at the polling place at exactly the right time could observe the tape, but when the poll workers leave, the tapes go with them. After the canvass, someone could request to inspect records, but this must be under the supervision of election officials and cannot disrupt the process.
In Arizona, most jurisdictions use central tabulation and do not post poll tapes at polling locations. The two largest counties (Maricopa and Pima) use different voting systems (Dominion and ES&S) with different tabulation architectures. Poll watchers can observe the tabulation process at some locations, but the general public cannot access the secure areas where tapes are printed. Results are available on county websites as they are tabulated.
These jurisdictional specifics illustrate why coordination — and local knowledge — matter. The state-by-state reference (B.3) catalogs what AC has learned about each jurisdiction it has worked in.
The non-adversarial philosophy
AC’s relationship with elections offices is fundamentally collaborative, not adversarial. The Henderson County model — AC contacts the office with a question, the office responds with information, the question is resolved — is the preferred outcome. Every cooperative interaction builds a relationship that makes the next interaction easier. Every confrontation burns a bridge.
This does not mean AC defers to elections offices when it has reason to believe something is wrong. It means AC asks questions before making accusations, presents evidence before drawing conclusions, and gives offices a genuine opportunity to explain before escalating. The cooperative posture is both ethically correct and strategically effective: it preserves the relationships that make future analyses possible.
10.5 — Paid Worker Operations
When and why AC pays for recording
AC’s default model is volunteer-driven, but in jurisdictions where volunteer networks are insufficient to achieve the desired coverage, AC has deployed paid workers. The 2024 General Election deployed paid workers in Georgia, Florida, and Pennsylvania alongside volunteer networks in North Carolina and other states.
The decision to pay for recording is driven by coverage needs: when a target jurisdiction lacks organic volunteer energy, when the target area is large and geographically dispersed, and when the analysis timeline requires a guaranteed minimum number of recordings. Paying workers converts an uncertain volunteer supply into a more predictable commitment.
Compensation structures
The 2024 General Election paid workers per poll tape recorded: 6.00 in Pennsylvania (adjusted for the number of precincts per polling location — in Pennsylvania, most polling places have at least two precincts and therefore at least two poll tapes). Most of the worker’s time is driving time, not recording time. A worker covering ten locations in a night might spend three hours driving and thirty minutes recording.
The per-tape structure has trade-offs. It aligns incentives (workers are paid for output, not time), but it means workers bear the risk of unproductive driving (when tapes aren’t posted, when locations are locked, when they arrive and the tape has already been removed). The GA2024 debrief identified this as a friction point: workers who drove to locations and found no tapes felt they were working for free.
Team leaders — coordinators who manage multiple workers — received flat compensation ($200 in the 2024 deployments). The GA2024 debrief recommended formalizing the team leader role with a handbook and specific responsibilities.
Recruitment and retention
Recruitment should begin months before election day. The GA2024 debrief identified late recruitment as a recurring problem: close to election day, potential workers are already saturated with campaign requests and don’t want additional commitments. Sources of paid recruits include Craigslist postings (which produced functional workers in Georgia), partner organizations, and outreach to civic groups and student organizations.
Retention depends on clear communication about what the work actually involves. Workers who understand upfront that they will be driving at night, that some locations won’t have tapes, and that the pay structure is per-tape rather than hourly are more likely to complete their assignments. Workers who are surprised by these realities are more likely to abandon the effort. The GA2024 debrief was emphatic on this point: all instructions, pay structure information, and realistic expectations should be available as soon as recruitment begins, not released days before election day.
Quality expectations and accountability
Paid workers are held to the same quality standards as volunteers. Recordings must be clear, complete, and properly framed. PII must be avoided. Metadata must be as accurate as possible. The fact that a worker is paid does not change what constitutes a usable submission.
Accountability is harder with paid workers than with motivated volunteers. A volunteer who doesn’t show up is disappointing; a paid worker who doesn’t show up after being assigned precincts creates a coverage gap that may be discovered too late to fill. The GA2024 debrief raised this concern explicitly: a worker who accepts an assignment and doesn’t follow through leaves those precincts unrecorded, and there is no mechanism to detect the gap in real time. The debrief recommended overlap in assignments (so that if one worker doesn’t show, another’s assignment covers some of the same territory) and a real-time dashboard for team leaders to track which locations have been covered.
Lessons from the 2024 paid deployments
The GA2024 paid deployment produced several lessons beyond those already mentioned. Training materials should be available from the first moment of recruitment. The app’s precinct field should include clear instructions about what to enter (a screenshot showing what the precinct field looks like and what value to put in it). Workers doing gig-style election-night work are sensitive to friction — every unnecessary step, every confusing interface element, every unclear instruction reduces the likelihood of completion. Pay for the next-day follow-up (returning to locations to check for tapes posted overnight) should have a different structure than election-night pay, since the probability of finding tapes drops rapidly. And the per-tape payment should not apply to recordings of missing tapes — a worker should not be driving around collecting payments for documenting empty doorways, though documenting the absence remains useful context.
The aspirational category: Citizen Auditor
Beyond the operational categories described in Section 10.1 through Section 10.5, there is an aspirational category that AC has used in some contexts: the Citizen Auditor. A Citizen Auditor is an Actual Vote user who participates in the audit phase — transcription, comparison analysis, investigation, reporting — beyond just submitting recordings.
The Citizen Auditor category is aspirational because in practice most users stop at submitting recordings. The audit phase is difficult, time-consuming, and requires a level of technical consistency that is hard to maintain across many distributed volunteers. AC has encouraged users to step into the audit phase with mixed results. The Citizen Auditor framing appears in the Privacy Policy and other AC materials as a way to invite users into deeper engagement with the work.
For this manual’s purposes, “Citizen Auditor” should be understood as a possibility that AC welcomes but does not depend on. The pipeline is designed so that lone users and video-recording teams can contribute valuable data without ever doing audit work themselves — AC handles the audit centrally. The Citizen Auditor path is open for users who want to take on more, but it is not a prerequisite for participation. If and when the Citizen Auditor path becomes more formalized, it will likely resemble a lightweight version of the Audit Committee model (Section 10.3), with AC providing training, tools, and supervision for distributed audit contributors.
10.6 — Safety and Privacy
Personal safety considerations
Actual Vote work, especially Method 1 (recording at polling places on election night), often happens in conditions that are outside the user’s everyday environment: at night, at unfamiliar locations, sometimes alone, sometimes in places where the user’s presence is unusual or visible.
Basic safety practices: be aware of your surroundings — don’t focus so much on the recording that you lose track of who’s around you. Stay in well-lit areas. Keep your phone charged (a dead phone is both a recording problem and a safety problem). Tell someone where you are and your expected route and return time. Have a way to call for help. Don’t approach hostile situations. Trust your instincts — if something feels wrong, leave. No poll tape recording is worth getting hurt over.
Take a bright flashlight. Election-night recording frequently happens in poorly lit areas, and a flashlight serves double duty: illuminating poll tapes for readable recordings and improving personal visibility and safety.
If recording with a team, the buddy system provides an additional layer of safety. Two people at a polling place are less vulnerable than one, can watch each other’s surroundings while recording, and can share the driving.
The risk of being misinterpreted
A specific risk for AV work is that an outside observer may misinterpret what you’re doing. A person standing outside a polling place at night with a phone, recording something, can look unusual or suspicious to someone unfamiliar with Actual Vote. The misinterpretation can lead to a poll worker or property owner asking you to leave, a bystander becoming verbally aggressive, a bystander calling the police, or in extreme cases a confrontation that escalates.
Be ready to explain in friendly, factual terms. Suggested elements: “I’m a volunteer with America Counts, a non-profit non-partisan organization.” “We video-record poll tapes after polls close so they can be compared against the official results.” “This is an independent audit of vote reporting.” “I’m happy to answer questions or to leave if you’d like.”
Don’t argue. The point of the explanation is to defuse the situation, not to win it. If the explanation doesn’t work, leave gracefully.
If a property owner tells you to leave, comply. Property owners have the legal right to set rules on their property. You can typically continue the recording from a different vantage point — a public sidewalk, a parking lot.
If a poll worker or election official tries to stop you from recording legally posted tapes, don’t push back if there’s any doubt about safety. The poll worker may be wrong about the legality, but the polling place is not the place to litigate the question. Comply, document the encounter as a context submission, and let AC follow up through official channels.
If law enforcement is called, be calm and cooperative. Show identification if asked. Explain what you’re doing in factual terms. Comply with any lawful instructions. Don’t escalate.
Document notable encounters. Even if you successfully complete the recording, a notable encounter is worth documenting in the submission’s comment field or as a separate context submission.
De-escalation scripts
Several common confrontation scenarios and suggested responses:
When someone asks what you’re doing out of genuine curiosity, explain Actual Vote and offer to show them the app. Many of these encounters end with the other person thanking you and asking how to participate.
When someone accuses you of election tampering or illegal activity, stay calm. “This is legal — voting results are public information and I have the right to photograph them. I’m happy to show you our website at americacounts.us.” If they remain hostile, leave. If they threaten you, leave immediately and document the encounter.
When a poll worker says you can’t record, comply and ask if there’s a different location from which you can see the tapes. If not, document the exchange and move on.
Privacy protections: what not to record
AV recordings should contain only poll tape content. Users should take care to avoid capturing: faces of poll workers, election officials, voters, or bystanders (frame the recording on the tape, not the surrounding area); audio of conversations between others; license plates or vehicles in identifiable ways; sign-in sheets, ballot stubs, or other voter-identifying materials; and children (exercise extra caution).
AC’s vetting process catches PII issues and rejects submissions that contain personally identifying information. But the user should make a reasonable effort to minimize PII risk at the recording stage rather than relying entirely on vetting.
Privacy protections: the user’s own privacy
Users should also protect their own privacy. Don’t put your face in the recording (watch for window reflections). Don’t include audio of personal information — a conversation with a companion while recording can be picked up. Don’t include other identifying personal items in the frame. Review your submissions in the My Actual Vote Videos tab after uploading and contact AC if you see something that shouldn’t be there.
AC records at a table for in-office viewing (Case 3) are a special case: the recording session can be tedious, and conversations at the table can drift into personal territory. AC has had to reject recordings that captured a participant discussing private health information. Users should be aware that everything said during an active recording is captured on audio.
What AC does to protect submitter identity
AC vets every submission for PII before publishing on the public archive. AC stores submissions on a secure server. AC limits access to submission metadata (GPS coordinates, account information) to authorized staff. AC does not collect information about the submitter’s voting history, demographics, or political views. AC allows submitters to delete their submissions by contacting AC. AC removes submissions from the public archive if PII issues are discovered after publication.
10.7 — Troubleshooting App and Submission Issues
The Google Play Store rejection of November 2024
The most significant app issue in AC’s operational history was the November 2024 Google Play Store rejection. In the days leading up to the 2024 General Election, Google Play rejected several builds of the Actual Vote Android app on the grounds that the app was “government information” and required additional disclosures. AC submitted multiple revised builds; the latest Android version was not approved until November 8, 2024 — three days after election day.
The consequences: Android users who needed to install or update Actual Vote during the critical pre-election and election-day window could not do so via the Play Store. Existing users could continue using the version they had installed, but users who needed the latest version with bug fixes or new features were blocked.
The episode established a recurring operational risk: app store gatekeepers are not neutral infrastructure. They are owned and operated by companies that make their own decisions about what to allow. Politically-adjacent apps — apps in spaces like elections, advocacy, and political organizing — are at higher risk of scrutiny than neutral utility apps. AC cannot guarantee the app’s availability through any app store at any given moment.
Fallback: native camera app + email
When the Actual Vote app is unavailable or malfunctioning, the fallback is to use the device’s native camera app to record the poll tape and then submit the recording to AC by email.
The protocol: open the device’s native camera app, switch to video mode, record the poll tape (frame it clearly, pan slowly, hold the phone steady), stop recording when done, note the metadata (polling place name and address, precinct identifier, date, voting method), and email the recording to j.flatley@democracycounts.org with the metadata in the email body.
The native camera fallback works but is slower, more error-prone (metadata must be manually communicated rather than automatically captured), and harder for AC to process. It is a genuine safety net, not a replacement for the app.
Google Drive folder workflow for bulk submissions
For users with many recordings to submit, email is impractical due to attachment size limits and inbox management. The alternative is the Google Drive folder workflow: email Jason Flatley to coordinate, receive a designated Google Drive folder link, upload the recordings (each as a separate file with a meaningful name), notify Jason when uploading is complete, and AC processes the recordings through the same pipeline as app submissions.
This workflow has been used during the 2024 General Election when some teams had more recordings than the app could comfortably handle. It works well for bulk submissions but requires advance coordination.
Common app issues
The app crashes when trying to upload. Possible causes: poor network connection, large video file, outdated app version, device storage full. Workarounds: try uploading later when network is better, use the fallback camera app, update the app, free up storage.
Recordings show as “Pending Upload” indefinitely. The app has the recording but can’t push it to the server. Workarounds: check network connection, force-quit and restart the app, contact AC if it persists.
Login fails. Either Google sign-in or email-and-password login fails. Workarounds: clear app cache, try the alternate sign-in method, contact AC.
GPS location is wrong. The phone’s GPS hasn’t gotten a fix or shows the wrong location. The recording will still be uploaded but may have inaccurate metadata. Workaround: enter the precinct identifier manually and note the actual location in the comment field.
Recording quality is poor. Low light, wrong focus, or motion blur. This is a technique issue rather than an app issue. The practice poll tape helps users develop technique before going to a real polling place. A flashlight solves most low-light problems.
Video file is too long. Some devices limit recording length. For very long poll tapes, the user may need to record in multiple files. The transcriber can stitch these together.
Precinct identification challenges
Identifying which precinct a recording came from is harder than it looks. The same precinct can be referred to in different ways at different times. The 2024 General Election produced several apparent discrepancies that were actually precinct mismatch artifacts: a tape recorded at “Marion 2” was submitted without the number, creating a mismatch in the comparison logic. GPS coordinates help but are not always sufficient — multiple precincts may share a single polling location, and the GPS puts the user at the building, not at a specific precinct.
AC’s recommended approach: enter whatever precinct information is visible on the poll tape header into the precinct field. If the header shows “PRECINCT 0042” or “MARION 02,” enter that. If the precinct field on the tape is ambiguous, take a screenshot of the tape header and include it in the comment. AC’s transcription team will sort out the mapping.
When to contact AC for support
The general rule: sooner rather than later. If you’re facing an issue you can’t resolve quickly, email Jason Flatley with a description of the problem, what you’ve tried, and any error messages you’re seeing. Include screenshots if relevant. During election week, AC monitors email closely and responds as quickly as possible. During off-season, responses may be slower but still typically come within a day or two.
10.8 — The Human Contact Layer
This subsection — new in v2, addressing the gap between AC’s technical architecture and its actual operational communication — describes how AC and its users actually communicate during an active analysis.
Jason as operational troubleshooter
In practice, the communication infrastructure of America Counts during an active analysis is built around Jason Flatley. Jason is the CTO, the primary point of contact for users, the coordinator for video-recording teams, the troubleshooter when the app fails, the person who responds to email at 2 AM on election night, and the person who reaches out to elections offices with questions about apparent discrepancies. This is not an idealized description of an organization with distributed staff — it is the operational reality of a small nonprofit. The system works because Jason is responsive, knowledgeable, and committed. The system’s vulnerability is that it depends on a single person’s availability and bandwidth.
Communication channels
AC uses several communication channels in parallel, depending on the nature of the communication, the user’s preferences, and the urgency.
Email is the default channel for most communication. The canonical address is j.flatley@democracycounts.org. Email handles initial outreach, coordination with organized teams, follow-up questions during the pipeline, bulk submission coordination, and most non-urgent communication. The general inquiry inbox is info@americacounts.us.
Direct messaging apps handle real-time communication during high-tempo periods. During the 2024 General Election, AC used iMessage, Facebook Messenger, WhatsApp, and Signal at various points to coordinate with specific users in real time. The choice of platform depends on what the user already uses.
Phone calls work better for urgent issues or detailed conversations that benefit from real-time back-and-forth: initial conversations with partner organizations, detailed planning calls with teams, real-time troubleshooting during election week, and sensitive conversations.
The 2024 General Election News Page (and its successors) is a public-facing page updated during active analysis. It is not a direct communication channel but lets users see what’s happening with the analysis and get general updates, including specific issues and apparent discrepancies as they’re being investigated.
Fallback workflows
When the app fails, the human contact layer becomes the primary channel for getting recordings from users to AC. The 2024 Google Play rejection triggered exactly this scenario: users who couldn’t use the app recorded with their phone’s native camera and sent videos to Jason by email, iMessage, or Google Drive. The fallback worked because the communication channels were already established and because users knew who to contact.
For users with many files, the Google Drive workflow (described in Section 10.7) routes recordings through a shared folder rather than through email attachments.
Response time expectations
During active analysis (Phase 3 and Phase 4), AC tries to respond to user communications within hours, not days. The 2024 General Election News Page documents multiple instances of same-day and even same-hour responses during the analysis sprint. During off-season (Phase 1), responses may take a day or two.
AC cannot guarantee 24/7 availability. Users in unusual time zones or with after-hours emergencies should not expect immediate response. The User Agreement is explicit that AC is a small organization with limited capacity.
The philosophy
AC is a small organization that prioritizes human responsiveness over automated systems. There is no ticketing system, no customer support platform, no phone tree. There is a person — Jason — who reads every email, answers every question, and knows the operational context of every active analysis. This is a deliberate choice, not a failure to scale. The personal responsiveness is part of what makes the operational model work: users trust the system because they know a person is behind it, elections offices engage because they’re corresponding with a named individual, and operational problems get resolved quickly because the person who knows the context is the person who answers the message.
The trade-off is that this model does not scale gracefully. The 2024 General Election produced moments when communication was overloaded — too many users with too many issues for one person to handle in real time. Some issues had to be deferred. AC’s modernization planning includes more scalable communication infrastructure — possibly distributed responsibility across more team members, possibly a lightweight ticketing system — but the current informal, personal model is what’s available, and it works at the current scale.
10.9 — The Electoral Cycle
This subsection — new in v2, addressing the gap between the static description of methodology and the dynamic reality of how election work happens over time — describes the rhythm of the year as it relates to Actual Vote operations.
The phases
A typical electoral year has six operational phases for Actual Vote, plus a continuous track that runs throughout.
Off-season (January–June of even years, odd years). This is when most of the underlying capacity gets built. For AC: new jurisdictions are researched, comparison logic is developed and tested against historical data, the state-by-state reference is updated, tools are developed and tested, documentation is written and revised, partner relationships are built and maintained, and fundraising is done. For organized teams: team structure is clarified, new volunteers are recruited (often through partner organizations like Scrutineers), training materials are reviewed, local research is done about the jurisdiction’s poll tape practices, and equipment is checked. For individual users: new users discover AV and practice with the practice poll tape. Phase 1 is calm but not idle. The work done here makes the high-tempo phases possible.
Pre-primary / Pre-general (3–6 months before). Active analyses are scoped. Comparison logic is brought up to date, checked against any reporting changes the jurisdiction has made since the last cycle. Outreach to elections offices happens — confirming poll tape practices, scheduling in-office visits if needed, building relationships. Jurisdiction targeting, volunteer recruitment, SOE outreach, local knowledge gathering, and coverage planning happen in earnest. For how to allocate volunteers across counties for maximum statistical value — breadth over depth — see the deployment playbook in Statistical Inference from AV Results and the full paper’s Section 9.
Active recruitment (1–3 months before). Training sessions, practice runs, precinct assignment, and team leader preparation. For paid deployments, the pay structure is communicated, workers are signed up, and the assignment process begins. The GA2024 debrief recommended that all training materials and instructions be available from the first moment of recruitment rather than released days before the election.
Election week. The highest-tempo period. Final preparations, real-time coordination, recording operations, immediate troubleshooting. Submissions arrive; vetting begins; the communication channels are at full capacity. The 2024 General Election experience showed that this phase can be operationally intense, with submissions arriving faster than they can be processed and new issues emerging unpredictably.
Before-certification window (days to weeks after election). The analysis sprint. Vetting, transcription, comparison, investigation. Findings that result in escalations to elections offices are pursued during this phase. The before-certification window is the window in which AC’s findings can in principle still affect the certified results. Time pressure is real: the certification date is fixed.
Post-certification. Report publication, SOE follow-up, lessons learned, data archival. The analysis transitions to the after-certification window, focused on transparency and building the historical record rather than on changing certified results. This eventually fades back into the off-season for the next cycle.
Continuous. Archive maintenance, website updates, research, fundraising, and ongoing communication between AC and users run through all phases.
Primary vs. general election considerations
Primary elections are lower-profile and lower-turnout, which makes them better environments for testing methodology, building relationships with SOEs, and training new team members. The NC 2024 Primary analysis was conducted in part as a proving ground for the methodology ahead of the General Election.
General elections are higher-stakes, higher-volume, and more operationally demanding. They produce more submissions, involve more jurisdictions, and have tighter certification timelines. AC’s operational planning for a general election begins months before election day.
10.10 — Coordinating with AC Before an Analysis
This subsection — new in v2, providing a concrete protocol for teams and audit committees planning to participate in an analysis — describes what coordination with AC looks like in practice.
Who to contact
Jason Flatley: j.flatley@democracycounts.org. For first-time inquiries, info@americacounts.us also works. Email is the preferred initial channel.
When to reach out
As early as possible. The recommended minimum lead times: six months before the election for new partnerships and major coverage planning; three months before for team formation and preliminary coordination; one month before for detailed precinct assignment and logistics planning; one week before for final confirmations and communication channel testing.
Teams that reach out at the last minute — days before the election — can still participate, but AC’s ability to provide meaningful support (jurisdiction knowledge, analysis planning, comparison logic preparation) is limited by the compressed timeline.
What information AC needs
When a team contacts AC to coordinate, AC needs to know: the target jurisdiction (state, county, specific precincts if known); the target election (primary or general, date); the team size and structure (how many people, who’s leading, what’s the skill level); the access method (Method 1, 2, 3, or a combination); the expected coverage (how many polling places the team can reach); local knowledge the team has (what’s the SOE like, where do tapes get posted, are there known issues); and what the team needs from AC (analysis support, vetting priority, methodology guidance, communication availability).
What AC commits to in return
AC commits to: analyzing the team’s submissions through the central pipeline (the team does not need to do the analysis themselves); vetting the team’s submissions with appropriate priority during the before-certification window; being available for communication during election week (subject to the capacity constraints of a small organization); providing jurisdiction-specific knowledge where AC has it; sharing comparison logic and methodology guidance; and publishing findings through AC’s channels (with appropriate credit to the team).
What AC cannot commit to
AC cannot guarantee: that every submission will be vetted before certification; that analysis will be complete before certification; that AC will have jurisdiction-specific knowledge for every target; immediate response to every communication during high-tempo periods; or outcomes of any kind. The User Agreement documents these limitations. AC does its best within the constraints of a small, under-resourced organization.
The coordination timeline
6 months out: Introduce the team to AC. Discuss target jurisdictions and overall approach. AC shares what it knows about the jurisdiction. The team starts local research.
3 months out: Firm up the team’s coverage plan. AC begins preparing comparison logic for the target jurisdiction. The team starts recruitment if needed. AC and the team align on which contests and precincts matter most.
1 month out: Final coordination. Precinct assignments, route planning, communication channel testing. AC confirms pipeline readiness. The team distributes practice poll tapes and ensures everyone has the app installed.
1 week out: Final confirmation call or email exchange. Last-minute updates on jurisdiction-specific issues. Backup plan review. Everyone knows how to reach everyone else.
Election day and after: Real-time coordination. The team records and submits. AC vets and begins the pipeline. Both sides communicate about issues as they arise. After the election, the team stays available for follow-up questions.
10.11 — Known Operational Failure Modes and Lessons Learned
This subsection — new in v2, documenting specific failure modes from AC’s operational experience — serves as practical reference for users and teams planning their own work, and as institutional memory so the same lessons don’t have to be relearned.
App store rejections
What happened: In November 2024, the Google Play Store rejected several builds of the Actual Vote Android app days before the General Election. The latest version was not available until November 8, 2024 — three days after election day.
What was learned: App store gatekeepers are not neutral infrastructure and cannot be relied upon for time-critical distribution of politically-adjacent apps. Fallback recording and submission methods must exist, must be fully documented, and must be communicated to users before election day. AC now treats app store availability as uncertain and plans accordingly.
User confusion about recording quantity
What happened: During the 2024 General Election, some users were uncertain about what counts as a usable submission and submitted many recordings of various subjects — multiple takes of the same tape, recordings of closed doors, recordings of the surrounding area. One user submitted 90 videos when the expected count was 5.
What was learned: Users need clear, simple guidance on what makes a usable submission. Without it, well-meaning users over-submit in ways that multiply the vetting load and bury their actual usable contributions. AC plans to improve onboarding documentation, provide quicker feedback in the app about whether initial submissions were approved, and develop a self-check protocol users can run before submitting.
Account and workflow breakdowns
What happened: Certain account-related issues — login failures, the “AVA” account issue where recordings were associated with an incorrect account, and Google sign-in problems — created friction during the 2024 General Election that prevented some users from submitting through the app.
What was learned: Account workflows need to be simple and resilient. Users doing time-sensitive election-night work cannot troubleshoot authentication issues in the field. AC plans to simplify the account flow and ensure that the fallback (native camera + email) is clearly communicated as the immediate backup.
Zero tapes submitted alongside results tapes
What happened: Some users recorded zero tapes (printed when the machine boots up, showing all zeroes) alongside or instead of results tapes. Users recording everything posted at a location don’t distinguish between zero tapes and results tapes — they just record what’s there.
What was learned: Zero tapes are important election-integrity artifacts but are not what AV compares against official results. The vetting process must flag and separate them. Users should be told what a results tape looks like versus a zero tape, but the operational reality is that users will record whatever is posted, and the system must handle the sorting.
Poll tapes not posted despite legal requirements
What happened: Even in jurisdictions with posting requirements, compliance is uneven. AC has documented cases of tapes posted behind locked school or church doors after hours, tapes not posted at all, and tapes posted in locations that are physically inaccessible. In the 2024 General Election in Florida, multiple users found polling places where tapes were not posted despite Florida’s statutory requirement.
What was learned: The posting requirement is a statute, not a guarantee. Training should set realistic expectations: not every location will have tapes posted, and the user should document the absence and move on. The next-day return (going back to the location the following morning to check whether tapes have been posted or to visit the SOE office) is worth building into the operational plan. AC has also considered having an attorney send a certified letter to polling locations ahead of the election, informing them of the recording project and the posting requirement — the GA2024 debrief discussed this possibility.
Poll tapes obscured, taped together, or posted behind glass
What happened: Poll tapes have been found crinkled, torn, taped together in ways that make sections unreadable, posted behind reflective glass, or posted with low ink density that makes numbers faint. In some cases, SOE staff have cut poll tapes to tape them onto paper for xeroxing — which AC considers potential destruction of evidence.
What was learned: Recording conditions in the field are frequently imperfect. Users should do their best with what’s available, including recording from multiple angles if the tape is behind glass (to reduce glare), using a flashlight for low-ink tapes, and documenting the physical condition of the tape as part of the submission. AC’s transcription team works with what it gets; a partially legible tape is better than no tape.
Upload failures in low-connectivity areas
What happened: Users in rural areas or areas with poor cellular coverage experienced upload failures. Recordings were captured by the app but sat in a “Pending Upload” state until connectivity improved.
What was learned: The app should handle delayed uploads gracefully (which it does — recordings queue and upload when connectivity returns). Users should be told to expect this and not to force-quit the app out of frustration when uploads are slow. For bulk submissions, the Google Drive workflow (upload from a location with Wi-Fi after the recording session) is a better approach than relying on field connectivity.
Precinct identification difficulties
What happened: Multiple apparent discrepancies in the 2024 General Election turned out to be precinct mismatch artifacts. The “Marion 2” case, the “Marion 2 Retake” confusion, and the GLENWD apparent discrepancy all stemmed from difficulties in correctly identifying which precinct a recording came from.
What was learned: Precinct identification is harder than it looks. Per-jurisdiction precinct lookup tables that include common variants and synonyms would help. The app’s precinct field could suggest known precincts based on GPS. The comparison code should include a normalization step that handles common variants automatically. Until these improvements are built, users should enter whatever precinct information appears on the poll tape header and let AC sort out the mapping.
Transcription errors and quality control
What happened: Transcription errors — transposed digits, misread values, missed contests — entered the dataset and produced apparent discrepancies downstream. In most cases, the errors were caught during the investigation phase (the transcription error check is the first of the seven checks), but each error consumed investigation time.
What was learned: Transcription is where errors most easily enter the pipeline. Quality control approaches include re-watching key frames when values seem unusual, cross-checking sums, spot-checking by a second person, and using early comparison runs as a sanity check. AI-assisted transcription (using vision models to do a first pass with human verification) is the direction AC wants to move to reduce the error rate.
Timing pressures
What happened: The before-certification window imposes a hard deadline on the analysis pipeline. In the 2024 General Election, the multi-state effort produced findings over many weeks as a rolling analysis rather than completing before any single state’s certification date. Some investigation threads were still open when certification occurred.
What was learned: Completing the full pipeline before certification is the goal but is not always achievable, especially for large or complex analyses. Rolling publication (the 2024 General Election News Page model) makes findings visible as they emerge rather than waiting for the complete analysis. Post-certification publication still produces value for the public record, even if it cannot affect the certified results.
SOE cut/damaged poll tapes
What happened: AC has observed cases where poll tapes appear to have been cut or damaged — potentially by elections office staff in the course of handling them. In some jurisdictions, tapes are cut from the printer and taped onto paper for xeroxing. The GA2024 debrief flagged this as potential destruction of evidence.
What was learned: Poll tape handling practices vary and some are destructive. AC documents the physical condition of tapes as observed in the recordings. Where evidence of damage is observed, the analysis notes it, and if AC has occasion to communicate with the SOE, the handling practice is raised. AC recommends that elections offices handle poll tapes as the official records they are and avoid cutting, taping, or otherwise altering them.
Wake County PDF obfuscation
What happened: Wake County, North Carolina published official results in PDF files with non-standard character encoding, hidden positioning offsets, and image-substituted text — making it extremely difficult to extract the numeric values programmatically. The obfuscation may have been intentional or may have been an artifact of the county’s PDF generation process.
What was learned: The format of official results is a significant variable in the analysis pipeline. Machine-readable formats (CSV, Excel) are strongly preferred. When results are published in obfuscated formats, AC can still extract them but at much greater analyst cost. AC’s recommendations to elections offices (Section 15.1) include publishing results in machine-readable formats.
The improvement philosophy
Across these failure modes, several general lessons emerge. The pipeline has known weaknesses at specific points, and resilience requires deliberate fallback paths and cross-checks. Communication is the connective tissue — when communication breaks down, other things break down too. Documentation is high-leverage: time spent writing things down pays back in time saved later. And each cycle makes the system better. Failures are expected, documented, and addressed. The fact that AC documents its failures openly (including in this manual) is itself a reflection of the methodology’s commitment to transparency: the same principle that drives AC to publish its findings drives it to publish its operational lessons.