11 — Training and Volunteer Onboarding

11.1 — Purpose and Audience

This section describes how new participants — volunteers, paid workers, team coordinators, and aspiring analysts — get from “I want to help” to being effective contributors to the Actual Vote ecosystem. It is a practical guide to onboarding, not a restatement of the methodology (which is in Section 8) or the operational practices (which are in Section 10). Readers should treat this as the “how to get started” companion to those sections.

The training needs differ substantially by role. A lone user recording a single poll tape on election night needs fifteen minutes of preparation. A video-recording team coordinator managing ten workers across a county needs hours of planning and a detailed understanding of the operational landscape. An analyst building comparison logic for a new jurisdiction needs deep familiarity with the methodology, the seven-check investigation process (which triages initial non-matches and informs investigation before any heavier ask to the elections office), and the jurisdiction’s specific reporting practices. This section is organized by role, with shared foundations first and role-specific modules after.

11.2 — Shared Foundations (All Roles)

Every participant, regardless of role, should understand four things before doing any AV work.

What Actual Vote does

Actual Vote independently records the numbers printed on poll tapes at voting locations and compares them against the officially published results. If the numbers match, the vote reporting was accurate. If they don’t match, something went wrong in the reporting chain — and the recording provides the evidence to investigate what.

This is the reporting layer, not the counting layer. AV does not check whether scanners read ballots correctly. It checks whether the numbers the scanners produced were faithfully reported in the official results. The distinction matters because it defines what AV can and cannot find. A new participant who understands this distinction will make better decisions in the field and will have appropriate expectations about what their work produces.

The one-paragraph version for conversations with the public: “We video-record the results tapes that voting machines print after polls close and compare them against the official election results. In almost every case, the numbers match — which is good news. When they don’t match, we investigate why. We’re non-partisan and independent.”

Why it matters

New participants sometimes ask why this work is necessary if the government already has safeguards. The answer is concrete and documented: the safeguards don’t check the reporting layer. Virginia’s risk-limiting audit confirmed counting accuracy with 99%+ confidence while a 4,000-vote reporting error went undetected. Monmouth County’s post-election audit reported “100% accuracy” while the wrong person held a school board seat. These are not hypothetical scenarios — they are documented recent events in jurisdictions with modern voting equipment and established audit procedures. AV fills the gap that these safeguards leave.

What’s legal

Recording publicly posted poll tapes is legal. Poll tapes posted in public view after polls close are public documents. A person standing on a public sidewalk or in an area open to the public has the legal right to photograph or video-record them. The legal basis is detailed in Section 6, but for training purposes, the key points are: you have the right to be there, you have the right to record, and the police have confirmed this in real encounters (the Broward County 2018 sergeant confirmed the team’s right to be present after a precinct captain objected). That said, a property owner can ask you to leave their property, and you should comply — then record from a public vantage point instead.

What a poll tape looks like

Before going into the field, every participant should have seen a poll tape and know what they’re looking for. A poll tape is a paper printout — typically a long, narrow strip — produced by a voting machine after polls close. It lists the contests on the ballot, the candidates, and the number of votes each received at that machine or precinct. The header usually shows the precinct name, the date, and the machine identifier. The format varies by equipment manufacturer and jurisdiction.

AC provides a practice poll tape — a downloadable PDF — that shows what a real tape looks like. Recording the practice tape at home is the single most valuable pre-field preparation a new user can do. It develops the motor skills (panning speed, framing, focus distance) and the visual recognition (what the numbers look like, where the contest names are, where the header and footer are) that make field recording faster and more reliable.

The practice tape lives on the website at /practice-poll-tape/, where it can be downloaded as a printable PDF. One generic tape is provided; it’s enough to build the motor skills and visual recognition that matter, even though real tapes vary by equipment vendor and are often longer. (A recording of a blank wall or a grocery receipt also works for a first dry run — the goal is simply to confirm the app works and rehearse the steps.)

11.3 — Role 1: Lone User

Who this is: A single person, typically recording at one to five polling places on election night. No team, no coordinator. The most common user type and the one AV was originally designed for.

Time to onboard: 15–30 minutes.

Checklist

Before election day:

Download the Actual Vote app (iOS or Android). Create an account.
Read the in-app instructions. Complete the consent flow.
Download and print the practice poll tape. Record it with the app. Review the recording — can you read every number? Is the frame steady? Is the pan slow enough?
Identify the polling place(s) you plan to visit. Confirm that your state requires poll tapes to be posted (check the state classification in Section 7.1 or the state-by-state appendix in B.3). If your state is classified as “dark green” or “light green,” tapes are likely to be posted. If “gold” or “red,” consider Method 3 (in-office viewing) instead.
Plan your route. Know where each polling place is, how long it takes to drive between them, and where the tape is likely to be posted (typically on a window, wall, or door visible from outside).
Charge your phone. Bring a portable charger and a flashlight.
Know the fallback: if the app isn’t working, record with your phone’s native camera and email the video to j.flatley@democracycounts.org with the polling place name and precinct identifier.

On election night:

Arrive at your first polling place around closing time or shortly after. Tapes may not be posted immediately — expect a wait of thirty minutes to several hours after polls close.
Find the posted tapes. They are usually visible from outside the building.
Open the app. Record the tape by slowly panning from top to bottom. Keep the camera steady and close enough to read every number. If the tape is behind glass, try different angles to reduce glare. Use your flashlight if it’s dark.
Confirm the upload in the app.
Move to the next location.
If tapes aren’t posted, record a brief context video showing the location and noting the absence. This is useful information.

After election night:

Check the My Actual Vote Videos tab to confirm your submissions were received.
Be available by email for a few days in case AC has follow-up questions about your recordings.

Common first-timer mistakes

Recording too fast — the camera pans faster than the transcriber can read. Recording too far away — the numbers are too small to transcribe. Recording the zero tape instead of (or in addition to) the results tape — the zero tape is the one printed when the machine boots up, showing all zeroes, and is typically posted alongside the results tape. Not bringing a flashlight — election-night recording frequently happens in poorly lit areas. Assuming every polling place will have tapes posted — compliance is uneven, even in states with posting requirements.

11.4 — Role 2: Video-Recording Team Member

Who this is: A member of an organized group recording poll tapes across multiple locations, with a coordinator handling logistics. The team submits recordings to AC; AC handles the analysis.

Time to onboard: 1–2 hours (including team briefing).

Everything from the Lone User checklist, plus:

Coordination. You’ll receive assignments from your team coordinator — specific polling places, a route, a communication plan. Know your assignments before election night. Save the addresses. Test the communication channel (group text, Signal, Slack, whatever the team uses).

Coverage discipline. If your assigned location doesn’t have tapes posted, document the absence and check in with your coordinator before improvising. Your coordinator has the big picture and may redirect you to a higher-priority location. Don’t freelance coverage unless you can’t reach your coordinator.

Real-time reporting. Report back to your coordinator as you complete each location: “Location A done, 2 tapes recorded, moving to Location B.” This lets the coordinator track coverage in real time and redirect resources if needed.

Buddy system. If the team has enough people, pair up. Two people at a location are safer, can cover more tapes simultaneously (if multiple tapes are posted), and can watch each other’s surroundings.

De-escalation. Know the de-escalation scripts from Section 10.6. The short version: be friendly, explain what you’re doing (“I’m a volunteer with America Counts, a non-partisan organization that independently verifies election results”), don’t argue, and leave if anyone becomes hostile. No poll tape recording is worth a confrontation.

11.5 — Role 3: Team Coordinator

Who this is: The person who organizes a video-recording team — recruiting members, planning coverage, assigning routes, managing communication on election night, and debriefing afterward. This is the most operationally demanding volunteer role.

Time to onboard: Several hours over several weeks.

Pre-election phase (weeks to months before)

Recruit. Sources include partner organizations (Scrutineers, Common Cause, League of Women Voters, local voter protection coalitions), personal networks, campus civic groups, and AC’s own outreach. Begin months before election day — the GA2024 debrief identified late recruitment as a recurring problem. Close to election day, potential volunteers are saturated with campaign requests.

Scope. In coordination with AC (contact Jason Flatley), identify which election, which jurisdiction, and which polling places to target. Get the list of polling places from the county elections office or website. Prioritize based on coverage feasibility, contested races, historical issues, and AC’s guidance on which precincts matter most for the analysis. 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.

Assign routes. Map each team member’s assigned locations and driving order. Consider drive time, expected posting times, and geographic clustering. The RouteXL tool has been used for route optimization. Budget at least thirty minutes per worker for this step — it is more time-consuming than it appears.

Build backup plans. Assign overlap so that if one member drops out, another’s route covers some of the same locations. Identify a standby list of members who can fill gaps on short notice.

Establish communication. Choose a channel (group text, Signal, Slack), test it with all members, and agree on the protocol: when to check in, what to report, how to handle problems.

Distribute materials. Every team member should have, before election day: the app installed and tested with the practice tape, their route and assignments, the communication plan, the fallback recording procedure, the de-escalation scripts, and realistic expectations about what they’ll encounter (some locations won’t have tapes; that’s normal).

Election night

Monitor. Track which locations have been covered as team members report in. Redirect members to uncovered locations as needed. Be available by phone for troubleshooting.

Handle problems. Common election-night problems: a member can’t find the polling place, tapes aren’t posted, the app isn’t working, a member encounters a hostile individual. For each: stay calm, give the member clear guidance, and document the issue.

Don’t record yourself (unless you have to). The coordinator’s most valuable election-night contribution is coordination, not recording. If you’re driving between locations recording tapes, you’re not available to manage the team. Record only if you have spare capacity or if coverage is thin.

Post-election

Confirm submissions. Check with each team member that their recordings were uploaded successfully. Follow up on any that are stuck in “Pending Upload.”

Debrief. Within a few days, collect notes from each team member on what worked, what didn’t, what was surprising. Write a short summary and send it to AC. The GA2024 debrief — the single most valuable operational document in AC’s history — was built from exactly this kind of post-election feedback.

Stay available. AC may have follow-up questions during vetting, transcription, and investigation. The coordinator is usually the best person to route these to the right team member.

11.6 — Role 4: Paid Worker

Who this is: A volunteer who is paid for their recordings — not by America Counts, but by another organization. The large majority of Actual Vote work to date has been done by unpaid volunteers. Some partner organizations, however, choose to pay people per poll tape recorded in order to increase the number of recordings collected in a given area.

It’s important to be clear about America Counts’s role here: AC does not pay workers directly, and AC does not offer paid roles itself. What AC does is support other organizations that run paid recording efforts. So for any individual, we can’t promise that a paid role will be available — it depends entirely on whether an organization is funding that work in your area at that time.

Time to onboard: 1–2 hours.

What’s different from volunteer work

The core recording task is identical. What’s different is the relationship with the funding organization and its expectations.

Compensation. Different organizations have used different compensation structures for paid workers, often paying them by the number of results tapes recorded. Payment methods vary, and workers are usually paid within days of submitting their recordings.

Quality expectations. Paid workers are held to the same quality standards as volunteers. Recordings must be clear, complete, and properly framed.

Reliability. A paid worker who accepts an assignment and doesn’t follow through creates a coverage gap that may not be discovered until it’s too late to fill. If something prevents you from completing your assignment, notify the organizing team immediately so the gap can be addressed.

Onboarding for paid workers

The onboarding checklist is the same as for team members (Role 2 above), with the addition of:

Understanding the funding organization’s compensation structure and payment process.
Confirming availability for the assigned date and route.
Providing contact information for the organizing team.

11.7 — Role 5: Transcriber

Who this is: A person who watches poll tape recordings and enters the numeric values into a structured format (typically a spreadsheet) for use in the comparison pipeline. Transcription is the bridge between field recording and analysis.

Time to onboard: 2–4 hours (including supervised practice).

What transcription involves

The transcriber watches a video recording of a poll tape — typically 1–3 minutes long — and enters the values into a transcription spreadsheet. For each contest on the tape, the transcriber records the contest name, the candidate names, and the vote counts. The transcriber also records metadata from the tape header: precinct name, date, machine identifier.

The work is painstaking. A single tape may contain dozens of contests with hundreds of individual values. Accuracy is critical because transcription errors produce apparent discrepancies that consume investigation time downstream. The first check in the seven-check investigation process (Section 8.9) is “transcription error” — precisely because transcription is where errors most easily enter the pipeline.

Skills required

Attention to detail. Comfort with spreadsheets. The ability to read numbers from imperfect video — low resolution, poor lighting, partial obscuration, faded ink. Patience. The work is repetitive and the stakes of each individual number are high.

No specialized election knowledge is required to start. Transcribers learn the format of specific vendors’ poll tapes through exposure. A transcriber who has processed ten ES&S tapes can process the eleventh much faster because they recognize the layout.

Training sequence

The recommended training sequence, once materials are available:

Orientation. Read this section and Section 8.1–6.3 (methodology overview through transcription). Understand where transcription fits in the pipeline and why accuracy matters.
Template walkthrough. Review the transcription template. Understand what goes in each field.
Supervised practice. Transcribe 3–5 practice recordings with a mentor available to answer questions. Compare your transcription against the answer key. Discuss any discrepancies.
Independent practice. Transcribe 5–10 additional recordings independently. Submit for review.
Pipeline integration. Begin transcribing real submissions from the current analysis, with spot-checking by a senior transcriber or AC staff.

Common transcription errors

Transposed digits (entering 546 as 564). Misread digits caused by poor video quality (a 3 that looks like an 8, a 1 that looks like a 7). Skipped contests (missing a race entirely, especially near the bottom of a long tape). Duplicate entry (entering the same contest twice from a tape that printed it twice, as some machines do for different voting methods). Misattributing values between candidates in a race. Entering values from the zero tape rather than the results tape.

AI-assisted transcription

AC is moving toward AI-assisted transcription using vision models to perform a first pass, with human review of the AI’s output. This approach has the potential to substantially increase throughput and reduce the error rate for routine tapes, while still requiring human judgment for edge cases — poor-quality video, unusual tape formats, contests with unusual candidate names or write-ins.

11.8 — Role 6: Analyst

Who this is: A person who builds comparison logic, runs the comparison between transcribed poll tape values and official results, investigates apparent discrepancies, and produces findings. This is the highest-skill role in the AV pipeline and typically the bottleneck.

Time to onboard: Weeks to months, depending on background.

Prerequisites

Proficiency with data manipulation — at minimum, advanced spreadsheet skills; ideally, scripting ability in Python or R. Familiarity with the manual’s methodology section (Section 8), especially Section 8.4–6.11 (obtaining official results through investigation and asking the elections office). Understanding of how official results are structured in the target jurisdiction (reporting grain, voting method breakdowns, precinct naming conventions). Comfort with ambiguity — the investigation process frequently involves tracking down explanations for puzzling discrepancies that may turn out to be mundane.

Training sequence

Read the methodology. Section 8 in its entirety, with particular attention to the seven-check investigation process (Section 8.9) and the worked examples.
Read the comparison analysis guide. The “Doing an Actual Vote Comparison Analysis” section in this manual walks through the full process from scoping through investigation, asking the elections office, and — in the rare case of a fully unresolved apparent discrepancy — handing off to a court-capable party.
Study a completed analysis. Read the NC 2024 Primary Report end to end. Understand how the transcription data, official results, comparison logic, and investigation notes fit together.
Shadow an existing analysis. If timing permits, observe an active analysis in progress — sitting in on investigation discussions, watching how discrepancies are triaged, seeing how comparison logic is adapted when assumptions break down (as happened with the Georgia vote-center grain problem in 2024).
Build comparison logic for a practice jurisdiction. Using transcription data and official results from a previous election, build the comparison from scratch. This is where the real learning happens: discovering that precinct names don’t match between the tape and the official results, that the official results aggregate early voting differently than the analyst expected, that write-in candidates are reported in a separate column. Each jurisdiction has its own idiosyncrasies, and building the comparison logic is how the analyst learns them.
Run a supervised analysis. Conduct a comparison analysis on real data — either a current election or a historical one — with oversight from Jason or an experienced analyst. Triage initial non-matches through the full seven-check process, investigating each discrepancy before making any heavier ask to the elections office. Produce a findings document.

The seven-check investigation process (summary for training)

When a comparison produces an initial non-match, the analyst runs through seven checks in order. The checks triage whether the non-match is an AC process error (resolved to a match, unreported) or a discrepancy — and, if a discrepancy, whether a benign hypothesis exists — before any heavier ask to the elections office:

Transcription error. Re-watch the video. Did the transcriber enter the wrong number?
Comparison logic error. Is the comparison matching the right precinct to the right official result? Are naming conventions consistent?
Missing poll tapes. Is the comparison expecting a tape that wasn’t recorded? Missing tapes produce systematic discrepancies.
Duplicate tapes. Was the same tape transcribed twice? Duplicates inflate the tape-side total.
Aggregation logic error. Does the comparison logic correctly aggregate precinct-level tape values to match the grain of the official results?
Undocumented vote reporting logic. Is the jurisdiction doing something with the data that the analyst didn’t expect? This is the Georgia vote-center problem — official results reported at a different grain than the analyst assumed.
Operational context. Is there a real-world explanation? A ballot jam causing re-scans, a precinct relocation due to a hurricane, a machine replacement mid-day?

Most apparent discrepancies resolve at checks 1–3. Those that reach checks 6–7 are the most interesting and the most likely to surface real findings. The full treatment of this process is in Section 8.9.

11.9 — Role 7: Audit Committee Leader

Who this is: A person who organizes and leads a full Audit Committee — a local group that runs the entire comparison analysis pipeline, from recording through investigation and reporting. This is the most demanding role and is relatively rare; most groups operate as video-recording teams with AC handling the analysis.

Time to onboard: Months.

When to pursue this path

The Audit Committee model makes sense when: the jurisdiction is unfamiliar to AC and the committee can build local knowledge faster; the committee has specific expertise (legal, technical, journalistic) that AC doesn’t have for that jurisdiction; the committee wants to publish findings under its own name for local credibility reasons; or the committee is in a jurisdiction AC isn’t planning to analyze.

For most groups, AC recommends the video-recording team model. Centralized analysis is more consistent, AC has accumulated jurisdiction-specific expertise, and findings published under AC’s name carry AC’s institutional reputation. The committee model is high-effort and high-maintenance. The canonical example — Citizens Audit Broward (CAB) — eventually transitioned to operating as a video-recording team because the full-pipeline workload was difficult to sustain.

What it takes

An Audit Committee leader needs everything an analyst needs (Role 6), plus: the ability to recruit and manage a recording team (Role 3); familiarity with the local elections office and the relevant officials; the judgment to make scope decisions (what to analyze, what to skip); the communication skills to present findings publicly; and the sustained commitment to see an analysis through from recording to publication over a period of weeks.

Training path

Complete the coordinator training (Role 3).
Complete the analyst training (Role 6).
Conduct a full analysis of a past election in the target jurisdiction as a dry run, with AC providing guidance.
Coordinate with AC on methodology alignment — your findings will be more credible if your methodology matches AC’s published approach.
Plan and execute a live analysis in coordination with AC.

11.10 — Training Materials and Resources

This section catalogs the training resources available to participants at each level.

Currently available

[coming soon]

Recommended additions

Based on the operational lessons in Section 10.11 and the training structure described above, the following materials would strengthen the onboarding pipeline:

For all roles: A 5-minute video showing the complete recording process from arrival at a polling place through upload confirmation. Seeing someone do it is faster than reading about it.

For lone users: A one-page quick-start card (printable) with the election-night checklist, the fallback procedure, and the de-escalation scripts. Something a user can fold and put in their pocket.

For team coordinators: A coordinator’s handbook covering recruitment, route planning, assignment templates, the communication protocol, election-night coordination, and the post-election debrief. The GA2024 debrief recommended this; it should be formalized.

For transcribers: A set of 10 practice recordings with answer keys, graded from easy (clean recording, standard ES&S tape, good lighting) to hard (behind glass, faded ink, unusual format). A transcription accuracy benchmark: what error rate is acceptable before a transcriber works on live data.

For analysts: A worked example analysis for a training jurisdiction — the full dataset (recordings, transcriptions, official results, comparison logic, investigation notes) with annotations explaining each decision. The NC 2024 Primary is close to this but would benefit from being packaged as an explicit training exercise.

11.11 — Onboarding Workflow

The standard onboarding path, from first contact to active participation:

Step 1: Initial contact

A new participant reaches out to AC — by email, through the website, through a partner organization, or at an event. Alternatively, AC reaches out to a potential participant or group.

Step 2: Role identification

In the initial conversation, identify the appropriate role: lone user, team member, coordinator, paid worker, transcriber, or analyst. Most new participants start as lone users or team members. The conversation should cover: the participant’s location (which determines the relevant jurisdiction and access method), their availability (election night only, or sustained engagement), their skills (technical background, organizational experience), and their motivation (why they want to participate).

Step 3: Self-service onboarding (lone users)

Lone users can onboard entirely through self-service: download the app, practice with the practice tape, review the quick-start materials, and show up on election night. AC welcomes a brief introductory email but does not require it.

Step 4: Coordinated onboarding (team roles)

Team members, coordinators, and paid workers onboard through coordination with AC. The sequence: introductory call or email with Jason → role-specific training materials → practice session (recording, transcription, or analysis as appropriate) → assignment for the next election.

Step 5: First deployment

The participant’s first real election. For lone users, this is self-directed. For team members and paid workers, this is coordinated by the team leader. For coordinators, this is planned in partnership with AC. For transcribers and analysts, this is supervised by AC or a senior analyst.

Step 6: Debrief and retention

After the first deployment, a debrief captures what worked and what didn’t. The debrief is both operational feedback (for AC’s improvement) and a retention touchpoint (for the participant’s continued engagement). Participants who have a positive first experience and feel their contribution mattered are more likely to return for the next election.

11.12 — Scaling Considerations

The current training model is one-to-one: Jason onboards each new coordinator, each new team, each new analyst personally. This works at the current scale but will not work at the scale AC aspires to. Several approaches to scaling have been discussed or are in development:

Train-the-trainer. Experienced coordinators train new coordinators in their region, with AC providing materials and quality assurance. This is the most natural scaling path for the team model.

Self-service training for recording roles. The app, the practice tape, the quick-start card, and the introductory video should enable a lone user or team member to onboard without any direct contact with AC. This is largely possible today but would benefit from the materials improvements described in Appendix X.

Structured analyst training. The analyst role cannot be fully self-service — it requires jurisdiction-specific knowledge and methodological judgment that comes from experience. But a structured training program (read these sections, work through this practice analysis, submit your work for review) could reduce the time Jason spends on each new analyst.

Regional hubs. As volunteer networks grow in specific states, regional coordinators could take on the onboarding and coordination responsibilities that Jason currently handles. North Carolina, with its established volunteer base and well-documented jurisdiction-specific knowledge, is the natural first candidate.

11.13 — Quick Reference: Role Summary

Role	What you do	Time commitment	Skills needed	Onboarding time
Lone user	Record 1–5 tapes on election night	2–4 hours on election night	Phone, app, willingness to show up	15–30 min
Team member	Record assigned locations as part of a coordinated team	4–8 hours on election night + pre-election briefing	Same as lone user + following instructions	1–2 hours
Team coordinator	Recruit, plan, assign, manage a recording team	Weeks of pre-election planning + election night coordination + debrief	Organizational skills, people management, local knowledge	Several hours over weeks
Paid worker	Record assigned locations for per-tape compensation	4–8 hours on election night	Same as team member	1–2 hours
Transcriber	Convert video recordings to structured numeric data	Flexible, ongoing during analysis period	Attention to detail, spreadsheet skills, patience	2–4 hours
Analyst	Build comparison logic, investigate discrepancies, produce findings	Weeks during analysis period	Data manipulation, scripting, methodological judgment	Weeks to months
Audit Committee leader	Run the entire pipeline for a jurisdiction	Months	All of the above + leadership, local expertise	Months