The Parallel Vote Tally: How Parties Ran Shadow Counts in 2022

On the night of August 9, 2022, while the IEBC was receiving electronically transmitted results at Bomas of Kenya, two other operations were running simultaneously in different locations across Nairobi. UDA's war room was aggregating results from their agents. Azimio la Umoja was doing the same. Both were running what's known as a Parallel Vote Tabulation (PVT) — an independent count of results as reported by party agents stationed at every polling station.

By the morning of August 10, UDA's parallel tally showed William Ruto leading with approximately 50.5% of counted votes. The final IEBC result gave Ruto 50.49%. Their projection was accurate to within a tenth of a percentage point.

Azimio's parallel tally, by contrast, showed Raila Odinga leading — a result that was at odds with both the IEBC count and UDA's independent tally. The discrepancy contributed to Azimio's decision to reject the results and file a Supreme Court petition.

What Is a Parallel Vote Tally?

A parallel vote tally (PVT), also called a parallel vote tabulation or quick count, is an independent process of collecting and aggregating election results from polling stations. It works like this:

1. Agent deployment: The party or organization deploys trained agents to every polling station (or a statistically representative sample)
2. Result recording: After votes are counted at each station, the agent records the results from the official Form 34A before it's sealed
3. Data transmission: The agent sends the recorded results to a central command center, typically via SMS, phone call, or a dedicated app
4. Aggregation: The command center aggregates incoming results in real time, building a running tally
5. Projection: Once a sufficient sample is in, the tally can project the overall result with a stated margin of error

In Kenya, political parties have the legal right to deploy agents to every polling station. The Elections Act requires the IEBC to allow accredited agents to witness the count and receive copies of Form 34A. This means parties have access to the same primary source data as the IEBC itself.

UDA's Operation

UDA deployed approximately 92,000 agents across Kenya's 46,229 polling stations — roughly two agents per station. Their parallel tally operation was reportedly managed from a command center in Karen, Nairobi, and used a combination of a dedicated mobile application and SMS reporting.

Key features of UDA's system:

• Photo verification: Agents photographed the Form 34A and uploaded it alongside their typed results, allowing cross-verification
• Real-time dashboards: Results were aggregated into constituency and county dashboards that campaign leadership could monitor
• Quality control: Agents whose reported numbers were outliers (too high or too low compared to registered voters at that station) were flagged for verification
• Coverage: UDA reportedly received results from approximately 95% of polling stations within 24 hours of polls closing

UDA's final parallel tally reportedly showed Ruto winning with approximately 7.1 million votes out of about 14.1 million valid votes — very close to the official IEBC figure of 7,176,141 votes (50.49%).

Azimio's Operation

Azimio's parallel tally operation was reportedly less cohesive. The coalition comprised multiple parties — ODM, Jubilee, Wiper, KANU, and others — each of which had deployed their own agents. Coordinating results across multiple party structures created logistical challenges.

According to media reports and post-election analysis, Azimio's operation suffered from:

• Fragmented data collection: Different member parties used different reporting channels, making real-time aggregation difficult
• Coverage gaps: In some Rift Valley and Mt. Kenya constituencies where Azimio had weaker organizational presence, agent coverage was incomplete
• Selective reporting: Some analysts suggested that agents in Azimio strongholds reported promptly while those in opposition areas were slower, creating an early bias in the running tally
• Technology limitations: Reports indicated that Azimio's system relied more heavily on phone calls and manual data entry, increasing the risk of transcription errors

Why UDA's Tally Was More Accurate

The difference between UDA's accurate parallel tally and Azimio's inaccurate one comes down to three factors:

1. Organizational discipline. UDA was a single party with a unified command structure. When UDA headquarters issued instructions on reporting format and timing, agents followed a single protocol. Azimio was a coalition of parties, each with its own chain of command. Getting ODM, Jubilee, and Wiper agents to report through a single system proved unwieldy.

2. Technology investment. UDA reportedly invested more heavily in their parallel tally technology platform. A purpose-built application with photo verification, GPS tagging, and real-time dashboards provides significantly better data quality than SMS or phone call-based systems.

3. Full coverage. A parallel tally is only as good as its coverage. If you're receiving results from 95% of stations, your projection will be very close to the official result. If you're receiving from 70% — and those 70% are disproportionately from your strongholds — your projection will be biased.

Civil Society PVTs

Political parties weren't the only organizations running parallel tallies. The Elections Observation Group (ELOG), a coalition of Kenyan civil society organizations, conducted a statistical PVT using a representative sample of 1,500 polling stations (about 3.2% of all stations).

ELOG's PVT projected Ruto winning with 50.3% ± 1.2 percentage points — confirming the IEBC result. The Catholic Church's Justice and Peace Commission also ran an independent count from approximately 32,000 stations and reached a similar conclusion.

These independent PVTs are crucial for election credibility. When multiple independent sources — the IEBC, a civil society PVT, the Catholic Church, and the winning party's own tally — all converge on the same result, it becomes very difficult to credibly argue the result was manipulated.

The Future of Parallel Tallies in Kenya

The 2022 experience demonstrated that parallel vote tallies have moved from a nice-to-have to an essential component of Kenya's electoral architecture. For the 2027 election, several trends are likely:

• More sophisticated technology: Parties will invest in dedicated PVT platforms with AI-powered anomaly detection, photo OCR (optical character recognition) to automatically read Form 34A images, and blockchain-based audit trails
• Higher expectations: After UDA demonstrated that 95% coverage and 0.1% accuracy is achievable, anything less will be seen as organizational failure
• Integration with social media: Real-time PVT dashboards may become public-facing, allowing parties to share their running tallies with supporters — adding transparency but also pressure

The lesson from 2022 is clear: if you're running a campaign and you don't have a robust parallel tally, you're operating blind on election night. You'll be at the mercy of whatever the official count says, with no independent verification. That's not a position any serious campaign should be in.

Run your 2027 parallel tally with confidence. Votrack has processed millions of results entries and can handle every polling station in Kenya simultaneously. From agent deployment tracking to real-time anomaly detection, we give you the tools to know your numbers before the announcement. Talk to our team.