Case Study 33-1: The Riverside Problem

Background

Three weeks after Nadia Osei launched the voter contact dashboard, the Garza campaign faced a specific problem that the dashboard had both revealed and complicated: Riverside County was falling significantly behind its contact goal, and the reason wasn't what anyone had expected.

Riverside County was the campaign's most competitive geography — the county where polling showed the race essentially tied, where field staff estimated the outcome would be decided. The campaign had allocated 28,000 contact attempts to Riverside over the 35-day general election period. The dashboard showed that with 15 days left, Riverside was at 61% of its contact goal — the furthest behind of any county, and the one with the least time to catch up.

But Nadia's quality metrics complicated the picture. Riverside's persuadability targeting lift was +11 percentage points — the highest in the state. Canvassers in Riverside were reaching persuadable voters at a dramatically higher rate than canvassers elsewhere. The conversion rate — positive outcomes as a share of total contacted — was 24%, compared to a statewide average of 19%.

Riverside was behind on volume. It was ahead on quality.

The Data

The dashboard's county-level report for Riverside showed:

The not-home rate was the most revealing data point. In Riverside, canvassers were attempting doors but finding nobody home at a significantly higher rate. Nadia's first hypothesis: the county had a higher proportion of voters who were working multiple jobs and were simply harder to find at home during canvassing hours.

She queried the voter file to look at the occupational and income profile of Riverside's uncontacted high-priority voters. The data suggested a concentration of manufacturing and logistics workers — consistent with Riverside County's economic geography — whose work schedules made evening and weekend canvassing windows shorter and more competitive.

The Campaign's Response Options

Nadia presented the data at the campaign's Tuesday field call. Yolanda Torres and the Riverside regional director, Marcus Gould, listened to the analysis and then had a disagreement about what to do.

Marcus's position: The high quality metrics in Riverside justified the lower volume. "If I'm hitting the right doors and getting a 24% conversion rate, that's better than hitting every door and getting 19%. We're doing good work. Give us more time."

Yolanda's position: "Marcus, 61% of goal with 15 days left is a math problem. If we maintain current pace in Riverside, we finish at 81% of goal. That's 5,300 contacts short. In a race this close, that's potentially 1,000 or more votes we didn't get."

Nadia's data-informed assessment: The volume shortfall was real and significant. But the quality advantage was also real: Riverside canvassers were making better use of their contact opportunities than canvassers elsewhere. The question was whether the campaign should prioritize increasing volume (accepting some targeting quality reduction), improving canvassing efficiency (changing hours and methods to reduce not-home rate), or increasing resource allocation to Riverside from other counties.

What Nadia Found in the Prioritization Data

Nadia ran the prioritization tool filtered to Riverside's uncontacted voters. The output revealed something unexpected: the top 4,000 uncontacted priority voters in Riverside were concentrated in three zip codes — all adjacent to each other — that had been receiving very low canvassing coverage. The campaign's canvassing routes had been following the voter file's geographic clustering algorithm, which had distributed canvassing across the county's footprint rather than concentrating it in the highest-density areas of high-priority uncontacted voters.

In other words, Riverside canvassers were traveling more and knocking fewer doors because the routing algorithm wasn't optimized for density. A manual routing adjustment — concentrating effort in the three high-density zip codes — would reduce travel time and increase contacts per hour, potentially closing most of the volume gap without any additional resource allocation.

The Resolution

Marcus accepted the routing adjustment. The campaign's data director spent two evenings rebuilding the Riverside canvassing walk lists, concentrating the final 15 days' effort in the three priority zip codes. The canvassing supervisor in each of those zip codes was briefed on the rationale.

In the final 15 days, Riverside's daily contact rate increased by approximately 28%. The county finished at approximately 87% of its contact goal — still short, but within a range the campaign considered acceptable given the quality metrics.

Discussion Questions

1. Nadia's dashboard revealed both a volume problem (61% of goal) and a quality advantage (high targeting lift and conversion rate). How should campaigns weight volume vs. quality in their evaluation of contact programs? Is there a formula, or is this fundamentally a judgment call?

2. The routing adjustment increased contacts per hour without adding any resources — it was purely an analytical improvement. What does this suggest about the relative ROI of analytical work vs. resource allocation in field campaigns?

3. Marcus Gould's position — "I'm doing good work, give me more time" — is a common response from field directors when data challenges their performance. How should campaign data analysts handle this kind of tension between quantitative evidence and field judgment? What would have been the wrong way to handle the Tuesday call?

4. The not-home rate difference (28.4% in Riverside vs. 22.7% statewide) pointed to a structural difference in Riverside's workforce. What are the equity implications of this difference? Manufacturing and logistics workers — who are disproportionately working-class and non-white — are structurally harder for campaigns to reach. Does the standard contact model disadvantage these communities? What could campaigns do about it?

5. The campaign finished Riverside at ~87% of goal. In a hypothetical where the race was decided by fewer than 5,000 votes statewide, could the 13% shortfall in Riverside contact be causally attributed to the routing problem identified by the dashboard? What additional information would you need to make this attribution?

Quantitative Analysis

Using the data provided:

a) At Riverside's actual pace (61% in 20 days), project the total contacts if pace is unchanged for the final 15 days.

b) With the 28% pace increase, project the final total. What is the percentage of goal reached?

c) If each contacted voter in Riverside has a 24.2% chance of being converted to a confirmed/soft supporter, and each confirmed supporter has a 72% chance of actually voting, how many additional votes does the routing adjustment generate compared to maintaining the original pace?

d) If the race in Riverside is decided by a margin of 8,000 votes, and Garza needs a net margin of +12,000 in Riverside to offset her expected deficit in other counties, does the routing adjustment matter to the outcome? Show your reasoning.