Giving the outsider an edge

In 2020, Rob Mercuri decided to run for the Republican nomination to Pennsylvania’s 28th State House District. A husband, father, and Army Captain, he was running as a first-time political candidate. With party control at stake and a crowded primary ahead, Mercuri knew he had to run a smart and effective campaign to win the seat.

‍

He faced two other Republican contenders in the primary election: Libby Blackburn and Mike Heckmann. As a political outsider, Rob had little understanding of how to chart out a winning campaign. When Speaker of the House Mike Turzai introduced him to Numinar, he immediately saw the value of the integrated voter data, analytics, and contact tools packaged in an intuitive platform.

‍

Self-service polling

The first task for Mercuri’s campaign team was to understand how he measured up against the other Republican candidates in the primary field. Numinar had a solution right out-of-the-box: the world’s first self-service polling tool with just four steps. 

‍

1. Select a randomized, auto-generated list of 5,000 Republican voters.
2. Type the survey questions.
3. Send each voter the survey via text. 
4. Sit back and let the responses roll in.

Numinar’s targeting interface and conversational text surveys

With Numinar, Mercuri’s team didn’t need to spend tens of thousands of dollars and wait several weeks for a full-scale professional poll. They could build and send a poll from his laptop in just a few hours and at a cost of only $500. 

‍

Next-generation modeling

Those survey responses weren’t just used for a one-time topline report. The Numinar AI processed the survey results and automatically generated a customized predictive vote model. The team could immediately see which Republican each primary voter preferred and who were most likely to show up on Election Day. This model also enabled the campaign to target the most likely Mercuri supporters with limited campaign resources. 

Mercuri’s custom primary model among Republican voters

‍

The advanced modeling paid off. Numinar’s AI was strikingly accurate in its projections for each primary candidate’s vote share:

‍

R. Mercuri

L. Blackburn

M. Heckmann

Turnout

Numinar

61%

25%

14%

19.80%

Election Result

63%

20%

16%

18%

Margin of error

2%

5%

2%

1.80%

Primary election results compared to Numinar AI’s model prediction

‍

Was this remarkable accuracy really just remarkable luck? A deeper look at the precinct-level data would suggest otherwise. The average precinct in District 28 has just over 200 voters. Numinar’s median turnout error by precinct was 2.2%. For Mercuri, Blackburn, and Heckmann, the median precinct vote share error was 10.8%, 6.6%, and 4.8%, respectively. That means for each precinct, Numinar’s model was off by only about 20 votes for Mercuri, 13 votes for Libby, and 10 votes for Heckmann.

‍

On to the general election

With a huge primary victory in the bag, Mercuri advanced to the general election to face the well-financed Democratic candidate, Emily Skopov. The race was going to be tight. With a narrowing historical Republican majority in the district, most other pollsters had Mercuri running below Skopov. 

‍

Numinar’s general election model, on the other hand, consistently placed Mercuri ahead by a narrow margin. Over the course of the general election, the team sent over 17,000 text surveys, which allowed Numinar’s machine-learning algorithms to learn with each response. This combination of persistent surveys and advanced analytics gave Mercuri’s campaign team a real-time vote-share tracker through Election Day. 

Numinar’s live vote share tracker leading up to Election Day

‍

On November 3^rd, Mercuri won a close race with 53.8% of the vote against Skopov’s 46.2%. Numinar’s final prediction was right on the money, with Mercuri at 52.2% and Skopov at 47.8% yielding a 3.1% margin of error (compared to FiveThirtyEight’s average 2020 presidential polling margin of error of 4%). 

‍

Numinar AI’s general election model prediction compared to actual results

‍

Numinar’s groundbreaking analytics and campaign management platform were crucial components of the victory at each step of the race. Skopov outspent Mercuri’s team $826,828 to $345,865, but, in the candidate’s words, it was as though the Democrats were “throwing money away” without the right data-driven strategy. The results proved that money can’t always buy elections.

‍

Reflecting on the results he achieved with Numinar, newly-elected Rep. Rob Mercuri is bullish on it as the new frontier of campaign technology: “I’m a big believer in Numinar. I think Artificial Intelligence and modeling are the right approach.” 

‍