Areas of Expertise

Bespoke Tech Tools

Software built for one campaign's specific problem.

Why this is possible

The hard parts already exist

Modeling and Insights run on code that is written and maintained anyway: the voter file pipeline, the response collector, the precinct math, the page that renders a readout. Pointing that machinery at one race's particular problem is a few hours of work rather than a product launch.

Larger firms rarely build these, and the reason is structural rather than technical. Inside a large firm a small internal tool has to become a product: documented, supported, sold to every client, priced to cover all of that. A tool built for one race, for one cycle, carries none of those costs, so it can be built when somebody notices they are doing the same manual task twice.

What that means in practice:

  • Tool work is named in the written scope before anything is built.
  • It arrives as a link. Nothing to install, no app store, no training session.
  • The campaign's data stays the campaign's, and goes back to you at the end of the engagement.
  • It is built for your race, so it is not resold as a product to the next campaign.

Illustrative examples

Three kinds of tool

Illustrative only

The three examples below are illustrative. They describe the kind of tool that can be built for a race like yours. They show the format. They are not accounts of work already delivered.

Turnout tracker for one race

Illustrative. Early and absentee returns loaded each morning and set against the vote goal, precinct by precinct, with the gap stated in votes rather than percentages. On a Tuesday in October the only question is which three precincts are behind, and the page answers it before the morning call.

Precinct level canvassing prioritizer

Illustrative. Model scores, walk history and the campaign's own contact notes in one ranked list, so the volunteer coordinator picks Saturday's turf in two minutes instead of an evening. Doors already knocked come off the list automatically.

Message test readout

Illustrative. One page per wave showing which wording drew responses, which drew none, and every open answer in the voter's own words, filterable by precinct and by whether that person has already voted.

None of the three is complicated. Each one removes a job somebody on the campaign was doing by hand at eleven at night, which is where the time goes in a race with four paid staff or none.

How one gets built

How a tool gets built

  1. The problem gets named in one sentence, usually on a call where somebody describes the same manual task for the second time.
  2. A scope goes in writing: what the tool shows, what it will not show, where the data comes from, who can see it, and the date it is ready.
  3. It is built and handed over as a link, with the data source and the last update time printed on the page itself.
  4. It gets adjusted while the race runs, because the question in October is not the question in May.

Sometimes no tool is needed, because a spreadsheet already does the job or the question will not come up again. You will hear that.

These are small tools for one campaign. They do not come with a support desk at midnight, they are retired at the end of the engagement, and voter data is handled under the same rules as the research itself.

Describe the job you are doing by hand

Send the race, the election date and the task that is eating your evenings. We write back with a scope, a price and an honest answer about whether a tool helps.

Contact Upset Research