Kali, the Data Company

← All posts

Field Collection at Scale

Dozens of sites, one standard. The systems that keep field data consistent when the field does not cooperate.

A single recording session is a logistics problem. Fifty sessions across a dozen cities is a systems problem. Most of the work behind Kali is not the camera; it is the machinery that makes a Tuesday in Nairobi comparable to a Tuesday in Manila.

One playbook, many sites

Every collection program starts as a written playbook: a site profile, a kit list, camera positions, a capture script, and a calibration routine. Field leads are hired locally, trained against recorded examples, and certified on the playbook before their first paid session.

The playbook is deliberately boring. Boring is what survives a power cut, a shifted lunch break, and a manager who changes their mind about where the camera can go.

Instrumenting the work

Kits are built around a fixed capture layout: two or three camera angles, continuous audio, a wearable where the task requires it, and environmental sensors for light, temperature, and motion. All streams are synchronized at the start of a session, and the kit records a calibration slate that review can check later.

Uploads happen daily. A session that sits on a device for a week is a session that can be lost, corrupted, or recorded over.

Keeping quality constant

Each site gets a morning calibration check, a midday spot check, and a place in the remote review queue. Reviewers score framing, audio, sync, and task coverage. A site that drifts below threshold gets a call, a refresher, or a replacement kit before the next session is scheduled.

The scorecards are shared with field leads. The point is not to catch people out; it is to make the standard visible while there is still time to fix it.

When the field does not cooperate

Weather moves sessions. Permissions arrive late. A site contact forgets that a crew is coming. We plan redundancy into every schedule: backup sites, backup operators, and a rule that a missed session is cheaper than a bad one.

What comes next

We are expanding the network into new regions this year, adding collection protocols in more languages, and extending sessions to cover work that runs for hours rather than minutes. The playbook grows with it.

Work with our data.

Join the waitlist →