By SoundThinkingLearning Objectives
- Understand how SoundThinking's acoustic gunshot-detection sensors and machine-learning classification work
- Evaluate the contested effectiveness studies fairly, including the vendor's rebuttals
- Recognize the equity, evidentiary, and resource-allocation concerns behind recent city decisions
What Is SoundThinking?
SoundThinking — renamed from ShotSpotter in 2023 — operates rooftop networks of acoustic sensors that use machine-learning classification to detect and locate suspected gunfire, alerting police within seconds. Its broader SafetySmart suite also includes CrimeTracer (search), CaseBuilder (investigations), ResourceRouter (patrol allocation), SafePointe (weapons detection), and PlateRanger (automated license-plate recognition, powered by Rekor).
The company is publicly traded. It reported record 2025 revenue of $104.1 million (up 2 percent year over year) but a widening GAAP net loss of $9.4 million. Its footprint has been contested at the city level: it lost its flagship Chicago contract in 2024, Cambridge, Massachusetts voted to remove it in May 2026, and Detroit narrowly voted to extend its contract in June 2026 (about $2.1 million).
💡Key Concept
Acoustic gunshot detection: Sensors mounted on rooftops and poles listen for loud impulsive sounds. Machine-learning models classify whether a sound is likely gunfire and triangulate its location from the timing across multiple sensors. An alert then goes to police — often before, or instead of, a 911 call. The technology's promise is faster response; the debate is over how often those alerts correspond to actual gun crime.
Key Capabilities
- Acoustic sensor network — rooftop microphones detect loud impulsive sounds across a coverage area
- Machine-learning classification — models distinguish likely gunfire and locate it in seconds
- SafetySmart suite — CrimeTracer search, CaseBuilder investigations, ResourceRouter patrol allocation
- SafePointe — weapons detection for venues and facilities
- PlateRanger — automated license-plate recognition powered by Rekor
⚠️Warning
Effectiveness is genuinely contested — here is both sides. A 2021 study by Northwestern University's MacArthur Justice Center in Chicago found that about 89 percent of ShotSpotter alerts led to no gun-crime evidence on arrival, and about 86 percent led to no report of any crime. Cambridge's own decade of data showed only about 35 percent of alerts were confirmed as actual gunfire. Critics also cite over-deployment in Black and brown neighborhoods, officers diverted to dead-end alerts, and evidentiary and wiretapping concerns. Important balance: these figures measure alerts without recovered evidence, which is not identical to a technical misclassification rate — an alert can be genuine gunfire that leaves no recoverable casing or victim on scene. SoundThinking disputes the framing, citing high contractual accuracy guarantees (around 97 percent) and human review of every alert before it is sent. Both the criticism and the rebuttal deserve to be weighed.
Company Details
| Detail | Info |
|---|---|
| Company | SoundThinking (NASDAQ: SSTI), formerly ShotSpotter |
| Founded | 1996 |
| Headquarters | Fremont, California |
| 2025 revenue | $104.1 million record (up 2% year over year) |
| 2025 net loss | $9.4 million GAAP (widening) |
| Contract changes | Lost Chicago (2024); Cambridge voted to remove (May 2026); Detroit voted to extend ~$2.1 million (June 2026) |
| Product suite | Acoustic gunshot detection plus SafetySmart (CrimeTracer, CaseBuilder, ResourceRouter, SafePointe, PlateRanger) |
| Website | soundthinking.com |
Key Takeaways
- SoundThinking, formerly ShotSpotter, runs acoustic sensor networks that use machine learning to detect and locate suspected gunfire, with a broader SafetySmart public-safety suite
- The public company posted record 2025 revenue of $104.1 million but a widening loss, amid city-by-city contract fights (Chicago, Cambridge, Detroit)
- The honest caveat: studies found most alerts led to no gun-crime evidence, and critics cite equity and evidentiary concerns — but the vendor accurately notes these measure alerts without recovered evidence, not a misclassification rate, and cites human review of every alert
