Skip to content
Alshival.Ai

Patterns in public data

NHTSA Complaint Explorer.

Find related experiences. Follow a pattern. Read the accounts behind it.

—complaints
—clusters
—model years

Require these words. Every added word must occur.

Loading snapshot…
Refine complaints

Filters apply as you change them within the current map view.

Vehicle
Reported incident
Each dot is a complaint
Hover to explore · Double-click a group to open it

Explore the complaint atlas

Start with a company or describe an issue. Follow a cluster, then read the original accounts.

  1. Navigate the map. Drag to rotate and scroll or pinch to zoom. On touch screens, choose Pan to move the atlas, and use + and − for precise zoom. Scroll outside the map to move down the page. Double-click a dot to open its company or group. The list’s Open links also work by touch and keyboard. Copy the address bar to share a company, cluster, search, or filtered view. You can also open these links in a new tab. Full screen expands the explorer; use Exit full screen or Escape to return.
  2. Read an account. Click a dot to read its ODI details in the right sidebar. Click another dot to replace the details; × returns to the list. On a phone, tapping a dot opens a scrollable account below the atlas, keeping the map visible. Tap another dot to replace it, or × to close it. The Results tab opens the full company, cluster, or complaint list. Hover shows only the company or cluster.
  3. Find an issue. Keywords require every added word; semantic search finds similar descriptions. A higher similarity cutoff narrows the map.
  4. Interpret carefully. Nearby dots suggest related accounts. Similarity and reported incidents do not establish a vehicle defect. Search lists show up to 100 ranked matches. The map uses the same cutoff; large mobile views show a preview.

How to read this map

Related accounts, shared patterns.

Nearby dots suggest similar narratives. Screen distances remain approximate.

Read the original account.

Terms are distinctive words and phrases extracted from each group, with common filler excluded. Search relevance is a ranking, not a confidence score. Similarity does not establish a defect or confirm an incident. Injury and death filters use reported counts.

A defined snapshot.

The published atlas refreshes after completed ingestion and validation. Company views keep the same coordinates, camera, and scale as All companies, with colors showing company-specific clusters. Filters remove dots without rearranging them.

Download data on Kaggle ↗

Technical settings