Ticker Classification

Your agent labels each position with its sector and industry, then re-runs the attribution group by group to show whether the edge holds inside each one.

Two labels per ticker, and nothing else.

Classification answers one question: what kind of business is this company in? This dataset answers it twice for every ticker. Once at the sector level, eleven broad groups such as technology, finance, and healthcare, and once at the industry level, twenty-five narrower groups such as semiconductors, banking, and pharma_biotech.

There is no date dimension. Each row is one ticker and its two labels, ordered by ticker, and the labels are stored rather than inferred when you query. Values are lowercase and underscore-separated, which is what makes them safe to pass in a URL, keep in a column, and compare with a plain string match.

The peer group is what makes a signal mean something.

Most cross-sectional signals are partly a sector bet until you make them not one. Rank the whole market on a momentum or value measure and the top decile is often one or two industries, so the strategy lives or dies on those industries rather than on the measure you were testing. Grouping by alphanume_sector and ranking inside each group separates the two, and the same grouping tells you afterward where the return actually came from.

The second use is universe construction. An event study on semiconductors, a pairs screen inside banking, a dilution screen that leaves out reits: each one starts with a filtered ticker list. This endpoint returns that list in a single call, instead of a hand-kept list that silently goes stale as companies list, merge, and change what they do.

• Rank a signal inside alphanume_sector to strip the sector bet out before selection.
• Pull sector=technology or industry=semiconductors as the starting universe for an event study.
• Attribute a backtest's returns by group to see whether one industry carried the whole result.
• Build industry-relative measures by comparing each name with the median of its own alphanume_industry.
• Cap exposure per sector when sizing a basket, using the label as the grouping key.

Assigned from business activity, fixed at write time.

Each ticker is assigned to a sector and an industry from the underlying business activity, using Alphanume's own taxonomy rather than a licensed standard. The groups were sized to be coarse enough that every name lands in a populated bucket and fine enough that the buckets still separate: eleven sectors, twenty-five industries, no long tail of near-empty codes.

Assignments are served as stored values, not recomputed at query time, so two runs of the same query return the same labels. The mapping reflects the present state of each ticker and moves only when a company's business changes enough to warrant it. An invalid sector or industry value returns a 400 rather than an empty list, which makes a typo in a pipeline loud instead of silent.

Three columns and the filters that read them.

The headline fields below are a subset. Every field, with exact types and semantics, is documented in the API reference.

One call returns a whole sector.

One key works across the REST API, the hosted MCP server, and this dashboard. Every response is JSON with a { count, data } envelope.

import requests

url = "https://api.alphanume.com/v1/ticker-classification"
params = {
    "api_key": "alp_abc123",
    "sector": "technology"
}

r = requests.get(url, params=params)
print(r.json())

curl "https://api.alphanume.com/v1/ticker-classification?sector=technology&api_key=alp_abc123"

{
  "count": 2,
  "data": [
    {
      "ticker": "AAPL",
      "alphanume_sector": "technology",
      "alphanume_industry": "hardware_devices"
    },
    {
      "ticker": "MSFT",
      "alphanume_sector": "technology",
      "alphanume_industry": "software"
    }
  ]
}

What this data does not claim.

• Sector weights computed from this mapping will not match an index provider's, because the taxonomy is not theirs.
• There is no history. The mapping is current, so a company that changed businesses is labeled by what it is now, not by what it was during your backtest window.
• One ticker gets one sector and one industry. A company with two real businesses is placed in the dominant one, and nothing in the row tells you it was a close call.
• A ticker missing from a response is unclassified rather than misclassified. Write joins that expect misses instead of assuming full coverage.
• Invalid sector or industry values return a 400 error rather than an empty result set.