Why does a single public company show up under more than one name? This note explains entity resolution with Alphabet and its Google tickers as the example, and describes the splits a dataset suffers when the names stay unlinked.
The string is not the company
Filings, articles, and quote files do not share one label for a company. A legal name, a trading name, a ticker, and a former name can all refer to the same issuer. A dataset that groups rows by the raw string will treat those labels as different firms. Counts will be split. A history will look shorter than it is. A join between text and prices will drop the rows whose labels do not match exactly, and the drop will look like missing data rather than a naming problem.
The link has to be built on purpose. A stable identifier, such as a filing identifier or an exchange issue identifier, is a better key than a ticker. Tickers are reused, retired, and shared across share classes. Names are edited by writers. An identifier still changes when the corporate structure changes, so the map from identifier to identifier over time is part of the data, not a one-time cleanup. Without that map, last year's rows and this year's rows are strangers.
Entity resolution is the recorded link among the legal name, the trading names, the tickers, and the former names that refer to one issuer over stated dates.
Alphabet, Google, and two tickers
Alphabet is the public example. The company is widely called Google in ordinary writing, while the listed parent is Alphabet. The issuer has more than one share class, commonly quoted as GOOGL and GOOG. A news sentence that says Google, a filing that says Alphabet, and a price file that stores two tickers are one economic story and several keys. A study that picks only GOOGL has chosen a share class. A study that adds the two tickers' volumes without a rule has mixed classes. A study that searches articles for the string Alphabet and ignores Google has missed the way people write.
The same company can also change name. A reader looking for the current name will not find older articles that use the earlier one. The point-in-time version of the map says which name was current on which date, and which names were aliases rather than different issuers. Alias is not a fuzzy guess. It is a recorded decision: these strings denote this issuer for this span. A later correction of that decision should be stored as a correction, with its own date, so a replay does not silently inherit the fix.
Share classes are the detail that makes the Alphabet case useful. The classes can move differently, and they do not represent the same claim on votes. Collapsing them into one price series is a modeling choice. Leaving them apart and then speaking as if there were one price is also a choice, just an unstated one. The dataset should be able to answer a plain question: which class is this row? If it cannot, the row is not ready to be joined to a filing that describes the whole issuer.
What the map does not decide
Resolving names does not say whether the company is attractive, liquid, or related to another firm in a business sense. It says which records are the same issuer. Two suppliers can still disagree on a subsidiary, a joint venture, or a spun-off unit. Those disagreements should remain visible as unmatched rows, not be forced together by a loose string match. A loose match will happily join a company to a different company that shares a word.
A researcher accepts the map the way a researcher accepts any other feature. The acceptance names the identifier, the share class, and the date the alias became known. Alphabet is the example because the gap between the spoken name and the listed name is public and easy to see. The same gap exists, quieter, for every issuer with a former name or more than one class. The work is the map. The map is not a view about the business.
This piece does not claim that a name map is finished, that string matching is good enough, or that any share class should be preferred. It describes why one company becomes many rows, and what has to be linked before the rows are one history.
