Brand, Merchant and Subsidiary Mapping

A credit card transaction feed shows a merchant string of "SBUX #4521 SEATTLE WA." A researcher building a Starbucks revenue nowcast needs to know that this is a Starbucks-operated store, not a licensed store inside a grocery chain whose revenue Starbucks barely sees, and that the same feed also contains "STARBUCKS COFFEE CO," "TEAVANA," and dozens of other strings that may or may not belong to the same public company's consolidated revenue. This is entity mapping, and it is one of the least glamorous, most error-prone parts of working with alternative data — and one of the most consequential, because a bad mapping silently corrupts every signal downstream of it.

Why it's harder than it sounds

Public companies are not single, clean brands. A single ticker can own dozens of consumer-facing names through acquisition, licensing, and joint ventures, and the revenue attribution between the parent and each brand varies enormously:

• Wholly owned subsidiaries whose revenue flows fully into the parent's consolidated statements — straightforward once identified, but the brand name on a receipt or app store listing rarely matches the legal entity name.
• Licensed or franchised brands where the public company earns only a royalty or franchise fee, not the transaction's face value — treating a franchise's card swipe as full revenue overstates the read by an order of magnitude.
• Joint ventures and minority stakes where the parent consolidates none, some, or all of the JV's revenue depending on ownership percentage and accounting treatment, information usually buried in a 10-K footnote rather than in any data vendor's mapping file.
• Multi-brand portfolios where a spending panel might show ten merchant strings that all belong to one holding company, and a researcher unaware of the corporate structure treats them as ten unrelated data points instead of one company's ten revenue lines.

A worked example

Consider a consumer holding company whose 10-K describes three reportable segments built from six underlying brands, two of which are wholly owned, three of which are majority-owned joint ventures consolidated at 100% under accounting rules despite a minority partner holding real economic interest, and one of which is a licensed brand contributing royalty income only. A naive mapping that treats every merchant string bearing one of these six brand names as "100% revenue to the parent" would overstate the licensed brand's contribution by roughly the multiple between royalty and gross sales — often ten to twenty times — while correctly capturing the wholly owned and consolidated JV brands. The fix requires going brand by brand through the segment footnotes, tagging each with its correct revenue-recognition treatment, and weighting the alt-data signal accordingly before it ever reaches a forecasting model.

Brand string in raw data	Corporate relationship	Correct treatment
Wholly owned flagship brand	100% owned subsidiary	Full transaction value counts toward revenue
Consolidated joint venture brand	Majority-owned JV, consolidated	Full transaction value counts, footnote minority interest
Licensed/franchised brand	Franchise agreement	Only royalty percentage counts, not gross sales
Recently divested brand	Sold mid-sample	Exclude from mapping after divestment date

In practice

• Build the mapping as point-in-time, not static. A brand that was wholly owned in 2019 and divested in 2022 needs a mapping that changes on the divestment date, or the signal quietly mixes in a company's data after it stopped belonging to the ticker being traded. This is the alt-data version of the same discipline covered in Point-in-Time Data.
• Prioritise mapping effort by revenue materiality. A brand contributing 40% of segment revenue deserves careful footnote research; a brand contributing 0.5% is rarely worth more than a rough estimate.
• Cross-check against a second source. Institutional holdings and analyst models sometimes carry their own brand-to-segment breakdowns; see 13F and Institutional Holdings Signals for a related mapping problem on the ownership side rather than the revenue side.
• Keep a change log. When a mapping is revised, note why and when, so a backtest run today and one run after next quarter's 10-K don't silently disagree about history.