Industry two sides: consumers versus founders

94 industries and 18574 deduplicated search words, split into what people ask before they spend and what they ask before they open.

The numbers

What one record is

One row per industry. Columns: the industry name and short name, the number of search roots used, how many of those roots are startup-side and consumer-side, the number of deduplicated search words, the cleaned consumer and startup word counts, the consumer index (cleaned consumer words per consumer root), the startup index, the balance value, the band the balance value falls into, and a flag for whether the consumer reading may be driven by a holiday window rather than by the industry itself.

Records

Every industry is measured on both sides with the same kind of question: what people ask before they spend, and what they ask before they open something. The balance value says which side the words lean to; the consumer index says how hot the consumer side is.

Columns

• name, short, lang - the industry as recorded.
• type - the band the balance value falls into. The three band labels are written in Chinese in the file: both-hot, consumer-side, startup-side.
• balance - consumer words divided by (consumer words + startup words), after moving words that leaked across sides back to the side they belong to.
• consume_index, startup_index - cleaned words per root on each side.
• words_dedup, consume_words, startup_words, roots, roots_startup, roots_consume - the counts behind those ratios.
• external_driver_suspect - true when the collection date fell inside a holiday window and the consumer side dominates, so the reading may be a holiday effect rather than a standing one.

Sample rows

Boundaries

• The balance value compares how many words one kind of question can pull, so it is sensitive to how wide the roots themselves are; the uncleaned word counts are kept in the JSON file for comparison.
• An industry with barely any words on either side is not read as startup-side or consumer-side; it takes at least eight words to be banded at all.
• These are search-wording signals, not revenue. They say which side the questions lean to, not how much money moves.

Download

• industry-sides.csv text/csv — 94 rows
• industry-sides.json application/json — 94 rows, columnar

Files are served from this domain and stay at these URLs. Licence: CC BY-NC 4.0 — attribution required, commercial use needs written permission.

Method & definition

• Source. InfoGap own keyword collector, using the same roots as the industry search demand dataset and aggregating them per industry under the sides_v2 rule. Queries run against Baidu autocomplete (www.baidu.com/sugrec) and 360 autocomplete (sug.so.360.cn/suggest).
• Definition. One row per industry. Columns: the industry name and short name, the number of search roots used, how many of those roots are startup-side and consumer-side, the number of deduplicated search words, the cleaned consumer and startup word counts, the consumer index (cleaned consumer words per consumer root), the startup index, the balance value, the band the balance value falls into, and a flag for whether the consumer reading may be driven by a holiday window rather than by the industry itself.
• Last updated. 2026-10-05
• Update frequency. daily (collector runs 02:00 Asia/Shanghai)
• Time range. see temporalCoverage in this page's schema.org Dataset block.
• Licence. CC BY-NC 4.0 — attribution: InfoGap (infogap.store), link to this page. Commercial use requires written permission.

Where to go next

• Method pages: https://infogap.store/method/how-to-check-our-numbers/
• Hub: https://infogap.store/data/
• Machine-readable: registry.json
• For AI engines: llms.txt — what to cite, and with which definition.