Standing research · CPG

CPG research

Unit economics read from the filings themselves, against what the inputs actually cost. Free, public, dated, and almost nobody reads it this way. Each study below registers its question before any data is pulled, publishes every row, and is re-run on a schedule.

AI assistants cited General Mills’ websites in 0 of 94 answers

The question: When shoppers ask AI assistants which cereal, cake mix, taco kit, granola bar, dog food, frozen snack, soup or mac and cheese to buy, whose websites do the answers cite?

What it found: General Mills’ own sites were cited in 0 of 94 answers across 8 categories. No brand site made any top 10: a machine availability gap. Publishers such as tastingtable.com, health.yahoo.com and dogster.com owned three categories outright.

8 categories · 94 AI answers · 7 search sets · 7 registered predictions · published 2026-11-19 · re-run quarterly

Pet insurance grew 20% in 2025. More than half of it was price.

The question: How big is the US pet insurance market, is it growing on more pets or higher prices, and who holds it?

What it found: About $5.7 billion in US premium in 2025, up 19.7%, on 6.98 million pets, up 9.0%. Price made up about 52% of the growth, against 19% in 2023. Trupanion, Nationwide and JAB lead.

NAPHIA 2022–2025 · Trupanion FY2023–FY2025 · 4 registered predictions · published 2026-11-17 · re-run when naphia publishes each june

Pet owners know insurance exists. They doubt it pays.

The question: Fewer than one US dog or cat in twenty is insured. Is the gap awareness, or doubt, and does selling cover through a pet retailer close it?

What it found: 4.27% of US dogs and cats are insured. Owners search for the price and how it works about four times as often as whether it is worth it, but in forums and reviews the argument is whether it pays, and a denied claim is the top complaint. Costco’s member discount on Figo draws buyers; Chewy’s store-brand plan mostly doesn’t.

559 reviews of 7 insurers · 28 searched terms · 12 assistant answers · 2 registered predictions missed and named · published 2026-11-12 · re-run one-time study

Most line extensions are too small to lift their parent

The question: An extension's business case often counts a halo on the parent brand. Does launching an extension lift searches for the brand it came from?

What it found: Of 45 extensions screened across the same aisle, a new aisle and licensed partners, 4 drew enough attention at launch to test, and licensed drew 0 of 11. None of the tested parents showed a 20% lift that a year without a launch did not also show. AI assistants named the extension for 6 of 7 parents, but licensed ones were never described from the owner’s site, and in reviews licensed products mostly replaced the restaurant visit.

45 extensions · 3 search passes, 2 voided by their own design and named · 14 assistant answers · 30 coded reviews · published 2026-11-11 · re-run one-time study

Kraft Heinz is not in the answer to its own category’s questions

The question: A content budget is spent on the assumption that publishing good answers to category questions reaches the household asking. Does it?

What it found: Across nine answers to the questions the US meals category is actually asked, the brand held none. On the largest assistant nothing it publishes was read at all — 318 to 321 input tokens across 2,404,500 monthly searches. Where an assistant did read, on a question about the brand’s own two products, it cited a home-decor site and a salad guide ahead of the brand’s pages. Perplexity cited the brand’s own domain first, so the lever exists — on the operators almost nobody uses.

254 candidate phrases · 21 distinct question concepts · 4 assistants · one run per question · two hand-written question lists discarded against the demand data before anything expensive ran · published 2026-10-28 · re-run when an operator changes how it retrieves

Nobody stopped caring. They stopped buying.

The question: Did the Halloween occasion shrink, or did the basket shrink inside an intact occasion?

What it found: The occasion is at a six-year high and the basket is not. September search for the category ran 15.5% above its own 2019–2024 average while sales fell 6.7%, and decorations fell 58.6%. Across 409 comments about trick-or-treating, “skipping” appeared zero times.

5 search terms since 2019 · 4 comment corpora + a costume control · 2 of 5 registered predictions wrong and published · published 2026-10-26 · re-run as soon as october closes

Costs did not fall. They split.

The question: Did the gross-margin recovery of the last three years come from input costs falling, or from prices rising?

What it found: Neither, for most of the category. Food and household gross margin fell 0.92 points while a control outside it rose 1.99 — because cocoa rose 109% while soybean oil fell 27%.

1,633 filers screened · 42 category + 15 control · 8 commodities · 2 of 4 registered predictions wrong and published · published 2026-10-25 · re-run each filing season

What this category covers

Margin and input cost read from the filings themselves, and what the category's buyers are actually asking, measured against what the answers contain.

What it does not

Filings lag by a quarter or more, and search demand is attention rather than purchase. Neither one sees a shelf price on the day a shopper stands in front of it.

When each study is read again

  • AI assistants cited General Mills’ websites in 0 of 94 answers — Quarterly. Last read 2026-11-19.
  • Pet insurance grew 20% in 2025. More than half of it was price. — When NAPHIA publishes each June. Last read 2026-11-17.
  • Pet owners know insurance exists. They doubt it pays. — One-time study. Last read 2026-11-12.
  • Most line extensions are too small to lift their parent — One-time study. Last read 2026-11-11.
  • Kraft Heinz is not in the answer to its own category’s questions — When an operator changes how it retrieves. Last read 2026-10-28.
  • Nobody stopped caring. They stopped buying. — As soon as October closes. Last read 2026-10-26.
  • Costs did not fall. They split. — Each filing season. Last read 2026-10-25.

A study is corrected in place and dated when a re-read moves it. The correction is published; it is not applied quietly.

Keep your own calls on record

DEOS is the decision app built on this research. Talk through a plan or a decision, and it writes the record, checks it against your past calls and brings it back when it is time to review.

Start free

Free for 30 days, no card. Then $99 a month.

Or see the evidence for this category’s decisions in the Decision Brief →

CULT+MATH LLC · Houston, Texas · Ipalibo Da-Wariboko
Standing research. Figures are re-measured on the schedule stated on each page and the page is corrected in place, dated. Corrections are published, not applied quietly.
Cite as: CULT+MATH LLC, Ipalibo Da-Wariboko. Attribution to cultmath.com is requested.
Studies signed by a desk are made by the research desks and published by Ipalibo Da-Wariboko.