A working thesis

In CPG, brands are increasingly built to be acquired or replaced within a few years rather than held for decades.

The data in this deck shows the compression. The argument is about who captures the value when that's the new normal.

Press → to advance.

Where this is going

Two acts

Act I: What happened to the lifecycle of a CPG brand. Six charts, four from a database of 383 brands that appeared on credible breakout lists 2021–2026.

Act II: What it means for who actually captures value. The retailer-as-house-of-brands argument, and what it implies for the People Inc. acquisition strategy.

Act I

What happened to the lifecycle of a CPG brand

Chart 1

Fewer giant CPG hits are coming from each new founding cohort

Founding-year distribution of brands appearing on credible breakout lists (Bain Insurgent, Numerator, Circana, Inc 5000 F&B, Pear Commerce, Food Institute), 2021–2026 editions. "Independent" means the brand was an independent founding, not a line extension. Pre-2000 founders are bucketed in a footnote below the chart.

Founding-year distribution of brands that later made breakout lists. The 2010s built most of the inventory the 2020s breakout lists draw from. The 2020+ founding cohort is visibly thinner, even allowing for the fact that very recent foundings haven't had time to break out yet.

Chart 2

When they do break out, they break out faster

Years from first retail launch to first appearance on a credible breakout list, by retail-launch cohort. Independent brands only; extensions excluded. Black bar shows the cohort median. A 2020+ brand reaches list-recognition in roughly a third the time a 2010-14 brand did.

A 2020+ brand reaches breakout-list recognition about 3 years after first hitting retail. A 2010-14 brand took 11.

What's behind it: TikTok algorithmic distribution, retail-buyer category cycles getting shorter, and SKU-velocity expectations rising. A brand no longer needs a decade of word-of-mouth to register on these lists.

Chart 3

And exits come earlier in life

Years from founding to acquisition, by founding cohort. Each dot is one acquired brand (post-2000 founders). Black bar shows the cohort median. Pre-2000 founders are excluded; outliers like Pillsbury (acquired 129 years after founding) would flatten the modern cohorts to invisibility. The 2020+ cohort has had less calendar time to be acquired, so its median is biased downward, but year-zero exits aren't an artifact of that bias.

Years from founding to acquisition by cohort. Median: 21 → 13 → 11 → 6 → 3.

When RXBAR sold to Kellogg's in 4 years, that was unusual for the 2010s. By the 2015–19 cohort (n=41), 6 years was the median. The 2020+ cohort even includes year-zero acquisitions: brands acquired in their founding year by Mammoth Brands, Target, ZURU.

Chart 4

And no, the compression isn't a censoring artifact

Kaplan-Meier curves for time-to-acquisition, by founding cohort. Each line shows the probability that a brand from that cohort is still independent after N years. A steeper drop means faster acquisition. The method handles right-censoring properly: still-independent brands contribute up to today's age rather than their full eventual lifespan. By year 5, 19% of the 2015–19 cohort had been acquired, against 3% for 2010–14 and 6% for 2005–09.

Kaplan-Meier curves accounting for right-censoring. By year 5, 19% of the 2015–19 cohort had been acquired, vs. 3% for 2010–14 and 6% for 2005–09.

The 2015–19 cohort has 83 brands and 7+ years of follow-up. The finding stands without the 2020+ data point. The compression is real.

Why this happened

The launch economy bifurcated

⚠ Illustrative only. These numbers are invented, not measured. The stack covers pre-launch costs to put a credible-looking CPG SKU into market: branding, packaging, photography, storefront, ad creative, contract-manufacturing minimums. Numbers are rough order-of-magnitude estimates pieced together from agency rate cards, Shopify Plus pricing, and AI-tool pricing as of 2024, against equivalents from around 2010. No single source ties the stack together. Distribution costs (slotting, broker, marketing-to-scale) are not shown; those didn't fall, and likely rose.

Numbers above are illustrative composites, not measured.

Pre-launch costs collapsed: AI imagery, contract manufacturing, Shopify+TikTok, branding kits for the price of one designer-week. The cost to put a credible-looking SKU into market fell roughly an order of magnitude over the past decade.

Distribution costs did not collapse. Liquid Death and Olipop spent $50M and $90M+ respectively to reach scale. Slotting fees, broker relationships, and velocity requirements at major retailers got harder, not easier.

The result is an asymmetry: launching looks cheap, surviving costs the same as it did. So more brands attempt launch with insufficient capital to clear retail-velocity gates, and the failure happens at shelf instead of at concept. That's where the compressed lifecycles come from.

One reasonable objection

"But AI is making feedback faster — won't that change everything?"

The single-product gain is real: ~5 months out of an 18-month cycle.

The portfolio outcome is closer to zero.

Why faster nets to zero

The constraint isn't stage duration

⚠ Illustrative. Three stacked timelines. Top: a single product's pre-market cycle compresses with AI feedback (concept → iteration → package test → launch prep), saving roughly five months of an eighteen-month cycle. Middle and bottom: portfolio cadence over a 24-month window. Pre-AI shows four launches. AI-enabled shows four launches in the same window — internal gestation is shorter, but launch cadence is unchanged because the binding constraint at portfolio level is concurrent organizational bandwidth, not stage duration. Cycle counts and durations are typical of a major CPG category team, not a specific company.

Same window. Same launch count. Faster cycles don't widen the funnel.

The binding constraint is concurrent organizational bandwidth — reviewers, agency capacity, retail-buyer slots, trade-promotion budget, category-line review calendars. None of those got wider when AI feedback got faster.

So the saved time gets reabsorbed:

Extra iteration on the same concepts
Slack the org doesn't fill (next slot is already booked)
Competitive parity once every operator has the same tools

Better synthesis at the concept phase could shift outcomes. Different capability, separate question — deck doesn't take a position.

Act I summary

Brands as economic campaigns

What this all adds up to: the rational unit in CPG is shifting away from "build a brand and hold it for 30 years" toward something more like "spin one up, ride the slope, exit at peak." Operators are starting to plan for this explicitly.

That changes who captures the value, which is the rest of the deck.

Act II

Who captures the value when brands are ephemeral

Chart 5

A small set of strategic acquirers does most of the harvesting

Acquirers with two or more breakout-brand acquisitions in the dataset. The top eight absorbed roughly 20% of all acquisitions. A long tail of one-off strategic and PE buyers accounts for the rest. Most of the repeat buyers (Hershey, PepsiCo, Keurig Dr Pepper) are themselves channel operators with retail distribution that benefits from cycling brands through.

The repeat acquirers in our dataset. Hershey leads with 6 acquisitions, then PepsiCo with 5, then Keurig Dr Pepper, Bansk, L Catterton, Church & Dwight, Molson Coors, ZURU.

These are channel-tier operators. They own the shelf relationships and the retail muscle that benefits from cycling brands through. Discovery happens at the insurgent; consolidation happens at the strategic. The acquisition is the hand-off between tiers.

The structural pattern

Insurgents pay to teach the category. Retailers harvest it.

Conceptual diagram. The insurgent S-curve climbs through demand discovery: early adopters, taste-makers, then mass. At the inflection point, the private-label curve diverges and harvests the educated demand at no discovery cost. The insurgent paid to prove the category exists; the retailer captures the second half of the lifecycle with a Kirkland version.

The insurgent S-curve climbs through demand discovery: early adopters, then taste-makers, then mass. At the inflection point, the private-label curve diverges and harvests the educated demand. The insurgent paid the cost of proving a category exists; the retailer launches a Kirkland or Good & Gather version and captures the second half of the lifecycle without that cost.

When launching is cheap, the channel beats the IP.

The bigger threat

Private label is growing four times faster than national brands

H1 2025 dollar growth: +4.4% private label vs. +1.1% national brands, per Circana.

Private-label share at the major chains:

Trader Joe's: ~70% of sales
Costco Kirkland: ~33% (up from the high 20s, growing roughly half a point per year)
Target Good & Gather: ~25%
Walmart Bettergoods: launched 2024 and scaling

Trader Joe's launches around 1,500 SKUs per year and rotates them constantly. The brand customers trust is Trader Joe's itself; the individual products are ephemeral inside an immortal channel. Costco does a more selective version of the same play.

The retailer captures the demand the insurgent paid to discover, at zero brand-discovery cost.

Chart 6

What survives the cycle: brands private label can't replicate

Brands at least 7 years old, still independent, appearing on multiple breakout lists. X-axis is years since founding; Y-axis and dot size are breakout-list appearances. The brands that survive the cycle each have something private label can't credibly clone: personality (Liquid Death), regimen lock-in (AG1), functional positioning and IP (Olipop), founder narrative (Magic Spoon).

The 49 brands in our dataset that are 7+ years old, still independent, and appear on multiple breakout lists. The pattern is consistent. Each survivor has a moat private label can't credibly clone:

Liquid Death's personality
AG1's daily-regimen lock-in
Olipop's functional positioning and IP
Magic Spoon's founder narrative and category creation

The implicit rule: anything a Kirkland version could replicate, a Kirkland version eventually will. Anything it can't, the brand keeps.

The strategic frame

Two tiers won. The middle got squeezed.

Where value is captured when brands are ephemeral. The two diagonal quadrants both win: insurgent tier (high discovery risk, low channel ownership) captures the upside of demand discovery; channel tier (low discovery risk, high channel ownership) captures the durable customer relationship. The middle quadrants, where most legacy CPG brand IP and media-brand licensors sit, capture neither.

The diagonal wins. Insurgents capture the discovery upside, channel operators capture the distribution upside. The middle — brand IP without channel ownership — captures neither. That's where most legacy CPG brand equity, mid-market PE rollups, and media-brand licensors sit.

The People Inc. case

What this implies for the Diller bet

People Inc. is buying nostalgic brand equity — Food & Wine, People, the rest — in an era when that asset class is structurally squeezed. The acquisitions don't take demand-discovery risk (so no insurgent-tier upside) and don't own the retail channel (so no harvest-tier durability either).

The premium multiples make sense only if brand IP by itself is still a durable asset. The whole argument of this deck is that brand IP, on its own, is exactly what stopped being durable. People Inc. is paying 2010s prices for assets that fit a 2010s value-capture model.

Caveats

What we don't know

A few honest gaps in the underlying data:

Absolute launch volume isn't published. Mintel reports a novelty mix (35% genuinely new in 2024 vs. 75% in 1996) but not totals. The "more launches than 2010" claim leans on Amazon Brand Registry, Shopify Plus growth, and Expo West first-time exhibitor counts — directionally consistent, no clean primary source.
Shelf-life data is proprietary. NielsenIQ and Circana track SKU velocity by category but don't publish historical comparisons. The compression argument here uses cohort-level acquisition data as a proxy.
The 2020+ cohort is small (n=7 acquisitions, n=49 surviving independents). The 2015–19 cohort is the load-bearing one. The 2020+ data is consistent with continued compression but not yet authoritative on its own.

Working draft

Nine charts, first complete pass

Six charts from the 383-brand database, three illustrative. Feedback welcome at matt@happyrobots.com.

The underlying brand database is browsable on the Brands page. Methodology in Criteria. The lists ingested in Sources.