Six alternative branches. These are subjective probabilities from the seed, not observed frequencies or investment recommendations.
Why this trend
Load-bearing assumption
That the channel shift proceeds as a grind of roughly 0.8 percentage points of share a year rather than as a step change, so that the decisive competitive variable is which adjacent profit pool funds price investment, not how fast online grows.
Preconditions
Not recorded.
E-commerce grinds to roughly 22-24% of US retail by 2030 at about 0.8 percentage points a year from the 17.1% recorded for Q2 2026. Retail media consolidates into an Amazon-Walmart duopoly with a scaled second tier of three to five networks and a dead long tail. Agentic discovery becomes a material referral channel while agentic checkout stays a minority of it. Trade-down persists, stores stabilise as fulfilment and media infrastructure, and 2027 margins step down as the tariff-refund windfall lapses.
Mechanism
Online share gains come from category mix and same-day fulfilment rather than from a behavioural break; merchandise gross margin keeps eroding, so advertising, membership and marketplace commissions carry operating income. The decelerating element is retail media in aggregate: WARC's 9.8% ex-Amazon growth for 2027 is the lowest it has recorded, so the second tier's contribution to the profit pool shrinks even as the leaders compound at 26% and 38%.
Preconditions, early indicators and assumptions
- Early Indicators
- Indicator
Census e-commerce share progressing 0.7-0.9pp per year through 2027-28 without acceleration
- Source
S-12-01 Census Quarterly Retail E-Commerce Sales
- Would Be Visible By
2028-02
- Indicator
A year-over-year decline in gross margin rate at Target or Walmart in the quarter lapping the 2026 tariff refunds
- Source
S-12-09 Target IR; S-12-08 Walmart IR
- Would Be Visible By
2027-09
- Indicator
Instacart advertising growth staying below 20% for four consecutive quarters while Amazon advertising stays above 20%
- Source
S-12-09 Instacart IR; S-12-07 Amazon IR
- Would Be Visible By
2027-08
- Affected Industries
- 12
- 16
- 09
- 19
- What Businesses Should Do
Ask every retailer reporting 2026 margin expansion what the ex-refund number was, and budget 2027 off that figure; treat a sub-scale media network as trade-spend reclassification and stop funding it as a media business.
- Precedence Note
Distinguished from disruption by who mediates the transaction: in base, agents refer and humans check out. The first published agent-completed transaction count moves the outcome to disruption regardless of its level.
- Would Change Our Mind
Two consecutive quarters of e-commerce share gains above 0.30 percentage points quarter on quarter.
Tariff normalisation and energy mean-reversion restore real income, volume returns to grocery after a year of negative volume carried by price, and AI-mediated discovery expands consideration sets rather than redistributing existing demand. Retail media re-accelerates on off-site and connected-TV inventory as on-site saturates. Margins absorb the refund lapse without a visible step down because underlying merchandise margin recovers.
Mechanism
The named constraint that releases is real income. US grocery volume fell 1.0% in 2025 while price rose 2.2%, and wage growth at roughly 3.5% against 3.4% CPI is only marginally positive. If energy CPI falls from 16.3% toward its long-run range, discretionary budgets recover and volume returns; conversion lifts across the board because AI referral traffic already converts 54% better than other channels, so a larger denominator compounds rather than cannibalises.
Preconditions, early indicators and assumptions
- Early Indicators
- Indicator
US grocery volume turning positive year over year
- Source
S-12-19 McKinsey State of Grocery; S-12-02 Census MARTS food and beverage
- Would Be Visible By
2027-06
- Indicator
WARC revising ex-Amazon retail media growth back above 12%
- Source
S-12-14 WARC Future of Commerce Media
- Would Be Visible By
2027-12
- Indicator
Energy CPI twelve-month change falling below 5%
- Source
S-12-03 BLS CPI detailed tables
- Would Be Visible By
2027-09
- Affected Industries
- 12
- 19
- 16
- 05
- What Businesses Should Do
Hold promotional intensity rather than cutting it at the first sign of volume recovery, because trade-down quality perception has shifted and does not revert with income; invest feed and structured-data quality now, since AI-referred conversion is the one measured lift available.
- Precedence Note
Requires volume, not price, to be the source of growth. Nominal sales growth driven by inflation stays in base.
- Would Change Our Mind
Two consecutive quarters of negative grocery volume with private-label share still rising.
Rates rise, leveraged specialty retail fails in volume, energy stays elevated, and the tariff-refund windfall unwinds into a 2027 margin cliff that coincides with an advertising slowdown. Retailers that guided 2027 off flattered 2026 margins miss, and the miss lands in the same year as refinancing.
Mechanism
The parameter that deteriorates is the cost of capital, transmitted through balance sheets rather than through format failure: Saks filed with $3.4bn of debt and Sleep Number with $1.3bn, and the FOMC held at 3.50-3.75% with three members dissenting toward a hike. Advertising revenue is near-100% incremental margin, so an ad-market slowdown hits retail operating income far harder than it hits retail sales. The refund lapse shows up first in the quarter lapping Target's $994mn, worth 3.7 percentage points of margin.
Preconditions, early indicators and assumptions
- Early Indicators
- Indicator
Major US retail Chapter 11 filings exceeding 12 in a calendar year
- Source
S-12-22 Retail Dive bankruptcy tracker
- Would Be Visible By
2028-01
- Indicator
Any large retailer guiding down on advertising revenue while comparable sales also decelerate
- Source
S-12-07, S-12-08, S-12-09 retailer IR
- Would Be Visible By
2027-09
- Indicator
Energy CPI twelve-month change still above 8% in mid-2027
- Source
S-12-03 BLS CPI detailed tables
- Would Be Visible By
2027-07
- Affected Industries
- 12
- 16
- 18
- 09
- What Businesses Should Do
Refinance ahead of the 2027 maturities rather than into them; separate the refund benefit from underlying margin in internal reporting now, so the 2027 comparison is understood before the market makes it.
- Precedence Note
Downside requires both legs: a margin step-down and a credit or advertising deterioration. A margin step-down alone is the arithmetic consequence of the refund lapsing and stays in base.
- Would Change Our Mind
A 2027 in which retail bankruptcies stay below ten and advertising revenue at the top five networks grows above 15%.
Agent-mediated purchasing crosses roughly 5% of online transactions and retailers are disintermediated from discovery, losing both the advertising inventory and the merchandising lever in one move. Price becomes the main variable an agent can compare, and the alternative-profit model that funds retail price investment loses its inventory.
Mechanism
Two independent protocol stacks already exist — OpenAI and Stripe's Agentic Commerce Protocol with the Shared Payment Token, and Google's Agent Payments Protocol with signed Intent and Cart Mandates and more than sixty partners — which satisfies the framework's two-parties test. If agents complete checkout, the retailer's page stops being the decision surface, on-site advertising inventory loses its audience, and the retailer's product feed rather than its merchandising becomes the whole interface.
Preconditions, early indicators and assumptions
- Early Indicators
- Indicator
The first published transaction count, GMV figure or active-merchant number from OpenAI, Google or Stripe
- Source
S-12-20 agentic-commerce protocol documentation
- Would Be Visible By
2027-12
- Indicator
Adobe publishing an AI-referral denominator rather than only growth and conversion rates
- Source
S-12-16 Adobe Digital Insights
- Would Be Visible By
2027-06
- Indicator
A top-ten US retailer quantifying agent-referred or agent-completed GMV in an earnings release
- Source
S-12-23 SEC EDGAR full-text search on 'agentic'
- Would Be Visible By
2027-12
- Affected Industries
- 12
- 16
- 07
- 01
- What Businesses Should Do
Treat product feed and structured-data quality as a distribution asset rather than an SEO task — Adobe scores retail product pages at 66/100 against 75 for homepages — and negotiate agent-channel terms before volume exists, not after.
- Precedence Note
Requires a published transaction or merchant number. Growth rates over an undisclosed base, partner counts and integration announcements leave the outcome in base. That single disclosure is the branch point.
- Would Change Our Mind
A second full year of complete silence from all eight named protocol participants, combined with Adobe's AI-referral growth rate decaying below 50% year over year.
Marketplace unbundling under FTC v. Amazon or the EU Digital Markets Act separates marketplace, logistics and advertising, collapsing the alternative-profit model that currently funds retail price investment. Simultaneously de minimis is restored by executive action and the low-price import channel reopens, so the price floor falls at exactly the moment the profit pool that subsidised price is broken up. Direction is two-sided: the same instruments could equally harden the current structure.
Mechanism
The instruments are named and dated: FTC v. Amazon in its remedy phase, the EU DMA, the US de minimis suspension upheld by the Court of International Trade on 2026-08-13 on a privilege-revocation theory with statutory repeal due July 2027, and the EU's EUR 3 per-item duty in force from 2026-07-01 to 2028. The four-stage distinction matters: the de minimis suspension is in force and enforced; the FTC remedy is not yet even proposed in structural form.
Preconditions, early indicators and assumptions
- Early Indicators
- Indicator
Any remedy phase in FTC v. Amazon reaching structural relief rather than conduct remedies
- Source
S-12-23 SEC EDGAR and federal court filings
- Would Be Visible By
2027-12
- Indicator
A Federal Register notice restoring Section 321 de minimis treatment
- Source
S-12-10 CBP trade releases
- Would Be Visible By
2027-06
- Indicator
An EU DMA non-compliance decision naming marketplace self-preferencing in commerce search
- Source
S-12-06 EU customs and Commission releases
- Would Be Visible By
2027-12
- Affected Industries
- 12
- 16
- 09
- 07
- What Businesses Should Do
Model the marketplace, logistics and advertising businesses on a standalone basis now so that a separation order does not arrive as a surprise; do not build a 2027 sourcing plan that assumes either the de minimis closure or its restoration.
- Precedence Note
Regulatory outranks base when an order issues, even if measured shares look base-like. A complaint, an investigation or a proposed remedy is not an order.
- Would Change Our Mind
FTC v. Amazon settling on conduct remedies alone with no structural component, combined with the July 2027 statutory de minimis repeal taking effect unaltered.
The agentic-commerce infrastructure build is written off the way cashierless stores and Western social commerce were. Retail media growth goes negative excluding Amazon, retailers that capitalised network build costs take impairments, and retail reverts to a price-and-convenience business with no adjacent profit pool to fund price investment.
Mechanism
The thesis that fails is that first-party purchase data plus on-site inventory equals a high-margin advertising business for everyone. WARC already puts Amazon at 78% of US retail media and Walmart at 7.5%, leaving 14.5% for every other network combined, and an Ipsos study found memory encoding for ads on retail platforms 47% lower than for the same ads off-site. The holders of the impaired asset are the retailers that capitalised network and agentic-integration build costs; the historical analogue is Western social commerce 2019-2023, repeatedly declared imminent on the strength of Chinese adoption and never transferring, which resolved as a quiet de-scoping rather than a crash.
Preconditions, early indicators and assumptions
- Early Indicators
- Indicator
Two consecutive quarters of declining retail media revenue at any top-five non-Amazon network
- Source
S-12-09 retailer IR; S-12-14 WARC
- Would Be Visible By
2028-02
- Indicator
Public deprecation of the Agentic Commerce Protocol or the Agent Payments Protocol by its authors
- Source
S-12-20 protocol documentation and changelogs
- Would Be Visible By
2028-06
- Indicator
An impairment charge against capitalised retail media or commerce-integration assets in any top-ten retailer's filings
- Source
S-12-23 SEC EDGAR full-text search
- Would Be Visible By
2028-03
- Affected Industries
- 12
- 16
- 01
- What Businesses Should Do
Do not capitalise agentic-integration spend on an assumed multi-year benefit; hold network investment to a payback test that works at current, disclosed, on-site volumes only.
- Precedence Note
Failure outranks downside on occurrence. Downside is a cyclical advertising slowdown; failure is the adjacent profit pool being written down as a category.
- Would Change Our Mind
Any top-five non-Amazon network sustaining advertising revenue growth above 20% for four consecutive quarters.
Additional scenario notes
- Probabilities Sum
1
- What Must Be True To Grow
- Census continues to publish the quarterly e-commerce series on its fixed schedule, since it is the only Tier A measure of the channel shift
- Total retail sales keep growing, because share is a ratio and a weak denominator flatters it without any behavioural change
- Same-day fulfilment cost per order falls faster than energy and labour costs rise, with energy CPI at 16.3% against core goods at 0.7%
- An adjacent profit pool — advertising, membership, marketplace or fulfilment services — keeps funding price investment as merchandise margin erodes
- The 2027 margin comparison is understood as a refund lapse rather than an operating deterioration, or capital is withdrawn from the wrong businesses
- What Could Stop It
- Regulatory reversal: restoration of de minimis, or structural relief in FTC v. Amazon that unbundles the profit model
- Capital cost: leverage, not format failure, is what has been killing retailers, and the policy rate sits at 3.50-3.75% with hawkish dissent
- Demand deflation: the agentic-commerce pipeline is infrastructure without a disclosed transaction anywhere
- Measurement revision: the US 17.1% and UK 28.3% online shares are not comparable, and most published headroom estimates rest on that non-comparability
- Physical constraint: energy CPI at 16.3% hitting refrigeration and last-mile fuel precisely as same-day delivery scales
- Uncertain Assumptions
- Text
Census e-commerce share continues to be measured on a stable definition, with buy-online-pickup-in-store treated consistently
- Confidence
medium
- Load Bearing
false
- If Wrong
A definitional change moves reported share by more than a year of organic gain and every threshold in this set becomes unresolvable across the boundary
- Text
The tariff-refund benefit was concentrated in 2026 quarters rather than spread across several periods
- Confidence
medium
- Load Bearing
false
- If Wrong
The 2027 margin cliff flattens into a slope and the downside indicator fires late or not at all
- Text
The channel shift proceeds as a grind rather than a step change, so mechanism competition rather than growth rate is the decisive variable
- Confidence
high
- Load Bearing
true
- If Wrong
Base and disruption swap probability mass, and the 2020-21 error of extrapolating a shock as a slope is repeated in the opposite direction
- Text
Agent-completed checkout remains unmeasured because volume is low rather than because disclosure is withheld for competitive reasons
- Confidence
low
- Load Bearing
false
- If Wrong
Disruption is already under way and is invisible, and the base case is describing a world that has already changed
- Authored
2026-09-15