Xumulus Logo
Search

The Retail Website Monopoly Is Over. That's Not the Same as the Website Dying.

Date Published

Forrester published a report this year with a title designed to get attention: Death Of The Retail Website As We Know It. The accompanying blog post is more measured than the title, and the gap between the two is where the actual insight lives.

Forrester's own answer to whether websites will really die is, roughly, probably not. What's ending is something narrower and more specific: for about three decades, the retail website owned non-marketplace digital commerce outright. That monopoly is what's breaking. The site isn't disappearing — it's losing its automatic position at the center of every buying journey.

For catalog-heavy distributors, that distinction is the whole ballgame. A decline narrative tells you to panic. A repositioning narrative tells you what to build.

Where shopping starts is now a function of intent

The most useful thing in Forrester's analysis isn't a growth number. It's a segmentation.

Shoppers don't pick a channel out of loyalty. They pick based on what they're trying to accomplish in that moment. Research-heavy purchases and unfamiliar products pull people toward answer engines that can compare and explain. Purchases where the buyer already knows the brand — or is spending real money — tend to start on the retailer's or brand's own site. Cheap, repeat, commoditized buys still default to marketplaces where price and convenience win.

Three intents, three different starting points. The website didn't lose all of them. It lost one of them, and it happens to be the one that used to feed the funnel.

Forrester's February 2026 Consumer Pulse Survey found that more than a quarter of US online adults had used ChatGPT to search for products within the prior month. That's the discovery layer relocating.

Translate that to B2B and it gets sharper

Run the same three-intent framework against how distribution actually works:

Known part number, existing account. The buyer isn't shopping. They're transacting. This stays on your site or in your punchout, and agents mostly help by making it faster. Low risk.

Commoditized consumables. This is the marketplace bucket, and in B2B it has been leaking to Amazon Business and Grainger for years. Agents accelerate an erosion that already started.

Spec-driven and unfamiliar. "I need a fitting that handles 3000 PSI at this thread pitch, compatible with the unit I bought from you in 2019." This is high-consideration research, and it is precisely the bucket Forrester identifies as flowing toward answer engines.

That third bucket is where distributor margin lives. It's the work that justifies the relationship — the fitment expertise, the cross-reference, the "actually, you want this one instead." If that conversation now happens inside an assistant that has never read your catalog, you don't lose a session. You lose the consultative position that made you more than a price.

That's the real exposure, and it's larger in B2B than in consumer retail. Gartner projects that by 2028 roughly 90% of B2B purchases will be intermediated by AI agents. B2B buying was never a browsing experience — it was always a lookup problem wearing a storefront costume. Which makes it a better fit for agents than consumer retail, not a worse one.

The honest counterweight

Forrester addresses the obvious objection directly rather than dodging it: it's tempting to write off answer-engine commerce as immature, and current purchase volume genuinely is limited. They also note that some answer engines have retreated from their early native checkout efforts — a detail that gets left out of most agentic-commerce coverage.

Add the supporting data and the picture stays honest. AI-referred traffic is still well under 1% of total retail traffic. Adobe reports it growing 393% year over year in Q1 2026 with conversion running roughly 42% above traditional search — impressive rates off a very small base. Forecasts for 2030 range from Morgan Stanley's roughly $190 billion to McKinsey's $3–5 trillion, a spread of more than 200x that exists mainly because the firms are counting different things.

Forrester's framing of this is the right one: what matters isn't today's volume, it's the trajectory and the fact that interest substantially exceeds current usage. Treat present numbers as a floor rather than a ceiling.

There's also a second-order effect worth more than the traffic. Conversational shopping teaches people to use natural language, and that expectation doesn't stay in ChatGPT. It follows them onto your site. Even if not a single order ever arrives through an agent, your own search box is now being judged against an interface that understands "something that fits my 2019 unit."

That alone is a reason to fix your index.

What appreciates and what depreciates

If the site is no longer the default venue for the high-consideration decision, your assets split into two piles.

Depreciating: navigation architecture, category merchandising, promotional real estate, funnel choreography. An agent doesn't see your hero image and won't be moved by your brand story.

Appreciating: structured product data, attribute completeness and normalization, real-time availability, contract and tier pricing logic, fitment and compatibility relationships, and the ability to answer a machine query authoritatively in milliseconds.

Forrester lands in the same place — clean, comprehensive product data is now table stakes across every channel, not a nice-to-have on the one you own.

The catalog data underneath the storefront is becoming the product. The storefront is becoming one rendering of it.

This is the architectural fork. If your product data exists as a byproduct of your storefront — fields in a platform's schema, enrichment done by whoever built the PDP template — then every new surface is a fresh integration project. If it lives in an index the storefront merely reads from, then the site, the agent endpoint, the punchout catalog, the marketplace feed and the ERP integration are all just consumers of one source of truth.

One is a rebuild every time the channel mix shifts. The other is a configuration change.

What to actually do

Not a replatform. Four things, roughly in order:

  1. Audit your machine visibility. Query your top 20 SKUs and your ten most common spec-driven buying questions in ChatGPT, Gemini, Perplexity and Claude. Note where you appear, where a competitor appears instead, and where the answer is simply wrong. Costs an afternoon; tells you more than a deck.
  2. Fix attribute completeness first. Agents filter deterministically. A missing attribute isn't a soft ranking penalty, it's exclusion from the result set. Incomplete data is invisible data.
  3. Decide assortment by channel deliberately. Forrester's point here is underrated: not every product should be pushed into every non-owned environment. Choose what belongs where instead of syndicating everything by default and letting the channel decide your margin mix for you.
  4. Separate the index from the storefront. If a new channel requires a new data pipeline, the architecture is wrong. One index, many surfaces.

Forrester's closing note is worth borrowing: leaders who accept the role change early spend less energy defending the previous model and more designing the next one. The distributors who struggle over the next three years won't be the ones whose websites declined. They'll be the ones whose product data was only ever good enough to render a page.

References

Primary source

Supporting data

Context and counterpoint

Note on the figures: Adobe and Shopify numbers are platform telemetry; Salesforce figures are vendor-reported program data; Gartner, McKinsey and Morgan Stanley figures are analyst forecasts measuring different scopes and should not be read as competing estimates of the same quantity.