B2B Network Effects: The Moat AI Can't Melt

This is one of RoadmapOne ’s articles on Strategy for Product Leaders .

B2B network effects occur when a business product becomes more valuable as more organisations connect to it. Each new participant makes the product stickier for every existing participant, harder for competitors to displace, and cheaper for the vendor to sell. In B2B, the network usually forms between trading partners — buyers and suppliers, brokers and traders, firms and their clients.

I spend a surprising amount of my board and advisory life explaining the same thing: if your product has a network effect — even a latent one — you should double down on it. Hard. Ahead of almost everything else on the roadmap.

And I keep having to explain it because most boards, and most product leaders, don’t get it. In my experience, the only people who instinctively understand network effects are those who have lived through them — either as the beneficiary or as the victim. Everyone else files “the network” alongside the rest of the backlog. Just another initiative competing for capacity. It is not just another initiative. It is the difference between a product that gets ripped out in three years and a piece of infrastructure that a whole industry cannot function without.

Sign up (free) to put network-building Objectives on your roadmap

What Are Network Effects?

A network effect exists when the value of a product increases with the number of participants using it. The telephone is the canonical example. One phone is useless. Two phones are mildly useful. A million phones are indispensable — and the million-and-first customer gets more value than the first ever did, while paying the same price.

The economics are seductive because they invert the normal rules. Most products face diminishing returns: each additional customer is harder to win and worth roughly the same. Networked products face increasing returns: each additional customer makes the product better, which makes the next customer easier to win. Get the flywheel spinning and growth compounds. That is why venture investors obsess over the category, and why networked businesses trade at multiples that make CFOs blink.

The Three Types of Network Effects

There are three types of network effects: direct, indirect, and two-sided. Direct network effects mean each new user adds value to all users on the same side — think messaging platforms. Indirect network effects mean growth on one side attracts a complementary side — more developers attract more users, and vice versa. Two-sided network effects — the dominant form in B2B — connect two distinct groups who transact with each other: buyers and suppliers, brokers and traders, firms and clients. Technically, two-sided effects are the strongest expression of indirect effects, but they behave so distinctively in B2B that they deserve their own label.

NFX famously maps sixteen sub-types . The taxonomy matters less than the diagnostic question: does each new participant make your product more valuable to the participants you already have? If yes, you have a network effect. If yes and you haven’t built your strategy around it, you have a problem.

Why B2B Undervalues Its Strongest Moat

Network effects have a branding problem in B2B. They sound like a consumer phenomenon — social networks, marketplaces, ride-sharing. Boards nod politely and go back to the feature roadmap.

This is exactly backwards. B2B network effects are stronger than consumer ones, for a simple reason: businesses embed connections into their operations. A consumer can join a new social network in an idle moment. A business that has integrated its trading partners, its order flow, and its back-office reconciliation into your platform faces months of migration work, operational risk, and retraining to leave. Consumer networks hold attention. B2B networks hold workflow. Workflow is far harder to move.

I watched this play out in travel. The big online travel agencies used their aggregation networks to force entire supplier categories into their ecosystem on their terms. The big rental-car brands saw it coming. They complained, they resisted, they held out — and they were powerless to prevent it, because the demand had consolidated onto the network and no individual supplier could afford to be off it. That is what being the victim of a network effect feels like: you can see the noose tightening and every individually rational decision tightens it further.

Yet when I sit in board meetings and a latent network opportunity comes up, it gets evaluated like any other feature. Same prioritisation scoring , same capacity horse-trading, same quarter-by-quarter scrutiny. The strategic asymmetry — that this one initiative could change the slope of the business rather than its intercept — never makes it into the framework. If your strategy conversation needs a scaffold for this, 7 Powers is the right one: Network Economies is one of Helmer’s seven durable moats, and he would tell you that almost nothing else on your feature roadmap qualifies.

The Benefits, in Very Simple Terms

Strip away the theory and network effects deliver four compounding commercial advantages.

Stickiness

Churn is the tax every SaaS business pays. Networked products barely pay it. When your product connects a customer to fifty trading partners, leaving means re-establishing fifty connections on a rival platform — and persuading fifty other organisations, whom the customer does not control, to move with them. Nobody churns from their own supply chain. The renewal conversation changes from “justify your value again” to “here’s the invoice.”

Win-Rate

This is the benefit that sales teams understand instantly and boards underestimate perpetually. In a networked market, your competitor’s product can be better, cheaper, and shinier — and still lose. The line that wins the deal is brutally simple: “You can buy our competitor’s software, and it won’t connect to any of your trading partners.” Feature comparisons become irrelevant. The prospect isn’t buying software; they’re buying access to everyone already on the network. Win-rates in networked categories are not incrementally better. They are lopsided.

Pricing Power — and Monetising the Edges

Most software vendors monetise nodes: a licence per customer, a fee per seat. A network operator monetises edges — the connections between participants. That distinction sounds academic. It is worth a fortune. A network with 100 participants has 100 nodes, but potentially thousands of edges, and every edge is a billable relationship that deepens over time. Better still, edge revenue grows quadratically as the network grows linearly. Add one participant and you don’t add one revenue line — you add a connection to everyone already there.

Expansion Rights

An established network is a licence to expand. Adjacent products, adjacent workflows, adjacent markets — all become cheaper to enter because the hardest asset, the connected customer base, is already in place. Competitors entering the same adjacency must build both the product and the network. You only need the product. I’ll show you what that looked like in practice.

Trayport: The Network That Kept the Lights On

I spent years at Trayport , the platform that sits underneath European wholesale energy trading. The structure was a two-sided network in its purest form: brokers providing over-the-counter markets on one side, trading firms on the other. Every broker paid for every trader connection. Every trader paid for every broker connection. We monetised the edges, not just the nodes — and everybody’s trading volumes were locked together in one liquidity pool.

The scale still makes people sit up. For every ten seconds a light bulb burned in Europe, roughly nine and a half of those seconds had been traded across our network. One of our consultants put it more bluntly: this is critical infrastructure — if it disappeared tomorrow, the lights in Europe probably do go out for a few days.

Two lessons from that period are the heart of this article.

First: the network defended itself — against genuinely credible attackers. We faced a constant stream of well-funded alternative trading venues. Exchanges, consortia, broker breakaways. On paper, several should have won. In practice, the traders did not want another screen on an already crowded desk, and their back offices emphatically did not want another set of connections to reconcile. One by one, the would-be competitors ended up paying us to connect to the traders through our network. Read that again. The businesses that set out to displace the network concluded that their least-bad option was to become participants in it — arriving as conquerors, staying as customers. That is what a real moat looks like. Not a feature advantage. A structural position where even your competitors’ success requires your rails.

Second: the network was a springboard, not just a fortress. We started in power and gas. Because the traders, brokers and clearing connections were already in place, each adjacent market cost us a fraction of what it cost anyone else to enter. Coal came next. Then the emissions market as EU carbon trading exploded in the late 2000s — we captured it almost as a by-product. Then dry freight. Then iron ore. Then oil cracks. And the expansion wasn’t only horizontal: we moved along the workflow too, from the front-office trading screen into middle- and back-office — confirmations, risk, reconciliation — because once you carry the trade, every downstream process wants to consume it from the source. The network made Trayport the entrepôt of European commodities trading: everything flowed through, and every flow paid.

Nobody plans “capture six asset classes and three office functions” on day one. What you plan is the network. The expansion rights come with it.

CPA Global: The Network Hiding in Your Customer Data

Here’s the version of this story for anyone who thinks “we’re not a trading platform, this doesn’t apply to us.”

At CPA Global, the IP-services business, we served two customer groups that transact with each other constantly: corporates managing patent portfolios, and the law firms that do their filing work. Two platforms, two customer bases, no network between them. Sitting in the data was a latent network of enormous value.

We could show a large corporate client the hundred law firms they used around the world — and then show them that 68% of those firms were already on our law-firm platform. The connection was sitting there, waiting to be lit up. Connecting the two sides — instructions, status, documents, billing flowing firm-to-client through one pipe — was, on our analysis, a £200m-a-year opportunity. And here is the commercial kicker: we sold the biggest software deal in CPA’s history on the strength of that proposal. Not on features. On the promise of the network.

The generalisable lesson: you may not need to build a network from a standing start — you may only need to reveal one. Any business serving two sides of an existing commercial relationship is sitting on latent edges. Law firms and clients. General contractors and subcontractors. Insurers and brokers. Manufacturers and distributors. If both sides already use you separately, the cold-start problem — the thing that kills most network plays — has largely been solved without you noticing.

The Leverage Cascade: Each New Side Comes Cheaper

One more pattern, heavily anonymised. A construction software vendor I know connects general contractors with their subcontractors — bids, documents, schedules flowing across the network. A solid two-sided position. The interesting move is the third side: manufacturers.

Why would manufacturers join? Because the contractors on the network demand product data — specifications, certifications, pricing — and a manufacturer who won’t provide it through the platform is, in effect, inviting the contractor to specify a rival’s product instead. The network’s existing participants do the selling. The vendor barely has to lift a phone. Each new side of a network is cheaper to recruit than the last, because the gravitational pull of the existing sides does the work. That is the leverage cascade, and it is how two-sided networks become three- and four-sided ones.

See your capacity allocation against strategic Objectives — sign up free

The Cold-Start Problem — and How to Cheat It

Time for the honest part. Everything above describes networks after critical mass. Getting there is brutally hard, and most network plays die in the cold-start phase: too few participants on either side to be useful to the other.

Three rules from having done it and watched it done:

Start where you already have density. Don’t build a network for a market; build it for a corner of a market where you already serve a meaningful share of one side. CPA had 68% of the relevant law firms before anyone drew a network diagram. If you’re starting from zero on both sides, you are not building a network effect — you are funding two customer-acquisition problems simultaneously, and your business case should say so out loud. This is the same logic as Crossing the Chasm’s bowling-alley: dominate a niche completely rather than sprinkle presence everywhere. Density in one segment beats coverage of ten.

Give away connectivity early. Crank the pricing handle later. In the cold-start phase, every connection you charge for is a connection you won’t get. Free connectivity is not generosity; it is deferred pricing power. Once the network reaches critical mass — once being off it costs more than being on it — the pricing conversation transforms. Participants who joined for free will pay handsomely to stay connected to a network they now depend on, and new entrants will pay full freight for access to it. The sequencing discipline matters: monetise too early and you strangle the flywheel; wait for lock-in and the margin is yours for a decade.

Measure the network, not the pipeline. Y Combinator used to screen for virality in consumer startups because viral growth compounds. Network effects are the B2B equivalent of virality — each connected participant recruits the next. So instrument it like a growth loop: count nodes, count active edges, count the share of new participants who arrived because someone on the network pulled them in. When that last number climbs, you have escape velocity.

Guard the Gates: The API Strategy That Preserves the Moat

Now for the part that will annoy your engineers. A network effect is only a moat if the liquidity — the transactions, the data, the relationships — stays on your network. And the single most common way B2B companies bleed their moat is through their own APIs.

Developers instinctively favour openness. Big API surfaces feel elegant, modern, right-side-of-history. “Openness” has become a virtue word, and platforms that restrict access get sneered at. But strategically, an unguarded API is a siphon: it lets a competitor drain out your data, replicate your connections, and quietly move the liquidity to their venue while your engineering team congratulates itself on its developer experience. Openness is a pricing decision, not a philosophy — and it deserves the same scrutiny as any other pricing decision.

If competitors or third parties must be allowed in — and commercially or (in some industries) legally, they sometimes must — then let them in on terms that protect the asset:

• Charge for conformance testing. Every third party that connects should pay for certification, and re-certification on every release. This is standard practice in payments and exchanges for good reason: it monetises access and controls quality in one move.
• Price third-party API access above your own products. If a rival’s route to your network costs more than your native product, you have converted a competitive threat into a revenue line and a pricing umbrella. Access is a product. Price it like your most valuable one, because it is.
• Keep third-party surfaces narrow; keep your own integration deep. The third-party API should do what it contractually must — no more. Your own products should enjoy rich, full-fidelity integration the API doesn’t expose. The gap between the two is the product advantage, renewed with every release.

None of this is anti-customer. Your customers get the deepest integration you can build. It is anti-parasite — aimed squarely at the intermediary whose business model is harvesting your network’s liquidity without having paid the decade of investment that created it. The naive-openness failure mode has a long history: platforms that opened generous APIs, watched aggregators commoditise them into anonymous back-ends, and spent the next decade clawing back terms. Guard the gates from day one. It is far harder to narrow an API than never to have opened it.

The Moat AI Can’t Melt

Here is why this article matters more in 2026 than it would have five years ago.

AI has collapsed the cost of building software. A competent team — or increasingly, a barely competent one — can clone your feature set in weeks. Feature differentiation, execution speed, even UX polish: every conventional moat is melting , because everything made of code can now be reproduced at near-zero cost. I’ve written elsewhere that five of Helmer’s seven Powers strengthen in the AI era. Network Economies strengthen most of all — for a blunt reason.

A network is the one asset a competitor cannot prompt into existence. They can generate your software. They cannot generate your two hundred connected trading partners, the integrations wired into every participant’s back office, or the decade of accumulated trust that put them there. Convincing organisations to join — and to change their operational workflows — costs what it has always cost: time, salespeople, and credibility. AI has collapsed build cost; it has not collapsed sell cost. A moat made of other people’s commitments does not melt, because it was never made of code.

There’s a second-order effect too. Networks generate proprietary flow data — who transacts with whom, at what volume, in which direction. In an AI era, that data trains models nobody else can train: pricing intelligence, risk signals, matching algorithms. The network effect and the data moat compound each other. The rich get richer, structurally.

So when the board asks “what’s our AI defensibility strategy?” — and every board now asks — the honest answer for most B2B companies is not a model or a copilot. It is: own the network in your corner of the industry before someone else does. The software is becoming a commodity. The connections are not.

When Network Effects Fail You

Like every framework in this cluster , this one is a tool in the kit bag, not a universal law. Intellectual honesty demands the failure modes:

Not every B2B business has a latent network. If your customers don’t transact with each other — or with a counterparty group you could plausibly serve — you cannot manufacture a network effect, and claiming one in the board deck is self-deception. Single-player value first; check for product-market fit before dreaming of flywheels.

Multihoming kills weak networks. If participants can cheaply sit on your network and your rival’s simultaneously, the winner-take-most dynamics never engage. The Trayport lock held because a second trading screen and a second set of back-office connections were genuinely painful. If joining a second network costs your participants nothing, expect to share the market.

Edges can be shallow. A “network” where participants merely co-exist — a directory, a badge programme — is not a network effect; nothing of value flows across the connections. Judge your network by the volume and criticality of what moves along its edges, not by the number of logos on the slide.

Dominance attracts scrutiny. Run the playbook well and you will eventually meet regulators, as every dominant network operator does. Plan for it; consider it a graduation ceremony.

Putting Network Effects on the Roadmap

If you take one operational point from this article, take this one. The reason network opportunities die is not that boards reject them. It’s that they get framed as build projects — and build projects get sliced, deferred, and de-scoped until nothing remains.

In Run-Grow-Transform terms, network-building is usually Transform work: it changes the shape of the business rather than feeding this quarter’s numbers, which is precisely why it loses every capacity argument it’s ever entered. There is always more work than capacity, and this is the work that blinks first. The fix is structural, and it starts with how the Objective is written.

Frame it as an outcome, not an output . Not “build connectivity between contractors and subcontractors” — that’s a feature spec wearing a strategy costume, and it will be judged, fatally, on delivery cost. Write instead: “Get five clients of type X transacting with five clients of type Y.” Now the Objective is a network outcome. The Key Results fall out naturally — connections live, edges active, volume flowing — and the team is empowered to discover the cheapest path to liquidity rather than marching through an integration Gantt chart.

Then resource it honestly. A network play run off the side of someone’s desk is a network play that will fail slowly and expensively; give it a dedicated minimum viable team — two engineers and a product person, full-time and protected. In RoadmapOne , that means the Objective is tagged — Transform, or Network Economies if you tag by Power — and visibly allocated to a squad in the capacity grid , sprint after sprint. When the quarterly squeeze arrives, the analytics show the board exactly what they’d be cutting: not “a project”, but the moat. That visibility is what keeps strategic work funded — because a board that can see the network Objective in the allocation is a board you no longer have to convince twice.

Frequently Asked Questions

What are the three types of network effects?

The three types are direct, indirect, and two-sided. Direct network effects mean every new user adds value for all users on the same side, as in messaging platforms. Indirect effects mean growth on one side attracts a complementary side, as with developers and app users. Two-sided effects — dominant in B2B — connect two groups who transact with each other, such as buyers and suppliers or brokers and traders.

What are examples of network effects in B2B?

Energy and commodities trading platforms connecting brokers with traders, payment networks connecting merchants with card issuers, freight platforms connecting shippers with carriers, procurement networks connecting buyers with suppliers, and construction platforms connecting contractors with subcontractors and manufacturers. In each case every new participant makes the network more valuable to existing participants — and leaving means abandoning connections to trading partners the customer doesn’t control.

Are network effects a moat?

Yes — in B2B software they are arguably the strongest moat available. Hamilton Helmer lists Network Economies among his 7 Powers precisely because the advantage persists even when competitors copy the product. A rival can replicate your features but not your connected participants, whose workflows are embedded in the network. In the AI era this matters more: build cost has collapsed, but the cost of convincing organisations to join a network has not.

How do you build network effects in a B2B business?

Start from density: pick a corner of the market where you already serve a large share of one side, rather than starting cold on both. Give away connectivity early to accelerate critical mass, and defer monetisation until being off the network costs more than being on it. Measure nodes, active edges, and network-referred growth. Once liquidity is established, expand into adjacent markets and workflows, and control API access so the liquidity stays on your network.

What is the difference between network effects and virality?

Virality describes how fast users spread a product; network effects describe how much more valuable the product becomes as they do. A viral product can grow quickly and still retain nobody, because early users get no extra value from later ones. Network effects are the B2B equivalent of virality: each connected participant makes joining more attractive for the next, but the value — and the retention — compounds permanently rather than just the awareness.

Conclusion

Network effects are not another feature to prioritise. They are a different kind of asset — one that delivers stickiness, lopsided win-rates, edge-based pricing power, and cheap expansion rights simultaneously, and the only one that gets stronger as AI makes everything else cheaper to copy. I have sat on both sides: running a network so entrenched that competitors paid to join it, and advising industries where incumbents watched a networked player become inevitable. Nobody I’ve met who has lived either experience ever undervalues network effects again.

You don’t need to be a trading platform. You need two groups of customers who transact with each other, the discipline to reveal and light up the connections between them, and the nerve to guard the gates once the liquidity flows. Write the network into your roadmap as an outcome, give it a protected team, make the allocation visible, and let the flywheel do what flywheels do.

The software is melting. The network won’t.

Sign up (free) to build the roadmap your moat deserves