B2B eCommerce Strategy: How to Build and Execute One

Fig. Growth-led and cost-led B2B ecommerce strategies compared.

Whichever intention dominates, the B2B ecommerce model underneath it changes every answer that follows, so it is worth naming explicitly.

• Under direct sales you own both the customer relationship and the order.
• Under distributor-led sales someone else owns the relationship, and your portal exists to support that partner rather than to go around them.
• A marketplace model brings third-party sellers onto your infrastructure, while punchout links your catalog into the buyer's own procurement system, such as Ariba, Coupa or Jaggaer, so requisitions are raised against your live catalog and prices and flow back into the buyer's approval workflow without anyone retyping an order.

Demand generation decides who arrives; a commerce strategy decides what happens once they do, and companies that conflate the two end up with a well-promoted portal nobody can complete an order through.

💡 For the ground rules underneath all of it, see what B2B ecommerce is and how it works.

Most B2B ecommerce strategies that fail were never rejected by buyers at all; they were undermined by internal inconsistencies that took months to become visible. McKinsey's 2026 survey found that inconsistent information across teams has become the single biggest reason B2B buyers switch suppliers, ahead of not being able to reach someone knowledgeable and ahead of gaps in cross-channel order tracking. Those read as customer-experience failures and behave as operational ones, and each traces back to a decision that was made carelessly or deferred. Six of them carry most of the weight, and because each constrains the next, the order is part of the answer.

The instinct to move every account at once is where a lot of programs lose their own sales team, since the accounts most likely to resist tend to be the ones a rep has spent a decade cultivating.

The archetype split in the same McKinsey Surprising Economics of B2B Growth report (2026) gives that resistance a shape worth planning around:

• 53% of B2B buyers are relationship-oriented adapters who default to familiar suppliers and established processes, up from 44% in 2024;
• 29% are seekers who expect fully integrated channels;
• 18% are innovators pushing for advanced self-service.

Push adapters onto a portal with no rep coverage behind it and adoption flattens early, at which point the sales team concludes, reasonably enough, that the portal is optional.

Contract accounts rarely buy from the full catalog, purchasing instead from an agreed subset that may carry approved substitutes and may exclude items they are contractually barred from seeing.

Where visibility gets handled as a filter layered over one master catalog, the rules accumulate until nobody can explain why a given account sees a given product, and the explanation ends up distributed across three systems. The rules outlive whoever wrote them.

The problem announces itself through the support queue, in customers asking why something they have bought for years has disappeared and nobody being able to answer without opening the ERP.

Getting the catalog structure right removes most of the visibility problem before it starts.

Pricing is where B2B and consumer retail stop sharing assumptions. Contract rates, volume breaks, customer-specific units of measure, after-the-fact rebates and currency variation all exist somewhere already: usually in the ERP, occasionally in a spreadsheet a regional manager keeps privately.

What has to be decided is whether the storefront calculates price or displays a price the ERP has calculated, and choosing calculation means owning a second pricing engine that must stay identical to the first one indefinitely. The failure surfaces as a customer disputing an invoice against a portal price, which erodes trust faster than an outage does.

Buyers do transact at scale once the numbers hold. In the same 2026 McKinsey survey, 73% now say they are comfortable placing orders above $50,000 online, up from 59% in 2022, though McKinsey recorded a five-point drop in willingness to spend $500,000 or more against 2024.

💡 The mechanics are covered in how B2B pricing models are built.

Your customer's purchasing process does not dissolve because you have built a portal:

• someone raises the order,
• someone else approves it against a budget, and
• a third person needs the invoice coded to a cost center before finance will release payment.

Platforms that model a single user per account push all of that straight back into email, where the order arrives as an attachment and the efficiency gain evaporates. The tell is a portal with healthy login numbers and disappointing order numbers, because buyers are using it as a catalog and placing the order elsewhere. Email wins by default.

Roles, spending limits and approval chains cost far less to design in than to retrofit, since retrofitting means re-onboarding every account.

💡 Read how the B2B buying process actually runs before designing this one.

Every order has to land in the system of record, and the question asked too late is what the storefront does during the window when it cannot get there. Retry is not a failure path.

Synchronous integration turns routine ERP maintenance into a customer-facing outage. An asynchronous queue keeps orders flowing. The cost is a harder question: what do you tell a customer about stock and price while that queue drains? Skip the conversation in design and you will have it during your first incident instead, at speed, with customers watching.

Treat the connection between ecommerce and an existing ERP as a two-way design problem rather than as a data feed.

From the CTO — Oleg Zhuk, CTO

Of the six, the one I see fail most often is the fifth. Teams design the ERP connection as a data feed, orders going one way and confirmations coming back, and the failure path gets written down as "retry". Then the first real incident arrives, usually a planned upgrade that overruns its window or a month-end close that locks the ERP for a day, and nobody has decided what the storefront tells a customer while the queue drains.

The sign shows up earlier than the incident does. Look at how customer service closes tickets: if a few orders a week are being re-entered by hand and logged as user error, those are integration failures filed under the wrong category. See that and you have weeks to fix the design instead of minutes to explain it.

It is usually the fifth and not the third because pricing errors get argued back. A customer disputes an invoice and somebody investigates. An order that never reached the ERP argues with nobody, and you find out when the buyer calls to ask where the delivery is, days after the date you promised.

Architecture gets its real test in the second market, which arrives carrying different tax rules, a different catalog, a separate legal entity and, after an acquisition, often a different ERP as well. Market two is the audit.

The choice is what stays central and what goes local:

• Centralize too much and every country queues behind one release train;
• Localize too much and you are running several platforms under a single invoice.

The consequence usually becomes visible only when a new market takes about as long to launch as the first one did, and funding stops: the business has decided that ecommerce does not scale here.