The VP of Sales at an industrial distributor is reading the six-month report on the company’s new customer portal. It was not cheap, it looks modern, it is fast and it has everything the sales team asked for: catalogue, cart, order history, invoices, returns, saved lists, chat. Of 900 active accounts, 140 have logged in. Thirty-one have placed an order. Everyone else still phones their account manager or emails a spreadsheet, and the account managers still retype those orders into the ERP.
When IT asks the sales team why, two answers come back. The first: “customers say the prices in the portal don’t match what I quoted them, so they’d rather call.” The second, said more quietly: “if the customer orders on their own, who knows whether I get credit for it.”
Neither problem has anything to do with screens. This article is for sales directors and CTOs who are planning a B2B customer portal, or trying to rescue one that never found its users. It defends one position that many vendors will disagree with: a B2B customer portal is a data and permissions project, not an interface project, and its adoption depends on whether it does one recurring task faster than a call to the account manager, and whether the salesperson benefits when it does. That is why the first release should be narrow and built on data customers can trust, rather than broad and impressive.
How is a B2B customer portal different from an online store and from a CRM?
An online store sells a product to an anonymous buyer at one public price. A B2B portal serves a known account with which you have a contract, payment terms, a credit limit, a customer-specific price list and years of history. In a store, the user is a person. In a portal, the user is an employee of a customer company acting on its behalf, within limits: allowed to order up to a certain value, to see invoices for one branch only, or only to raise returns.
A portal differs from a CRM in who uses it. Your CRM is your salespeople’s tool and describes the relationship from your side. The portal is a window through which customers see part of the same data from their own perspective: what I ordered, what is on its way, what I owe, what I reported. If you are considering a custom CRM alongside the portal, read our piece on when custom CRM development pays off first, because the two decisions shape each other.
In practice, a B2B portal usually covers several self-service areas:
- ordering: catalogue with customer prices, cart, quick order by SKU, reorder from history, file upload of order lines, approval workflow;
- status and delivery: fulfilment status, dates, shipping documents, proof of delivery;
- finance: invoices, credit notes, open items, due dates, credit limit usage;
- returns and claims: submission with attachments, status, correspondence;
- documents: certificates, safety data sheets, contracts, price lists;
- account administration: users on the customer side, their roles, delivery addresses, sites.
Nobody needs all of these on day one. Choosing where to start is the most important decision in the whole project.
Why is a B2B portal a data project rather than an interface project?
Because the portal creates almost no data of its own. Prices, availability, delivery dates, invoices, open items and order statuses live in the ERP, the warehouse system, the billing system and the CRM. The portal displays them and accepts a few kinds of input: orders, claims, account changes. Everything a customer sees is only as good as the system it comes from.
This is the trap the distributor above fell into. As long as customers phoned their account manager, the account manager was a filter. They knew that stock in the ERP was refreshed once a day, that this customer’s price was in a side agreement nobody had keyed into the system, and that “3 days” really meant a week. They corrected the data from memory before it reached the customer. A portal removes that filter. It shows the customer the ERP’s version of the truth, uncorrected, and if the ERP is wrong, the portal is wrong faster and in writing.
So the first task in a portal project is not wireframes but a data inventory. For every piece of information the portal will show, answer four questions: which system is the source, is the data complete there, how often does it change, and is there programmatic access to it. The answers are often uncomfortable. Customer price agreements kept in salespeople’s spreadsheets. Stock levels without reservations. An ERP with no documented API, reachable only through its database or file exports.
Each of those answers moves scope and cost more than any decision about look and feel. If your ERP itself needs a decision about extension or replacement, start there; we covered it in our article on whether to build, extend or buy a custom ERP. A portal built on a system that will be replaced next year will need much of its integration rebuilt.
This does not mean the interface is unimportant. B2B buyers now carry expectations shaped by consumer apps, as we argued in our piece on the UX revolution in B2B. But a good interface on unreliable data loses to a phone call every time. The reverse often works: a modest interface on data customers trust can win users, because it saves them time.
Which feature should come first: ordering, invoices or claims?
Instinct says start with ordering, because orders bring revenue. That is often a mistake. Ordering is the hardest portal feature: it needs current customer-specific prices, availability, minimum quantity rules, delivery terms, credit limits and approval flows. It is also the feature where an error costs most, because an order placed at the wrong price becomes a dispute with a customer.
A better test for the first feature is this: which task is frequent, repetitive and currently costs your team the most time, and can it be served from data you already trust? In many companies the answer is “where is my order?” and “please resend that invoice”. Answering those questions consumes hours of account manager and customer service time every day, and the data needed to answer is in the ERP and usually correct, because it describes things that have already happened.
Before choosing, measure. Ask customer service and account managers to log every customer call and email for one week under a handful of types: status, invoice, new order, claim, price question, other. After a week you have a distribution showing where self-service would relieve the team most. That is cheaper and more reliable than workshops where each department presents its wish list.
It is also worth looking at how your customers order today. Eurostat data for 2024 show that e-sales accounted for 19.49% of EU enterprise turnover, of which 11.07 percentage points came from EDI-type messages and 8.39 points from websites and apps (Eurostat, E-commerce statistics). For large enterprises, EDI accounted for 14.25% of total turnover, most of their electronic turnover. The practical conclusion for a portal: your largest customers probably already order machine-to-machine, from their system to yours, and the portal should not try to pull them away from that. The portal serves the middle and the long tail of your account list, the customers who order today by phone, email or spreadsheet. Their habits determine what must work on day one.
A typical, sensible sequence is therefore: first, status and document visibility (invoices, shipping documents, certificates); then claims; then reorder from history; and a full catalogue with cart only once prices and availability are reliable. This order is not dogma. In a company where claims are the largest service cost, claims come first.
How do you model companies, sites and users with permissions?
This is the second part of a portal that wireframes do not show, and it consumes a surprising amount of time. A B2B customer is rarely one person. It is a company with a head office and branches, with procurement, accounts payable and warehouse managers who each need something different. Sometimes it is a group with several legal entities, each with its own tax ID and terms, but one person who orders for all of them.
The permission model has to be designed on paper before the first screen is built. The minimum vocabulary is: account (a legal entity with terms and a credit limit), unit (branch, ship-to address, cost centre), user (a person with a login) and role (what the user may do and across which units). On top come business rules: order value limits per role, approval flows above the limit, price visibility (not every customer employee should see commercial terms), invoice access only for accounts payable.
The most consequential decision here is who administers users on the customer side. If every new or departing customer employee requires an email to your service desk, a few hundred accounts produce a permanent ticket queue, and former customer employees keep active logins for months. A better approach is delegated administration: each account has at least one administrator who manages their own users within the limits you set. Your largest customers may also expect single sign-on through their own identity provider (OpenID Connect, built on OAuth 2.0) and automated provisioning and deprovisioning via SCIM (RFC 7644). You do not have to build this in the first release, but the architecture should not rule it out.
Why so much attention? Because access control failures are the most common serious security problem in web applications. The OWASP Top 10:2025 ranks Broken Access Control first, and OWASP reports that 100% of the applications tested were found to have some form of broken access control (OWASP A01:2025). In a B2B portal the typical scenario is simple and dangerous: a user at one customer changes an invoice or order number in the URL and sees another customer’s documents, prices included. Leaking one customer’s price list to another is not just a security incident but a commercial problem no apology will fix. Permission testing must therefore cover not only “can users see their own screen” but above all “can they see anyone else’s data under any combination of parameters”.
Which ERP and CRM integrations do you need, and which system is the source of truth?
Integration is the core of a portal and the main component of its cost. The basics of systems integration are well known: APIs, service buses, integration platforms, file exchange. In a B2B portal the difficulty is different. For every type of data you must decide which system is the source of truth, which way data flows, how fresh it must be and what happens when synchronisation fails.
The table below shows a typical arrangement. It is not a template to copy but a list of decisions that must be made explicitly.
| Data | Source of truth | Direction | Required freshness | When integration fails |
|---|---|---|---|---|
| Account, payment terms, credit limit | ERP | ERP → portal | minutes to an hour | portal shows last known state with timestamp |
| Customer-specific prices and discounts | ERP (sometimes a separate pricing engine) | ERP → portal | current at order confirmation | order accepted with price to be confirmed |
| Availability and dates | ERP or warehouse system | system → portal | minutes, verified at order | availability flagged as indicative |
| Orders | portal (until accepted), then ERP | portal → ERP, ERP statuses → portal | seconds to minutes | order queued, customer sees “received for processing” |
| Invoices and credit notes | finance system / national e-invoicing platform | system → portal | hourly | invoice list with delay notice |
| Open items and payments | finance system | system → portal | daily | balance with calculation date |
| Claims and returns | portal or service system | bidirectional | minutes | claim stored in portal, forwarded when the link recovers |
| Account manager, sales notes | CRM | CRM → portal (selected fields) | daily | account manager from last sync |
A few rules save the most money. First, the portal should not calculate prices itself if the ERP can. Duplicating discount logic in two places guarantees that sooner or later they diverge. It is better to ask the ERP for the price for a specific customer and basket, even if that needs a middleware layer and caching. Second, every integration needs failure handling that people can see: a queue of orders that did not reach the ERP, an alert for the team and a status customers understand. Third, identifiers: the customer number in the ERP, the CRM and the portal must map unambiguously, which is rarely true in a company that has been operating for years.
The CRM is usually a supporting player. The portal can push events into the CRM (logins, abandoned carts, new claims) so the account manager sees what is happening at the customer, and pull the account manager and customer segment from the CRM. Making the CRM the source of prices or statuses rarely makes sense if those live in the ERP.
How do customer-specific pricing, availability and delivery dates complicate a portal?
These are the three areas where B2B portals most often run over budget, because their complexity only surfaces when you work with real data.
Customer-specific pricing in B2B is rarely one table. It is usually a stack: base price list, customer discount group, product group discount, contract price on selected SKUs, volume breaks, time-limited promotions, sometimes a price that depends on currency and the day’s exchange rate. Some of these rules live in the ERP, some in contracts, some in salespeople’s heads. Before building a priced catalogue, audit how many pricing mechanisms you actually use and how many are in the system. Every mechanism that is not in the system must either be put into it or consciously dropped. There is no third option in which “the portal will somehow handle it”.
Availability is the second trap. Stock on hand is not the same as availability to a customer. You have to subtract reservations, goods in quality inspection, stock allocated to key accounts, and sometimes add inbound deliveries. If the ERP calculates availability once a day, the portal must not pretend to show live stock. It is more honest to show “available in the central warehouse as of 6:00 a.m.” than a number that misleads.
Delivery dates complete the list. A B2B customer plans its own production or its own deliveries around them. A date shown in the portal becomes a commitment in the customer’s eyes, even if your terms say it is indicative. If your logistics cannot calculate dates reliably today, it is better for the first release to show order status and a date confirmed by customer service than an automatic calculation.
The common thread is the same as in the rest of this article: the portal exposes the state of your data. Often the biggest benefit of a portal project is not the portal itself but the fact that the company finally cleaned up its pricing and availability, because it had to.
What do e-invoicing mandates, accessibility and privacy rules change in scope?
For portals serving European customers, the biggest change of recent years is mandatory e-invoicing. Germany has required businesses to be able to receive structured e-invoices since January 1, 2025. Belgium made B2B e-invoicing mandatory on January 1, 2026. In Poland, KSeF became mandatory for issuing invoices on February 1, 2026 for taxpayers whose 2024 sales exceeded PLN 200 million, with receipt through KSeF mandatory for all taxpayers from the same date, and issuing extended to everyone else on April 1, 2026, with transitional exceptions for the smallest amounts until the end of 2026 (KSeF, scope of mandatory e-invoicing). France began its rollout in September 2026, and the EU’s VAT in the Digital Age package (Directive 2025/516) makes e-invoicing the norm for intra-EU B2B supplies from July 1, 2030.
What does this mean for a portal? In those markets, a PDF in the portal is no longer the invoice in the legal sense, only a readable copy. The legally binding invoice is the structured one exchanged through the prescribed channel, and your customer will receive it there anyway. The portal still has value, but a different one: it ties the invoice to the order, the delivery and any claim, shows payment status and balance, and offers related documents. Showing the official e-invoice reference next to each invoice helps the customer’s accounts payable team match it. A project that assumes “the portal as the invoice delivery channel” should be rethought.
In the United States there is no federal B2B e-invoicing mandate, so for US accounts the portal often remains the main place where customers retrieve invoices. If you serve both markets, design the invoice area so that it works as a view on top of an e-invoicing platform in Europe and as the primary document store in the US.
Accessibility is a topic where it is easy to overshoot in both directions. The European Accessibility Act (Directive 2019/882), applied since June 28, 2025, covers e-commerce services provided to consumers. A purely B2B portal, where only businesses buy, generally falls outside that scope. In the US, accessibility claims under the Americans with Disabilities Act mostly target consumer-facing sites. Neither is a reason to ignore accessibility. Many distributors sell through the same portal to sole traders and consumers, and then the rules may apply. And your customers’ employees are people with varied needs; basic WCAG conformance (contrast, keyboard operation, proper forms) costs little when planned from the start and a lot when bolted on after a year. A shared component library helps; we wrote separately about when a design system pays back.
Privacy in a B2B portal is mostly about customer employees’ data: names, email addresses, phone numbers, login history. Under the GDPR you need a lawful basis, retention periods, rules for deleting accounts and access logging. In California, business contact data has been in scope of the CCPA since its B2B exemption expired at the start of 2023. None of this is a large piece of work, but it must be done before launch, not after the first questionnaire from a customer’s legal team.
What decides whether customers start using the portal?
A B2B portal does not win users because it exists. A customer who has phoned an account manager for ten years has a habit and a relationship, and changing the habit costs effort. They will change only if the portal gives them something clearly better. In practice four things decide adoption, and none of them is a system feature in the narrow sense.
First: one task faster than a phone call. On first use, the customer must get something done faster than before. Order status in ten seconds instead of an email and a reply three hours later. All invoices for the quarter in one file instead of a request to your accounts team. Last month’s order repeated in one click. If the first experience is slower or less reliable than a call, the customer will not come back.
Second: data that can be trusted. One instance of a portal price differing from the invoice is enough for a customer to stop believing the portal and start checking everything with the account manager. At that point the portal adds work instead of removing it. That is why a narrow first release on reliable data beats a broad one on partly correct data.
Third: active onboarding. Emailing an invitation to all customers is not enough. Onboarding led by account managers works better: creating the account during a call, showing one task that will save the customer time, and following up after two weeks to ask what is not working. Small things help too: a deep link to the specific order in every confirmation email, a link to the invoice in every payment reminder, passwordless magic-link login for customers who visit once a month.
Fourth: measuring the right things. The number of registered accounts tells you little. Better measures are the share of active accounts (logged in within 30 days), the share of orders or enquiries handled through the portal in a given customer segment, the number of status calls and emails reaching customer service, and the time from invoice issue to first view in the portal. These show whether the portal is actually changing how people work.
All four conditions need decisions made outside the development team: in sales, customer service and finance. That is why the business owner of the portal should be someone from sales or customer service with a mandate to change processes, not only the IT department.
Why can salespeople block a portal, and how do you prevent it?
This topic rarely appears in project documentation, yet in many companies it is decisive. A salesperson whose pay depends on sales attributed to them looks at a portal with suspicion. If the customer orders on their own, will the account owner get credit? If the customer stops calling, will the account manager lose the relationship and influence? If the portal shows the customer a price list, does the salesperson lose room to negotiate?
When these questions go unanswered, salespeople do not sabotage the portal openly. They simply do not encourage customers to use it, keep taking orders by phone and listen with some relief to complaints about the portal. A portal without sales support, in a company where customer relationships are personal, tends to stall at the small group of customers who preferred working on their own anyway.
The solution lies in sales policy, not in code. First, an order placed through the portal counts for the account owner exactly like a phone order. It is a simple rule that must be announced before launch, not after the first complaint from a salesperson. Second, the portal is also a tool for salespeople: the account manager sees what the customer browsed, what they left in the cart, which claims are open, and can place an order on the customer’s behalf in the same system. Third, the time freed from retyping orders and answering status questions should be explicitly redirected to selling: account expansion conversations, new business, key account care.
For a sales director this is the most important argument. The portal does not replace the salesperson. It takes away administrative work they do reluctantly anyway, and leaves time for the work the company pays them for.
Should you build a portal, buy a platform or extend your ERP’s module?
Before commissioning a custom portal, check three alternatives. The choice depends on how far your processes depart from the standard and how well your ERP exposes data.
| Option | When it fits | Main risk | Who controls the roadmap |
|---|---|---|---|
| Portal module shipped with the ERP | simple processes, ERP with a decent self-service module, customers accept standard look and flows | the module’s limits appear at the first non-standard requirement | ERP vendor and implementation partner |
| SaaS B2B commerce platform | catalogue selling, many SKUs, standard pricing and order flows | licence cost grows with revenue or users, ERP integration is still your work | platform vendor |
| Custom-built portal | unusual processes (product configuration, project-based ordering, service), complex permission model, portal as competitive advantage | build and run cost on your side, needs a stable team | you |
| Hybrid | platform or module for catalogue orders, custom components for unusual processes | two systems to maintain and keep visually consistent | shared |
Note that ERP integration appears in every option. A SaaS platform has a ready-made cart and catalogue, but it does not know your price agreements, your credit limits or the way you calculate availability. That data still has to be fed into it. So the comparison “subscription platform versus custom build” is often misleading: in both cases the largest item is the same integration work.
A custom portal makes sense when self-service is a source of advantage: customers configure products, order against a project schedule, manage a fleet of equipment, report consumption. We set out the broader criteria for this decision, including risks and engagement models, in our guide to custom software development.
What makes up the cost of building and running a B2B portal?
The cost of a B2B portal is calculated like any software project: team composition times duration times rate, plus external costs and contingency. But rate arithmetic matters less than what drives the number of person-months. In a B2B portal those drivers are mostly things wireframes do not show.
The biggest cost drivers are:
- access to ERP data: a documented API is one level of cost; integrating through database views, files or a new middleware layer is a different one;
- the number of pricing mechanisms: every type of discount and contract price that must be handled and tested on real data;
- the permission model: a flat “one customer, one user” model is simple; a hierarchy of entities, sites, roles, limits and approvals is a module of its own;
- the scope of the first release: status and document visibility is a fraction of the work needed for a full cart with prices and availability;
- data quality: cleaning customer master data, SKUs and price lists, which usually falls to you but delays the whole project;
- non-functional requirements: SSO with the customer’s identity provider, multiple languages and currencies, high availability, penetration testing.
The interface itself costs less than people expect, especially with a ready component library.
The build cost must be supplemented by items that often vanish from proposals. Running costs: hosting, monitoring, library updates, bug fixes and adjustments to ERP changes. Evolution: a portal that wins users immediately generates new requests, and freezing it means slowly declining adoption. Internal costs: the time of the product owner, the account managers running onboarding, and the ERP team that must prepare data and interfaces. In many projects this last item is the biggest surprise, because the ERP team’s calendar is already full.
A credible estimate for a specific company only emerges after the data sources have been checked. So rather than asking vendors to price “a portal with ordering, invoices and claims”, commission a short analysis: data inventory, permission model, choice of first feature and integration architecture. At ARDURA, the discovery phase in software development projects takes one to two weeks; we described how it runs in our article on the discovery workshop. The output is a scope and an estimate that can be compared across proposals, because everyone prices the same assumptions.
How should you plan the first 90 days and measure the result?
A B2B portal is best launched in stages, with a decision after each one. A plan for the first three months might look like this.
Weeks 1–2: discovery. One week of logged customer enquiries, data inventory, choice of first feature, the model of accounts, units, users and roles, a decision on the source of truth for each data type, an assessment of ERP access. Exit criterion: you know which task the portal will take over first and where its data will come from.
Weeks 3–8: first release. Login, the permission model with delegated administration, read integration with the ERP for the selected data, the first self-service feature, permission testing on production-like data. The team works in two-week sprints and demonstrates a working portal after each one, ideally to customer service staff who know what customers ask.
Weeks 9–12: pilot. Launch to a few dozen customers chosen together with account managers, including some sceptics. Personal onboarding, weekly review of feedback, fixes. At the end, measure: how many pilot customers use the portal actively, whether status enquiries from that group have dropped, which data errors have surfaced.
The decision after 90 days is not whether the portal “worked” but what comes next: roll out to all customers, add the next feature, or first fix the data the pilot exposed. The pattern resembles building an MVP, with one difference: B2B portal users are your existing customers, so the pilot must be stable from day one. You cannot experiment at the expense of relationships built over years.
Define the post-launch metrics before launch and give each an owner: share of active accounts in the target segment, share of orders placed through the portal, number of status enquiries and invoice copy requests reaching customer service, number of ERP synchronisation errors per week, claim handling time. Review them against your assumptions each quarter and decide on the next stage.
What are the key takeaways before deciding to build a portal?
A B2B customer portal looks like an interface project but is a project about data, permissions and changing habits. Companies that treat it as a pretty store for business customers pay for broad scope and then watch customers drift back to the phone. Companies that start with one task on reliable data build less and win more users.
Remember seven things. First, before any wireframes, take a data inventory: for each piece of information, establish the source of truth, completeness, freshness and access method. Second, choose the first feature from a week of logged customer enquiries, not from departmental wish lists; most often that will be status and documents, not the cart. Third, design the model of accounts, units, users and roles on paper, with delegated administration on the customer side, and test above all that nobody can see anyone else’s data. Fourth, do not duplicate pricing logic in the portal if the ERP can calculate the price. Fifth, in e-invoicing markets treat the portal invoice as a view that complements the mandated channel, not as the delivery channel. Sixth, announce before launch that portal orders count for account owners, and make the portal a tool for salespeople too. Seventh, launch in stages, with a pilot on selected customers and metrics defined in advance.
If you are planning a portal for your customers, or want to find out why an existing one never won its users, see how we work in custom software development and platform engineering, or book a call. Discovery takes one to two weeks and ends with the choice of first feature, a permission model, an ERP integration architecture and an estimate you can compare across proposals.