The Distribution Blog

You spent $300K on an eCommerce platform. Your customers still call in to reorder.

Here's why. Log into your own site as one of your accounts. What do they see? The same homepage every other customer sees: generic banners, a broad category menu, maybe a "Top Sellers" carousel pulled from site-wide data. Nothing that says we know your business.

Some platforms promise to fix this with "personalized recommendations." But look closer at what's powering them: they're only pulling from that account's online order history. If your customer buys 50 SKUs from you but only 12 came through the website, your recommendation engine is working off 24% of the picture. The other 38 SKUs — the ones ordered through sales reps, over the phone, via EDI — don't exist as far as your site is concerned.

That's the gap nobody talks about during the platform demo.

The result is that a fraction of all your company’s orders trickle through online. Customers search for exactly what they already know they need, skip everything else, and never discover the 200 other products in your catalog they should be buying. Your site takes orders, but it doesn't sell.

Meanwhile, Amazon Business shows up and does personalize. And your customers start to wonder why your portal feels like 2014.

This guide breaks down what's happening in B2B distributor eCommerce right now: why Amazon Business is forcing the conversation, why the platform you pick matters less than people think, what the 70–80% data gap is really costing you, how AI-powered personalization that works across all selling channels closes it, and what to look for when you evaluate vendors.

TL;DR

• The Amazon Business pressure is real. B2B buyers who shop on Amazon at home expect the same experience when they buy fasteners. Distributors who don't deliver lose share to a marketplace that already has account access.
• Your platform is table stakes. BigCommerce, Sana, Unilog, OroCommerce — they all handle catalogs, pricing, and checkout. But none of them proactively sell for you. 
• Your recommendation engine is where the revenue growth lives. Generic engines train on web orders only, while AI-powered B2B engines train on full ERP transaction history.

1. Why distributors are going online 

Amazon Business hit $35 billion in gross sales in 2023 and has since continued double-digit annual growth. For most wholesale distributors, that growth came out of categories they used to own: MRO, industrial supplies, safety equipment, electrical, fasteners. The buyers haven't changed. The way they buy has.

A purchasing manager who orders ball valves from your inside sales rep on Tuesday will buy bolts on Amazon Business from her phone on Saturday. She doesn't see two different shopping experiences, she sees one experience that's faster, cleaner, and gives her a tracking number in 30 seconds, versus another that involves leaving a voicemail and waiting to hear back.

Distribution leaders have noticed. Board-level conversations have moved from "should we be online" to "why aren't we converting more on the channel we already built." Companies have invested. Six-figure platform builds, two-year roadmaps, dedicated digital teams. The infrastructure is in place, but oftentimes the revenue isn't following.

The pattern repeats across categories. A Michigan-based HVAC/MRO distributor put it bluntly to our team: they had invested in a new website and marketing, but it wasn’t doing any suggestive selling. “It was no more than an order entry tool."  In other words, their website was essentially a working portal that functioned as a digital order form. Sure, the site worked, but it didn't earn its keep.

That's the problem this guide solves.

2. The platform conversation is mostly noise

Search "eCommerce for distributors" and the first ten results are platform comparison guides. Sana versus BigCommerce versus Unilog versus OroCommerce versus Adobe Commerce. Feature matrices. Integration checklists. Pricing tiers.

Those guides answer one question: which infrastructure to run your catalog on. They skip the harder one: once your platform is live, how does it sell?

Think of your e-commerce platform as a kitchen. It stores ingredients, manages the dining room, handles checkout. But it doesn't cook. The distributors winning on e-commerce right now aren't winning because they picked the right platform. They're winning because they put intelligent recommendation engines on top of the platform they already had.

The data backs this up. Across distributor portals running on different platforms like BigCommerce, Sana, or Unilog, click rates on platform-native recommendation widgets cluster in the 2-5% range. The platform doesn't drive the click rate. The recommendation logic does. And every major B2B eCommerce platform ships with the same kind of built-in widget: trained on web transactions, optimized for B2C-style "customers also bought" patterns.

For most wholesale distributors, that's a problem. Read more about the difference between AI-powered and rule-based recommendations in our deeper look at the future of AI in B2B eCommerce.

3. The 70-80% data gap

For most $100M-$3B wholesale distributors, online orders represent 20-30% of total volume. The other 70-80% flows through phone, counter, and rep-managed channels. Your platform's built-in recommendation engine sees that 20-30% and builds its model on it. But your full ERP transaction history (every order placed by every customer across every channel for the past decade)  stays invisible to the model that decides what to show your customers when they log in.

So when a customer who's bought hydraulic pumps from your outside rep every quarter for eight years logs into your portal for the first time, the platform-native recommendation engine has no record of them. They look like a brand-new customer with zero purchase history. The carousel on the product page shows the bestselling items across your entire customer base. None of it is relevant. 

The right competitor's rep, knowing the same customer, would have said: "You should look at our new Parker filter series. You've been buying the 941 series for years, these are the compatible replacements and they run 12% cheaper per unit."

The only way to close the gap is to train your recommendation engine on 100% of ERP transaction data. Every channel, every customer, every order, going back as far as your ERP holds data. The right engine knows what your customers bought from your counter, your reps, and your field team. It clusters customers by behavior. It identifies the patterns — this customer profile, buying this sequence of SKUs, typically adds an adjacent product within 60 days — and surfaces the recommendation at the right moment.

For more on how reorder modeling works under the hood, read how distribution reorder models increase revenue.

4. What AI-powered eCommerce does for distributors

B2C recommendation tools optimize for browsing behavior: what customers click, add, abandon. B2B distribution doesn't work that way.

Your customers aren't browsing. They're buying. They know what they need, they've bought it dozens of times, and they want to place the order fast. The intelligence play in B2B isn't "show them something they might like." It's "show them what they need and forgot to order."

AI-powered eCommerce engines built for distribution work differently from generic recommendation widgets for several reasons.

1. They train on the full ERP. Not just web orders — every transaction across every channel, going back years. A customer who's bought from your reps for a decade has a rich behavioral profile even without ever using your website. The model treats their first portal session as their thirty-second order, not their first.

2. They understand reorder cycles. If a customer orders cutting fluid every 28 days and it's been 26 days since their last purchase, a "due to reorder" recommendation surfaces at the right moment. B2C engines don't have this concept. B2B purchasing is predictable in ways consumer purchasing isn't, and the right engine takes advantage of that predictability.

3. They serve customer-specific pricing and inventory. A recommendation that shows a product at the wrong price, or out of stock in the customer's region, destroys trust. Distribution-specific engines pull live pricing and inventory from the ERP in real time.

4. They recognize order-placer behavior. Distributors name this gap directly. As one industrial distributor told us, the engine has to handle customers who reorder the same SKUs cycle after cycle differently from buyers evaluating a new product category. Generic engines can't make that distinction.

5. They surface signals to your sales team. When a customer browses a category they've never bought, that's a lead, but only if it reaches the rep covering that account. AI-powered eCommerce engines built for distribution route those signals into the CRM, so reps follow up with relevant context. See how eCommerce and distribution sales reps actually work together.

The output: customers see relevant recommendations, buy more, buy more often, and use the portal more. The portal stops being a self-service order form and becomes a revenue channel.

5. The metrics that matter

Pull together what generic recommendation widgets deliver versus what AI-powered B2B engines deliver:

• Industry baseline click rate: 2-5% on platform-native and rule-based widgets. This is what most distributor portals see.
• AI-powered B2B click rate: 13% on engines trained on full ERP history. Customers click more than double the industry baseline because the recommendations are accurate.
• Click-to-order conversion: Up to 73% on AI-powered B2B recommendations. When the recommendation is right, customers don't deliberate.
• Revenue lift: 10-17% on existing portal traffic. For a distributor doing $200M, that's $20-34M of incremental revenue from a tool running on the platform you already have.

Teams that adopt AI are seeing incredible results across their entire business. As Martina McIsaac, MSC's CEO, has said: "MSC made an early bet on Proton in 2021, believing leaps in distribution effectiveness would come from AI. That bet has paid off. Since that time, we have doubled down on AI-enabled solutions across the entire enterprise to move our business forward."

6. What's stopping distributors from making the move

Three patterns show up over and over in conversations with distribution leaders evaluating eCommerce AI:

1. ROI skepticism. Enterprise tech projects have a track record. Multi-year ERP migrations, custom development, integration work that consumes IT for quarters and delivers value years later. Distribution leaders name the skepticism directly. One HVAC distributor working with a large franchise told us: "For years we've done technology projects without ROI. This one will have ROI." The difference is the data foundation. Recommendations built on existing ERP history don't require a clean-data project, a new platform, or a rip-and-replace.

2. Sales team resistance. Reps see anything labeled "eCommerce automation" as a threat. The framing matters. eCommerce AI built for distribution doesn't replace reps. It routes signals to them. When a customer browses a category outside their normal pattern, the rep gets the alert. When a customer's reorder cycle is overdue, the system flags it. Reps cover more accounts with better context. Read how eCommerce changes (not replaces) the rep relationship.

3. Phased rollout preference. Distribution leaders almost always want to launch on the website first, prove the lift, then expand the same recommendations to the inside sales team and the counter. That phased approach is the right move, and it's how most distributors deploy AI-powered eCommerce. Website launch in 4-6 weeks, rep enablement layer 90 days later, counter layer after that. Each phase reuses the same recommendation engine and the same ERP data.

7. How fast you can go live

Traditional eCommerce platform migrations take months. Custom development takes months. Data migrations take months. By the time the project ships, the team that advocated for it has moved on.

Recommendation AI is different. It runs on top of whatever platform you already have. No migration, no rip-and-replace. Implementation runs 4-6 weeks:

• Weeks 1-2: ERP connection, data extraction, historical transaction processing.
• Weeks 3-4: Model training, recommendation logic configuration, carousel design and placement.
• Weeks 5-6: Testing, QA, staging review, go-live.

The carousels drop into existing product pages via a JavaScript snippet or API integration. The platform doesn't change. ROI tracking starts on day one of launch.

For distribution leaders who've sat through 18-month ERP projects, six weeks is defensible. "We'll be in production before end of quarter" wins budget conversations that two-year roadmaps lose.

8. How to evaluate eCommerce AI for distributors

Use these questions when evaluating vendors:

Data coverage

• Does the engine train on all ERP transaction data, or web orders only?
• What percentage of total order volume will the model see?
• How does it handle customers who've never placed a web order?

B2B-specific capabilities

• Does it support "due to reorder" recommendations based on purchase cycle?
• Does it surface customer-specific pricing and live inventory?
• Can it differentiate between order-placer and discovery shopping behavior?

Integration

• How does it connect to your ERP — direct integration or middleware?
• What eCommerce platforms does it support?
• How does it handle multi-branch inventory?

Sales team integration

• Does it route browsing signals into your CRM?
• Can reps see what customers viewed but didn't buy?
• Does it support recommendations in channels beyond the website?

Measurement

• What does the ROI dashboard show: click rate, click-to-order, revenue attribution?
• Can you A/B test placement and recommendation logic?
• What's the baseline click rate among your existing customer base?

Implementation

• How long does implementation take?
• Who manages it: your internal team or the vendor?
• What does go-live look like?

Total cost

• What's the pricing model: flat license, revenue share, or per-session?
• Does bundling with a CRM or PIM affect pricing?

The answers will tell you whether the vendor is solving the actual problem.

For a deeper look at why Amazon's recommendation engine works the way it does, read the 2 things Amazon does better than anyone else.

9. Where this leaves you

Most distributors who launched eCommerce in the past five years made a sound investment. The portal works. Customers can place orders. The platform integrates with the ERP, mostly.

The next phase isn't a platform switch. It's an intelligence layer on top of what you already built — one that turns the portal from a digital order form into a channel that sells.

The data you need is already in your ERP. The traffic you need is already on your site. The infrastructure investment is already made. What's missing is the model that connects them.

For distribution leaders evaluating eCommerce AI right now, Proton's eCommerce AI trains on 100% of ERP transaction history and runs on top of any major B2B platform. Go-live in 4-6 weeks.