What Samsung learned bringing B2C experimentation rigor to B2B

Guest: Anuradha Tempe, Lead Product Manager, Samsung Electronics America. Host: Ashley Stirrup, CMO, GrowthBook. Show: The Experimentation Edge.

Running an A/B test on a consumer storefront is the easy part. The hard part is running one on a B2B platform where a single bulk order can rewrite your results, the buyer's decision happens on a different device than the purchase, and the feature you were certain would win turns out to confuse the exact customer you built it for.

Anuradha Tempe has lived all three. A former hardware engineer who worked on chip design before pivoting to software, she is now lead product manager at Samsung Electronics America, where she manages both the B2C and B2B e-commerce platforms behind Samsung.com. On this episode of The Experimentation Edge, she walked host Ashley Stirrup through what happened when she carried a mature consumer experimentation practice into a young, fast-moving B2B business, and what the B2B side taught her in return.

B2B is not a sidekick

Samsung's consumer storefront is young but stable, with a rigorous testing practice. The B2B platform is newer and, in Anuradha's words, "a lot of interesting things" are still happening there. When she joined as one of the first product managers on the B2B side, she brought her B2C learnings with her, and the reaction was immediate: "Whoa, whoa, okay, what are we trying to do here? We don't want multiple meetings and want to introduce new processes."

Her first lesson was that the two businesses are not versions of each other. "Most of the times people mistake B2B as like a sidekick to B2C," she said. "It has a completely different consumer base, and that's something that's very important to keep in mind when designing specifically features for B2B customers."

The difference shows up in the statistics before it shows up in the features. On a consumer site, order values cluster tightly and traffic is high, so the statistical engine has plenty to work with. On a B2B site, one procurement order on one day can be worth what hundreds of consumer orders are worth. "A particular order on a one particular day can skew your results and inflate it towards a false positive," Anuradha explained. Her team now normalizes order data before reading any test, so that one customer's timing doesn't masquerade as a winning variant. "It needs a certain level of understanding to also understand how to lead A/B tests on B2B side."

That understanding extends to knowing when not to test at all. "Not everything needs to be A/B tested," she said. Experiments consume engineering, design, and stakeholder time. Lower-risk changes go to Samsung's UAT team for feedback. Some launches get a pre/post comparison instead of a randomized test. Full experiments are reserved for the features where being wrong is expensive, and where the team has the expertise to read the results correctly.

Bringing rigor without breaking the culture

Samsung is a conglomerate with counterparts in Korea and teams spread across geographies and cultures. Anuradha describes it as a huge company that works like a startup: fast-paced, chaotic, and resource-rich enough that you can build and test things. Spreading experimentation practice across that organization is a communication problem as much as a technical one.

Her approach is deliberately unforceful. "The goal is not to shove things down other people's throats and be like, 'You have to do things this way only because this is the best way,'" she said. When she shares what she has learned with other PMs, whether in the Americas or in Korea, she goes in expecting to learn something back. "If you share your knowledge with that open-ended mindset that I'm here to learn something from you as well, it goes down better always, and teams end up aligning much faster."

On the B2B platform specifically, that meant pairing the case for structure with humility toward the people who had been running the business. The pitch was practical: this is how systems scale, and for B2B to reach the next level it needs practices borrowed from the more mature consumer side. The question she kept asking was, "What can we learn from B2C without losing the soul or essence of our fast-paced B2B way of doing things?"

The add-on experiment that nearly doubled attach sales

The early win that earned buy-in came from Samsung's buying flow for hardware. Samsung.com is organized by category, subcategory, and product, each with its own buy page, which can make the experience fragmented. During flagship launches, a business customer buying a new phone might also need a security package, a software bundle, or a tool like WebEx for hosting meetings. Those add-ons were not visible where the buying decision was happening.

Anuradha's initiative was a unified add-on experience across those buy pages, with software solutions surfaced on the same screen as the hardware. But the more important insight was about which screen. "Most of the times people assume that B2B customers will only buy devices on desktop or they won't really scroll on mobile platforms," she said. "But we have a huge traffic on mobile devices as well."

The data came from daily traffic reports, and the interpretation came from conversations with sales and customer support. Business stakeholders rarely complete a purchase on a phone, because they need approvals and a procurement code. "But they will still browse the options on mobile platforms, and they'll make a decision on mobile device before deciding to go ahead with generating the code." The decision and the transaction were happening on different devices, and the mobile half was being ignored.

Her team broke the work into phases, shipped a version one, and A/B tested the add-on experience. Conversion went up. Average order value went up. Add-on sales nearly doubled. "It solved the problem of visibility," she said. Reduced scroll time meant customers saw all their options in one go, which led to follow-up questions about bundling, higher engagement, and more sales.

The decision lived on mobile and the checkout lived on desktop, and designing for that split is what moved the numbers.

The Buy Now test that lost

Ashley asked how Anuradha would coach a new product manager who was so sure their feature would win that they hadn't considered why it might lose. Her answer: "I would say I was that product manager."

The feature came from Samsung's sales and business teams. On the consumer site, customers could buy faster and earlier in the flow. The idea was to bring that to B2B. Anuradha admits she was biased toward shipping it. "I thought it's a great idea."

The A/B test disagreed. "Customers are actually getting confused by seeing the Buy Now option earlier in the journey," she said. "If you're buying, say, 100 laptops, you don't necessarily want to just buy it at the click of one button. You want to be able to see what are the specs and what color and all the different software packages."

The principle she drew from it has become a design rule for the platform: "B2B consumers value clarity over faster conversions." That is not a claim that business buyers dislike speed. They want to buy quickly, but only once they are certain of what they are buying, because returns and cancellations on bulk orders are painful. In hindsight, she thinks an engagement feature at that point in the flow would have served the customer better than a purchase button.

The test did its job. "That's where the A/B testing and experimentation leads you in the right direction," she said. Working with the data teams to normalize orders also mattered here; a million-dollar day from one customer would have made the feature look like it was working.

What she wants other PMs to take from it is less about the feature and more about the posture. "I love to make mistakes and learn from them, and that's the best way you learn. You humble yourself in the process." Sharing losing results openly with the team, she found, produced more leadership support rather than less. "The right leaders do understand where you're coming from, and you do get the support because you're not trying to withhold any information from anyone."

North Star first, guardrails second

Underneath both stories is a discipline about why a feature is being launched at all. Marketing might propose something for engagement. Sales might push for revenue. Anuradha's rule is to align on the goal before anything gets built. "If the goal is high engagement but not high revenue, then as long as we are aligned on the goal, we can proceed with that."

She gave an example where the North Star was neither. Samsung's employee partner program customers always receive a discount, and when a platform migration left certain products excluded, those customers reached cart and checkout expecting a price cut that wasn't there. Support tickets rose exponentially. The fix was messaging earlier in the flow that the product wasn't eligible for the discount. It wasn't A/B tested; a pre/post comparison of ticket volume showed the drop. The North Star was fewer support tickets, and the metric was chosen before the feature was built.

Guardrails follow from the same logic. An engagement feature isn't expected to lift sales, but it must not lower them. For B2C, that means watching conversion, NPS, and CSAT dashboards. For B2B, where there is no daily NPS view, it means keeping a direct channel open with sales stakeholders and listening for "our customer is not liking this input" or "they're not able to buy these many phones."

What comes next at Samsung

Anuradha is now experimenting on her own process. Frustrated by the time spent writing specs, she built an A/B testing copilot: feed it the roadmap and it generates hypotheses and an initial test plan she would otherwise spend 45 minutes drafting. It also triages features by confidence level, low, medium, or high, to decide which need a full experiment and which can ship through lighter channels.

She is clear about the limits. "Automating processes is obviously important, but also having that factor of trust is important," she said. A human checkpoint stays in the loop to catch hallucination and keep the model honest. "We may not need as many humans as we do today, but we still need a human touchpoint to make sure that we are not steering away from the result we actually intended to have from the get-go." Next on her reading list: sequential testing, and how AI-driven experiments might reduce manual intervention without removing judgment.

For product managers, engineers, data scientists, and growth leaders running experimentation across more than one customer type, the Samsung story is a useful check. The tooling can be shared. The rigor should be shared. The assumptions about who is buying, how, and on what device should not.

🎧 Listen to the full episode →