Linda Bustos came on Live with Intelligems recently. She runs ecomideas.com, a database of over a thousand of the boldest, weirdest, "I can't believe they shipped this" design patterns in ecommerce. The kind of thing you screenshot and drop in a Slack channel. Here's the idea worth putting in front of this group. A lot of the boldest patterns she's archived don't last. The animated PDP gets stripped back to text. The stepped cart cross-sell gets reverted to a simple PDP upsell. And when we see that as testers, we fill in the story ourselves: it lost the test, so they killed it. But how would we actually know? Linda's point is that conversion is usually not the reason. Sometimes the creative team turned over and nobody understood the feature. Sometimes it's a performance or accessibility cost. Sometimes a replatform is coming and nobody wants to rebuild it. Sometimes engagement was low and it quietly got dropped. The feature disappearing tells you almost nothing about whether it worked. That cuts both ways. When a competitor ships something bold, you can't assume it's winning either. Which is her whole case for treating other brands as inspiration, not instruction. A pattern that's a founder-endorsed winner for a tight hero-product catalog, like True Classic switching product types right on the PDP, might fall apart on a 30,000-SKU store. Same idea, different catalog, opposite result. So the skill isn't spotting the clever idea. It's knowing which one is worth a test slot on your store, and testing it instead of reading a competitor's roadmap off their live site. The full interview is attached below if you want to watch the whole conversation. Curious how this community handles it. When you see a competitor kill a feature you liked, do you read anything into it or ignore it? How do you decide a hot idea from another brand earns the test slot on your own site? And has a pattern ever crushed it for a brand you admired and then flopped when you tried it?

Nobody in this community needs convincing that conversion rate isn't the whole story. Profit per visitor over CVR, we mostly get it. When we had @Luka Nikolić on Live with Intelligems recently, the interesting part wasn't the pitch for it, it was how much time he spent on why it's so hard to actually run a program this way. His answer came down to data and financial literacy. Getting profit into a test means getting COGS out of a client, and that's a harder conversation than it sounds, especially when they aren't used to sharing it. He's had to fight for those uploads. An agency can run a retainer for months, tweaking buttons, and never touch the number that actually decides whether the business is healthier, because nobody ever put the real financials in front of them. His pitch was to close the laptop on the AI tools for a bit and go read the boring spreadsheets instead. Return rates, supplier terms, how customers actually segment. The unglamorous stuff practitioners tend to skip is usually where the profit lever is hiding. His line for it: "revenue is vanity, profit is sanity." The full interview is attached below if you want to watch the whole conversation. Curious where this community lands. For those tracking profit per visitor, how did you get COGS out of a client who didn't want to hand it over? Have you ever shipped a test that won on conversion rate and later realized it hurt margin? And is there a "boring spreadsheet" you've gone into that ended up reshaping what you tested next?

This is a topic this community keeps coming back to. We already had a thread on it and we're taking it live tomorrow (Jul 2) at 11 am EST. Joining us for the debate: Barry Hot — $600M+ in paid media managed since 2008 across brands including True Classic, Harry's, and Athletic Greens. Founder of Hott Growth. Brings the media buyer's lens to the debate. @Nate Lagos — Former CMO at Adapt Naturals and Dugout Mugs, former VP of Marketing at Original Grain, now a performance copywriting and creative strategist. Has lived on both sides of the paid vs. on-site divide from inside the brand. Ryan Levander — Started in CRO and conversion optimization, now focused on paid media and full-funnel measurement as a fractional CMO. Brings a cross-discipline perspective to where testing and paid media collide. Total editorial freedom on this one. No agenda, no script, just three people with strong opinions and the receipts to back them up. Register here and come with an opinion! Do you think on-site testing has a real impact on paid performance, or is the correlation a red herring?

We had Barbara Bonfim on "Live with Intelligems" this month. She runs the experimentation program at Levi's, and at some point she said something that I think deserves more airtime than a 30-minute livestream: "The learning library is something no competitor can copy." She was talking about the practice of treating your test results not as a log of what you shipped or didn't ship, but as a compound knowledge base. A relational database of what you've learned about customer behavior, tagged by page, metric, segment, and user type. One-sentence insights. The rationale behind each hypothesis. What happened in Europe vs. the US. What new users did differently from returning ones. The test result is the starting point, not the output. I've seen this gap everywhere. Teams that run solid programs still tend to treat the learning phase as a formality. The report gets filed, and the institutional knowledge lives in whoever ran the test. When that person leaves, or just gets busy, the knowledge evaporates. What Barbara has been building doesn't depend on any single person. She frames it this way: the purpose of your program isn't the tests you run. It's the knowledge you compound. Two programs running the same number of tests per month can diverge dramatically in value over 18 months depending on whether one of them is building a relational database or not. She also laid out an MVP version for anyone starting from scratch: a relational database (even a spreadsheet), dropdown fields for test outcomes and next actions, an open text field for key learnings, and tags for browsing. Something the whole org can access. She's now building an AI layer on top of it so stakeholders can query it in natural language. Not just storing learnings, but making them retrievable at the moment a decision needs to be made. The full interview is attached below if you want to watch the whole conversation. Curious where this community is at with this. Are you maintaining something like a knowledge library, and who in your org actually reads it? And if anyone has experimented with putting an AI layer on top of one, I'd love to hear what that looks like in practice.