What I did when the B2B playbook backfired

For the first nine months at GlamAR, I was the only person doing sales calls. Not because I couldn’t hire. Because sales calls were the most honest product research I had access to.

When you’re selling something this complex, your sales calls are the only place you’ll hear the real objection. Customer interviews are too polite. Surveys never tell you what almost made them buy. You only learn what your product is actually missing when someone is deciding whether to give you money, and you watch them say no.

The most valuable output of a sales call is rarely the close. It’s the brief.

When AI Facial Skin Analysis became GlamAR’s top-selling product, it didn’t come from a roadmap exercise. We’d put a landing page live as a hypothesis: no product behind it, just a “request early access” form to test whether anyone would care. A partner saw it, asked if we could deliver it for an Indian beauty brand they were working with. We built it with that brand. The market told us skin analysis was the one.

The signal isn’t one buyer asking. It’s the same ask landing in three or four unrelated calls. For skin analysis, it came up across multiple conversations before we built. You can’t hear that from a deck review. Only from watching people decide whether to say yes.

You cannot run that play if your sales team is between you and the buyer.

At zero-to-one, don’t hire a salesperson until you’ve personally closed enough deals to know what the objection pattern looks like.

The standard SDK demo doesn’t close deals, because the buyer can’t visualize their version of it.

So we stopped showing the SDK. Two plays work better, both follow one rule: don’t pitch the technology, show the product.

First: a universal demo store. We built a public one at glamar.io/demo-store, a fake e-commerce site that mimics a real PDP/PLP flow. When a brand comes in cold, they don’t see a “request a demo” form. They see what their product could look like, live. Same shape as a real store, just generic products.

Second: a custom front-end, built for the buyer, before the call. This is what AI-assisted coding made cheap enough to do per prospect. For AI Skin Analysis, I started coding a front-end for each prospect: their colors, their fonts, a fake PDP with their actual products listed. A few hours of work per prospect, max. Here’s one I built for a skincare brand I was pitching. This breaks once you’re past founder-led sales. By then the demo store should be doing the same job at scale.

When the call started, I didn’t open with “let me explain virtual try-on.” I opened with: “Here’s what your skin analysis flow looks like, embedded in your skincare PDP. Walk through it.”

The buyer doesn’t have to imagine. They’re looking at it.

A working prototype shows the tech works. A real prototype shows the buyer their version works. Different things.

Never demo just your SDK. Build something closer to their product: the demo store if you’re moving fast, a custom prototype if the deal matters.

When GlamAR was 6 months old, I put up a pricing page. Featured tiers, $250/month starting price, scan limits, the whole thing. Our competitors didn’t have one. I figured “we’re transparent, they’re not” was a differentiator.

What happened: every B2B inbound conversation started with the buyer trying to negotiate down from $250. We’d been selling custom enterprise deals at premium price points. The published price became the ceiling, not the floor.

We removed the prices. Kept the page. Listed features per tier. Said “request quotation.”

Suddenly we had room to negotiate from a position of unknowns. A buyer who needed deep skin analysis integration with white-labeling and a custom front-end couldn’t anchor against $250. They had to engage with what they actually needed.

Second pricing lesson: we sold monthly when competitors were pushing annual contracts. Every B2B SaaS playbook tells you to chase ARR.

Brands that churn are your real R&D budget.

I went monthly because monthly attracts the brands most likely to churn. They onboard fast, push the product hard, and tell you exactly what’s broken when they leave. You can’t get that signal from an enterprise client locked into a 12-month contract.

A quarter of churning monthly customers gave us a faster signal loop than our annual-contract peers had access to. Then, when the enterprise pipeline matured (premium beauty deals, eyewear chains expanding regionally), the product was ready. Pricing was structured. The motion worked.

Obvious caveat: if you have brand recognition like Salesforce or HubSpot, publish. This is for when you don’t have that yet.

If you’re zero-to-one and unknown: don’t publish pricing. Start monthly. Treat your churning customers as the most honest product feedback you’ll ever get.

These deals are long. The buyer has to convince three or four other stakeholders internally. Your job is to make sure when their CMO finally raises the topic in a quarterly review, your name is the one that comes up.

You don’t do that by emailing them more.

Stop selling. Stay present.

A leading Indian beauty retailer told us no when I first pitched. Our quality, they said, wasn’t where it needed to be. The dominant incumbent in skin analysis, a global beauty-tech player, had better outputs. They went with the incumbent.

Most sales playbooks would tell you to add them to a 12-touch follow-up sequence. I did the opposite.

I stopped following up. I kept showing up in their world without asking for anything. LinkedIn engagement. Updates when we shipped meaningful product changes. WhatsApp greetings on festivals. No sales follow-ups. Just visible.

Eight months later, two things lined up. Their CMO saw a LinkedIn post about something we’d shipped. And their team had hit the wall with the incumbent: an off-the-shelf integration, no real customization, no flexibility on outputs. They were not happy. They reopened the conversation. I can’t prove the LinkedIn visibility was the deciding factor versus the product gap closing or the incumbent finally letting them down. What I can say: we were still on the shortlist when most vendors who got a “no” wouldn’t have been.

If you have access to their world: after a “no,” stop selling. Stay present. Be visible without asking for anything.

For products this new, battlecards aren’t enough. You need to physically use your competitor’s product to know what to build.

When that beauty retailer told us no, I didn’t read about the skin-analysis incumbent’s API. I bought it. Set up an account. Built test integrations. Took screenshots. Compared outputs scan-by-scan.

I’m not talking about a competitive analysis exercise. I’m talking about operating their product like a customer would, to know what felt right and what didn’t.

Most of the meaningful improvements we shipped in 2025 traced back to buying and operating it: I knew exactly what the dominant product looked like at the layer the buyer experienced it. Not from a webinar. Not from a Gartner report. From spending money to use it.

Most growth leaders know the competitor’s brand. Not their product.

That was the lever we pulled when the retailer came back to us. I opened both consoles side by side on the call. Every gap they’d flagged at “no” was gone. That’s what made it a yes.

When the product is testable: spend the money. Use the competitor’s. Then build the better one. That’s how competitive positioning actually works when your category is still forming.