Perxify

After NFT Tallinn in 2023, brands kept asking me how to launch membership and reward programmes like the ones they had seen from airBaltic and MetaBrewSociety. I researched providers and made introductions. Some brands came back when a provider changed direction or the solution did not fit. There was enough repeated interest for Alvar and me to start working on our own platform.

We wanted a brand to be able to set up membership tiers, reward rules and customer accounts from a dashboard. Transferable rewards were part of the idea, but customers should be able to use them without learning about wallets or blockchain. As we worked through the purchases behind those rewards, the catalogue, payment and delivery requirements became a much larger part of the product.

In January 2026, we approached customers with a complete e-commerce store, marketplace or loyalty programme built on our stack. In May, we refocused on selling the stack through APIs. Over the summer, we clarified the offer; by late August, we had agreed to start with Codex, Stock, Insights and Hugo, plus a storefront for manufacturers. We planned to combine the software with the initial sales work and earn through transactions. We did not launch that manufacturer programme.

At first, we thought we could connect existing tools and have an early version ready within a month or two. That was optimistic. The expected June 2024 launch slipped, and by August we were rebuilding. I worked on customer needs, the offer and priorities, while Alvar and the engineers led architecture and development. Each change also meant explaining a different product to prospective customers.

We built catalogue and order management, customer profiles, automations and marketplace functions. Products could have variants with their own prices. We supported vouchers with ownership histories, barcode-assisted listings, offers grouped across sellers, wishlists and multi-cart orders. API access allowed the storefront to run separately from these systems. Enterprise payment and delivery requirements still needed substantial work in 2025.

Raamatuhai gave us somewhere to put the infrastructure to use ourselves. We began it in spring 2025 and had launched BooksShark by December. The marketplace supported photo-assisted listings, price suggestions, checkout, shipping handoff, payouts and returns. AI catalogue enrichment, deduplication and exception handling were already in live use alongside Stripe Connected Accounts and DPD. We also developed buyer assistance based on product data, with questions passed to a person when needed, and reporting from structured commerce data.

For manufacturers, we organised catalogue data in Codex, availability in Stock, buyer questions in Hugo and reporting in Insights, brought together in the Perxify Dashboard. We planned a reusable storefront and six months of search, content, campaign and sales work, followed by a handover to the customer's team. We still needed to complete the standard manufacturer configuration and end-to-end integrations. A self-service Cloud offer and further modules remained on the roadmap.

Seeing people list, buy and receive books through software we had built was the most satisfying part for me. Running Raamatuhai let us see how catalogue data, payments and delivery behaved together in everyday use. All marketplace workflows were automated except withdrawals.

We also found a more specific customer for the wider infrastructure. The manufacturer programme was designed for a business with proven products, an existing direct-sales channel, usable product data and a team that could take over its operations. I helped define that offer and the responsibilities around it. We planned to configure a shared system, which gave us a clearer scope than the broad enterprise discussions we had started with.

I defended the expanding scope because requests came from customers. Many of those requests made sense, but we still had to decide what a small team could finish. We spent too long treating customer interest as a reason to keep adding work. In the manufacturer programme, we finally set firmer limits around standard configuration and excluded bespoke design and new launch-time integrations. I would have made those decisions earlier.

Our pricing work exposed another constraint. We proposed no monthly platform fee and planned to agree twelve historical monthly direct-sales figures, so each future month's sales could be compared with the corresponding historical month. We would take a share of baseline sales and progressively lower shares of growth, collected at the transaction. The model reduced the customer's upfront commitment, but left us financing the implementation and six months of commercial work before the resulting income arrived.

We set the 2026 fundraising terms in January: a €1.0 million target, a €1.2 million hard cap and a €4.0 million pre-money valuation. Those terms never changed. We updated the sales plan as we learned more about the customers and the offer. Our manufacturer programme later targeted a first launch in October and 18–20 live customers generating €180,000–€200,000 in monthly recurring fees by August 2027. These were targets for a funded programme, not results we had achieved.

We had resumed fundraising in January 2026, and eight months of work produced no financing. Of three potential leads, one cancelled, one lost funds through bad investments and one delayed until 2027. I had been too patient with discussions that kept giving us reasons to wait. After bootstrapping since 2024, we paused active development in October. We needed external funding for at least six months of operations before we could responsibly commit to customers.

I now give more weight to the work a customer would depend on after launch. With Perxify, a product request often led into catalogue quality, payment rules or delivery exceptions that were easy to underestimate in a demo. Running Raamatuhai made those dependencies much easier to see. I would test a complete purchase and its exceptions earlier, before agreeing to more features.

I would also put an earlier limit on how long we could keep building while waiting for investment. The manufacturer programme combined implementation, ongoing software operations and six months of sales work. We needed enough cash to cover those commitments even if sales took longer than expected. Funding discussions should have had a clearer deadline alongside the product plan.