We hit $1M ARR with Tectonic. Then we took it back to zero to build Spectrum.

There is a version of this story we have not told publicly yet. It is the part where, a few months ago, we wound a $1M ARR business back to zero on purpose. The version that asks our customers to start over with us. The version where we changed almost everything about how we go to market, while keeping every conviction we had about what was actually broken. This is that version.

Founders love telling lightning-strikes pivot stories. Ours was not that. The recognition was not dramatic. It was the steady, uncomfortable accumulation of "this is not where we add value," meeting after meeting, customer after customer, until the call to change shape was the only honest one left.

The hypothesis we made three years ago

When we started Tectonic, we believed two things. The first was that the Shopify growth stack was structurally broken, because the merchant was the integration layer between a dozen apps that did not coordinate. The second was that fixing it required owning more of the storefront than apps alone could touch.

So Tectonic was an AI-native storefront. We cloned the brand's existing Shopify frontend, ran it on a unified frontend layer on top of Shopify's commerce backend, and rebuilt three things at the same time: fast digital storefronts, growth features, growth agents. The backend stayed on Shopify. The storefront became Tectonic.

We earned scale the only way we knew how. Performance-gated traffic ramp: 1% A/B, then 5%, 20%, 50%, 100%. Each step required proven lift on revenue per session, AOV, add-to-cart, conversion. Brands earned the right to scale with us through performance, not through contract length. The model worked. Dozens of D2C Shopify brands signed on across apparel, beauty, personal care, home goods, and lifestyle, mostly India-based with select US brands. We hit $1M in ARR.

The diagnosis was right. The structural pain was real. The customers were real. The growth lift was real. We had built a thing that did the work it promised to do.

The pattern we kept hearing

Three years of operator conversations crystallised into a small set of phrases that operators kept repeating, almost verbatim, on call after call. The vocabulary was nearly identical across two countries and five verticals. The thing we were solving had a language.

When the same phrases keep coming back, almost word-for-word, across an audience that does not know one another, the problem is not idiosyncratic. It is structural. We had been hearing it for three years. The thesis was earned, not invented.

The shape of the work versus the edge

Here is the part that took us a while to admit. We were great at growth. We were also spending most of our time not doing growth.

The frontend replacement model was heavy. Every brand engagement opened with the same sequence: clone the existing storefront, port the theme primitives, rebuild the integration points, get to feature parity, then earn lift on top. The migration was solvable. It was also where most of the calendar went, on our side and on the brand's. We were doing months of frontend and agency-style UI work before the growth work could start. We were becoming something we had not set out to become: an agency with an excellent thesis.

Our edge was the growth layer. The agents, the experimentation harness, the brand-context system, the cross-surface coordination. That is where we added value the way nobody else did. But the path to those layers ran through a months-long migration that gated everything else. The thing we were best at was being delayed by the thing we had built around it.

When your shape diverges from your edge, you can do two things. You can keep doing the heavy lifting because that is what your shape requires. Or you can change the shape. The first is what most companies do. The second is what discipline looks like.

What 2024 made expensive that 2026 made cheap

There is a structural reason the original shape was right when we built it, and wrong by the time we changed it. In 2024, delivering AI-driven growth to a Shopify brand actually required owning the storefront. The model capabilities, the latency, the cost of running agentic systems at storefront scale, the maturity of the Shopify APIs, the depth of the metaobject and metafield layer, all of it pointed toward "you need your own surface."

By 2026, the math had moved. Frontier models could run agentic workflows reliably at production cost. Shopify's own data layer had matured. The APIs supported the depth of customisation we needed without forcing a frontend replacement. The capability surface had quietly walked past the original bet.

Timing is part of the product. When timing changes, the shape changes. The growth bet stays. The execution moves to where the friction is lower. The thing we wanted to deliver, AI agents running the growth work for D2C brands, was now deliverable without months of frontend migration. So the decision was clear: change shape, keep the bet.

The hard call: scale to zero

The conversation took a month inside the founding team. Stick with Tectonic and tweak. Run both products in parallel. Wind Tectonic down and start Spectrum cleanly. We argued through each. Many disagreements. Several agreements. The shape that emerged was not designed in a single whiteboard session; it surfaced as we kept testing each option against real operator pain.

Running both in parallel was the most tempting option. We had a $1M ARR business with a working delivery model. Keeping it running while building Spectrum was the safe call. It was also the wrong one. Two products meant divided focus. Divided focus meant Spectrum would ship at half speed. And the customers we were running Tectonic for needed the new shape more than they needed the old one. Splitting our attention to protect revenue would have failed them, and us, slowly.

So we scaled Tectonic back to zero. We told customers what we were building. We told them why. We offered each of them a path into Spectrum on terms that made sense for their store. And then we did the thing nobody talks about in pivot stories: we waited to see who would come.

The validation we did not control

Over 90% of the brands we ran Tectonic for have signed on to continue with us on Spectrum. That is the only validation that matters. If the original thesis had been wrong, our brands would have left. If the execution had been right and only the timing had drifted, they might have stayed without enthusiasm. What they actually did was tell us, in numbers, that the bet was right and the new shape was the version they had been waiting for.

A subset of those brands are already running pilots on Spectrum. The phased rollout (two apps, five use cases, duplicate theme, A/B against the live store, expand on proven lift) is the same earned-trust model we ran at Tectonic, just deployed without the frontend migration. The discipline of "prove lift, then scale" did not change. The shape it ships in did.

What is different now

Spectrum is a lighter shape for the same bet. Code lives in the brand's Shopify theme, not in a vendor sandbox. The brand keeps every app it loves; we replace only the apps that show up to a job no app is good at, which is reading the whole customer profile and writing across every surface. The pilot starts in the brand's actual store, not a year-long migration. Growth work starts on day one.

The pricing inverted with the model. Tectonic was a services-tinted SaaS bill: you paid us for a thing, and the thing took months to install. Spectrum is software-free, intelligence-priced. Every app is included at every tier. You pay only for the AI work the agents do across your store, in credits, at a per-credit rate that drops as you scale. The pricing matches the product. Both became lighter together.

Why we are building this in public

There is a temptation, after a story like this, to ship the new product quietly and re-emerge a year later with a polished launch. We are doing the opposite. Spectrum ships in public, week by week, kit by kit. Eight Optimization Kits over eight weeks, every drop accompanied by a teardown of what shipped, what worked, what surprised us, what is harder than we thought.

Two reasons. The first is honest: we got the shape wrong once, and the only way to keep getting closer to the right one is to ship in front of the customers who tell us when it is wrong. Build-in-public is not a marketing technique for us. It is the operating model. The second is product-led: the work itself is the best demo. Every drop is a real capability, available to the brands running with us, audited by them in real time. The campaign is the product, and the product is the campaign.

We are not done. We are at the start of the version of this that we should have built first if 2026 had been here in 2024. The work the agents do compounds every week. The brands we are running with are part of the iteration loop, not a launch list. The next kit drops on the Wednesday after this post. Read along, or watch from a distance. Either way, we are shipping in front of you.