How Immi Turned a 17.4% TTFB Improvement Into a 7.1% Conversion Rate Lift

Immi is a low-carb, high-protein ramen brand. The category is unusual for ecommerce. Customers don't grab and go. Buying high-protein food with novel ingredients is a trust-building exercise. Shoppers research the nutrition panel. They read reviews. They browse recipes and topping ideas. They look for cooking guidance and pairing suggestions. They come back to the site more than once before subscribing, and when they do, the browse sessions are long, the bounce rates are low, and the curiosity is high.

That shopper behavior changes what speed means. On a fast-moving impulse-buy site, latency shows up in the first paint and disappears. On Immi, latency compounds across a longer, deeper journey. Every product page, every recipe view, every ingredient panel, every topping selector, every subscription choice, every return visit is a place where TTFB either supports the trust-building process or interrupts it. And once a shopper does convert, Immi's subscription model rewards the win at LTV, not just at first order. A subscriber earned today keeps paying every month after.

Immi wanted to measure Nostra's impact rigorously, without taking the lift on faith. At Nostra, they’re already a paying customer and we always have a holdout at a 95/5 split, with a statistically significant control group held back to receive the unaccelerated experience while the rest of the traffic was served Nostra's edge-accelerated version. Same site, same products, same campaigns, same promotions. The only variable was speed.

‍

Nostra's job at the network layer is TTFB. Time to First Byte is the moment the browser starts receiving content from the server, and it sits upstream of every visible thing that happens on the page. Immi's TTFB improved by 17.4% at the p75 on the Nostra side of the test. That's a material shift, sustained across real production traffic and real customer conditions.

TTFB is Nostra's primary lever, but faster first-byte reads through the rest of the render sequence. On Immi, LCP came down 8.1% and FCP came down 9.3% at the p75 as follow-on effects. Nostra doesn't optimize LCP directly. But when the server responds sooner, the browser can start painting sooner, and the hero content lands on screen faster. It's a chain reaction that starts at the network edge and shows up in the pixels the shopper actually sees.

The reason TTFB matters is because every 100 milliseconds of latency moves the conversion needle. According to Shopify's own ecommerce research, each incremental delay compounds into fewer add-to-carts, fewer checkouts started, fewer orders completed. Move TTFB down and the funnel moves with it. (Source: Shopify.)

For Immi's shoppers, that math compounds across an unusually long browse. A shopper researching recipes clicks through more pages. A shopper reading the ingredient panel loads more content. A shopper comparing topping options triggers more session activity. Every one of those touches is a place where TTFB either supports the trust-building process or interrupts it. Nostra's improvement showed up across all of them at once, which is part of why the conversion result was as clean as it was.

On Immi's holdout, conversion rate on the Nostra side lifted 7.1% over the control. Same shoppers, same products, same campaigns. Only the milliseconds were different. And in Immi's business model, that CVR lift is worth more than it looks on paper. Immi has a subscription model. A converted shopper isn't always a single order, it's a recurring one. A 7.1% lift in conversion rate compounds through subscription LTV every month after. The revenue impact of a fast site keeps paying out long after the shopper has moved on.

There's one commercial detail worth flagging. Immi brought Nostra on with no financial risk. Nostra doesn't get paid unless we're driving incremental revenue against a live holdout and continue to speed up the site year round. The 7.1% CVR lift, and every dollar of subscription LTV it compounds into, sits on top of a pricing model that only invoices when the math works.

‍

"We always have Nostra on against a 95/5 live holdout on our traffic. TTFB came down 17.4%, LCP followed 8.1%, and conversion rate lifted 7.1% on the Nostra side. That lift matters more for us than it would for most brands. Our shoppers browse recipes and reviews before subscribing. Faster milliseconds don't just win an order, they compound through subscription LTV every month after.”