Case study 05
A $700k-a-week meal delivery brand fell to $64k, then rebuilt around lifetime value.
A meal delivery brand built for fitness, muscle gain, and weight loss, chef-made meals from local ingredients, had been the category leader in the US for years and grew explosively in 2020. Then it declined, slowly, over the years that followed, until weekly revenue that had once touched $700k was down to $64k. Lifetime value against acquisition cost had fallen to 0.7, which meant the average customer cost more to acquire than they were worth. The brand had plenty of signups. Ads worked, and its referral program was so generous it brought people in without ever teaching them to actually buy.
At the trough
$64k a week in revenue, down from a $700k peak
LTV to CAC ratio at 0.7, losing money on the average customer
A referral program so generous it attracted signups with no reason to buy
No lifecycle at all: no welcome sequence, no reactivation, ads that never changed
6 months later
$300k a week in revenue, still below peak but profitable
LTV to CAC ratio at 14
Annual lifetime value climbing toward $9,000 per customer
A $4,000 weekly ad budget doing the work a much larger one used to
Finding out who was worth keeping
Every customer got sorted into a pool by lifetime value and by tenure. Most of them turned out to have placed a handful of orders and disappeared fast. Quality had slipped over time, but the bigger problem was silence: no lifecycle, no fresh ad creative, and no affiliate strategy at all.
A lifecycle system that didn't exist before
Cash stopped going to ads first. In its place: an automated welcome sequence that educated new customers, then followed up gently or aggressively depending on their profile, seasonal promotions (standard for a food brand, but never built here), and reactivation sends at 30, 60, and 90 days.
A referral program worth keeping, not killing
The old double-sided referral bonus got dialed back from its overly generous terms, but kept alive, because it was still what got people to activate and feel like they belonged to the brand. Social content shifted too, aimed at existing customers instead of cold strangers.
Selective acquisition, once the lifecycle proved out
Weekly revenue had already doubled within the first 90 days on lifecycle alone. From there, acquisition got rebuilt around quality: heavy copywriting on ads instead of anything trying to appeal to everyone, plus influencer partnerships that hadn't existed before and cut affiliate acquisition costs by 5 times.
On day 100, a monthly subscription option went live, auto-activated the first time a customer ordered. Menus got rebuilt too, with a chef-led keto lineup and options aimed specifically at the fitness, weight-loss, and GLP-1 audience the brand was actually attracting, updated seasonally alongside the marketing calendar. Cold-traffic landing pages got rebuilt continuously: session recordings and AI tools kept auditing the copy and layout, and somewhere in the range of 500 to 600 page versions got tested before the winning combination landed. Cold conversion settled in around 30%, which is what let the brand generate real revenue from cold acquisition again, not just from its own list.
The brand didn't get back to $700k a week. It got back to profitable, which is what actually kept the lights on.
A generous referral program got people through the door in 2020. A lifecycle system is what finally made them worth keeping.
Not everything worked
The recovery wasn't fast, and it wasn't a full comeback.
Cutting ad spend to zero while the lifecycle system got built meant weekly revenue kept falling before it turned around. The referral program's original terms had to get dialed back too, a trade-off that risked losing some of what made the brand's earliest fans stick around in the first place, but the loyalty held anyway. Six months in, weekly revenue sat at $300k. That's a real business again, and it's still a fraction of the $700k peak the brand once had.
Client name withheld and identifying details generalized, per the engagement's NDA. The dates and numbers are real.