The idea
A friend pitched me a marketplace where supplement brands would compete on nutrition instead of vibes. It started as a side project.
Post-mortem · 2023 – 2026
A supplement marketplace that ranked products by what's on the label instead of the vibes in the ad. We catalogued 1,404 products and signed affiliate deals with 19 brands. Then the work outgrew the returns. This page records what we built and what I learned.
Ranked by cost per serving, protein
Criteria tap to add or remove
Scoring: grams of protein per dollar
This uses NutraStack's ranking formula and the label data from our database. Try adding Sugar. Prices are as we recorded them in 2024.
Timeline
A friend pitched me a marketplace where supplement brands would compete on nutrition instead of vibes. It started as a side project.
From here until November, NutraStack was close to a full-time job.
Sorting and filtering came first, then side-by-side compare, proprietary blends broken into their ingredients, and a landing page with a waitlist.
Every product got a Buy link that routed to the brand or to Amazon. Category rankings, search and sign-up followed, along with partner discount codes and referral links.
The largest catalog update we made, a new landing page with a partner carousel, and % Daily Value and category percentiles on every product page.
I made a new brand deck and pitched a few brands, but held off on going back to 10-hour days until something gave me a reason to expect a different result.
What we built
Every product was stored as its full Supplement or Nutrition Facts panel, one row per ingredient and dose. Everything else came from that data.
66 categories, from high-stim pre-workouts to frozen desserts. Sort by up to 10 criteria at once, such as creatine per dollar, and filter out what you don't want, like added sugar or artificial flavors.
Up to three products side by side, ingredient by ingredient, with blends expanded and flavors you could switch between.
The full label for every flavor, % Daily Value, and a percentile showing how each dose compared with the rest of the category.
1,880 tracked ingredients, each with an article explaining what it does and a ranking of the products that contain the most of it.
Shelved
For brands
The other side of the marketplace. Our brand deck promised that products would rank on relevance, not ad budget, and that we only earned when a brand made a sale.
We added the brand's products for them: images, nutrition labels, descriptions and prices.
Featured brands appeared in our social posts and the landing page carousel.
A flat 20% on referred sales, open to negotiation.
Monthly through PayPal, or Venmo, direct deposit or Shopify Collabs.
A referral URL with a 30 to 60 day cookie, plus a discount code such as NUTRASTACK10 where tracking wasn't possible.
Partners
19 brands signed affiliate agreements. Archaneon, Byoform, Live Evolved, Swollhouse, Toxic Masculinity Nutrition and Yumtein signed after the last site update and never went live. Products from every other brand linked to Amazon.
Sep 2024
The landing page led with a direct comparison, because the numbers were the whole argument.
Elevate your fitness with confidence
Looking for a healthy soda?
Compare Sodas
Traction
What went in catalog and partners
What came back Nov 2026
Catalog counts are from the code as of November 2024. The partner count includes six brands signed after that. Ad spend is totaled from 161 Meta and 14 TikTok receipts, August 2024 to March 2025. Follower counts are current.
Under the hood
| products | product_id · brand · name · product_type · description · price · flavors |
|---|---|
| product_flavors | parent_product_id · child_product_id · flavor |
| flavor_ingredients | product_id · ingredient · value |
| product_servings | product_id · servings · serving_size |
| compare_events | compare_id · time · session_id · product1 · product2 · product3 |
| outbound_clicks | click_id · time · session_id · page_url · product_id · destination |
Excerpt from our 2024 schema.
Why it stopped
NutraStack ran on data. Every flavor of every product was a full label, entered and checked by hand, and every new brand, flavor or reformulation meant more of it. Through most of 2024 that was more than ten hours a day.
The returns didn't follow. $5,689 of Instagram and TikTok ads grew the accounts to 1,191 and 1,652 followers, but followers didn't become shoppers. Sales through the site were minimal, and a 20% commission on minimal sales never came close to paying for the work.
In 2025 I pitched a few more brands, but I wasn't willing to go back to 10-hour days without a clear sign that the result would be different.
Lessons
We were already logging every outbound click. We should have spent $500 per channel, measured clicks and purchases, and stopped as soon as one didn't convert. Instead we bought reach and counted followers.
Brands said yes far more easily than shoppers did. Getting shoppers was the actual business, and affiliate revenue needs volume before it's worth anything. I'd pick one niche, like pre-workouts, and win search traffic for it before adding a 66th category.
A label database is never finished. Brands reformulate, add flavors and change prices, and each of 108 brands added to the work. I should have automated label capture early, or kept the catalog to the categories that actually sold.
Team
Designed and built the website, the ranking engine and the product data model. Mathematics-Computer Science, Statistics and Political Science.
In our words
“Rank on relevance, not budget.”
from our 2025 brand deck
Thanks to every brand that trusted us with its products, and to everyone who used the site to compare a label.