Our Numbers, Honestly: Why You Won't See "95% Accuracy" on Our Website

If you've shopped around for a subscription tracker, you've seen the claims: "saves users $512 a year", "99% detection accuracy", walls of five-star testimonials. Maybe some of those numbers are real. The problem is you have no way to know — none of them come with a methodology, a sample size, or a date.

WinnowFi is a new product, built by one person. We don't have thousands of users yet. So here is our public commitment about numbers, in writing, where you can hold us to it.

What we won't publish (yet)

We won't publish a detection accuracy rate. Our AI scores every bank transaction against known merchants, recurring patterns, billing categories, and your own confirmation history. We believe it works well. But "believe" is not a percentage. Our system already measures precision in production — how many detections users confirm versus dismiss — and until that metric is stable across real users over a meaningful period, you won't see an accuracy number on our site.

We won't publish "average savings per user". An average requires users, and enough of them for the number to mean something. What you'll see instead is plain arithmetic you can check yourself: one forgotten $20/month subscription is $240+ a year. That's not a statistic — it's multiplication.

We won't publish testimonials we didn't receive. When you see a quote with a name on our site, it will be from a real person who used the product and agreed to be quoted. Until then, we show you what the product actually catches, not what imaginary people say about it.

What we publish today

Numbers that are either live data or verifiable facts:

• Our donations. 20% of gross revenue goes to autism research — 12% to SPARK for Autism, 8% to the Autism Science Foundation. The Transparency page shows the running total, pulled live from our systems, with receipts published monthly.
• Our pricing. Free plan with up to 3 manual subscriptions, no credit card. Premium at $5.99/month or $59.99/year. No hidden tiers.
• Industry statistics, with sources. When we cite research — like C+R Research's finding that households underestimate their subscription spend by 68% — we name the source and the year, so you can check it.

What will change, and when

This isn't false modesty; it's a timeline. As real users confirm and dismiss detections, our precision metric accumulates data. Once it's stable over a meaningful window, we'll publish it — along with how it's calculated, over what period, and what counts as a hit or a miss. The same goes for savings: when we can compute what users actually cancelled through WinnowFi, we'll publish the real figure with the methodology next to it.

If the number turns out to be 87% instead of a marketing-friendly 95%, we'll publish 87% and keep working on it. A real number with a methodology beats a round number with neither.

Why we work this way

WinnowFi asks for something serious: read-only access to your bank transactions. That only works on trust, and trust isn't built with inflated numbers — it's built by being accurate about small things so you believe us on big ones. It's the same reason 20% of our revenue goes to autism research with public receipts instead of a vague "we give back" badge.

This post is the standard we're setting for ourselves. If you ever catch a number on our site that doesn't survive the question "can you prove it?" — email us at support@winnowfi.com and we'll fix it or pull it. That's the deal.