๐ Originally published on InsightRaider, a data platform tracking digital product revenue across Gumroad, Systeme.io and Whop.
Four models, ยฑ15-25% accuracy, zero public revenue data to start from. Gumroad, Systeme.io, and Whop never publish what a product earns. So how can anyone estimate it? This is the honest answer: no black box, no hand-waving, just the signals we collect, the models we run, and the limits we accept.
The benchmark we copied: BrandSearch
Before building InsightRaider, we studied how others estimate revenue for private businesses. The gold standard is BrandSearch, a company valued at $110M that estimates revenue for Shopify stores and Amazon sellers. They proved you can hit +/-20% accuracy by combining public signals with proprietary algorithms. We adapted that playbook for infoproducts.
Pillar 1: scraping the public signals
Digital product platforms expose more than you would expect. We systematically collect ranking data (bestseller position, overall platform rank, trending flags), social proof (review count, average rating, rating distribution, review velocity), and product metadata (price, launch date, creator follower count, products per creator).
Why it matters: public signals correlate strongly with revenue.
- Products in the top 10 of a category consistently outperform products ranked 11-50.
- Review counts correlate with sales at roughly 2-5% (100 reviews implies 2,000 to 5,000 customers).
- Rating scores above 4.5 correlate with 40% higher conversion rates.
These are patterns extracted from the 480,000+ products we track, not opinions.
Pillar 2: web traffic analysis
Revenue estimation needs to know how many people see a product page. We analyze traffic volume (estimated monthly visitors, trends, sources) and engagement (time on page, bounce indicators, return visitors), sourced from third-party analytics APIs (similar to SimilarWeb), backlink tools, social APIs, and search ranking data.
The conversion formula, at its simplest:
Estimated Revenue = Traffic x Conversion Rate x Average Order Value
For digital products, conversion rates typically range 1-4% depending on traffic temperature:
- Cold traffic: 1-2%
- Warm traffic (email, returning): 3-5%
- Hot traffic (referrals, affiliates): 5-10%
We calibrate the conversion assumption based on the traffic source mix.
Pillar 3: four models, then triangulate
We do not trust a single method. We run four and cross-validate.
Model What it uses A. Ranking-based Correlation maps between rank position and revenue across thousands of products B. Review-based Review-to-customer ratio (2-5%) to back out customers, times price, adjusted for refunds C. Traffic-based Traffic x conversion x price, refined by category, price point, and source mix D. ML ensemble Trained on products where creators publicly disclosed revenue (podcasts, tweets, Indie Hackers)The final estimate weights the four models by data availability and confidence. Strong traffic data pushes weight to Model C. Reliable review counts let Model B dominate.
How we know it works
We validate continuously against known data points. Creators like Pieter Levels, Tony Dinh, and Marc Louvion share revenue publicly, so we compare our estimates to their disclosures:
- Estimates for known products land within +/-15-25% accuracy.
- We update models when discrepancies appear and track accuracy over time to catch drift.
- When creators verify estimates for their own products, that anonymized feedback feeds back into calibration.
What the estimates can and cannot tell you
Honesty matters more than a clean number here.
They tell you: order of magnitude (a $1k/month vs a $10k/month product), relative comparison (Product A likely beats Product B), trend direction, and market sizing for a niche. For the full picture, see our digital product market size report.
They do not tell you: exact dollar amounts (we aim for +/-20%, not +/-1%), net profit, bundle or upsell revenue, or currency swings.
Accuracy drops for products with very few reviews, very new products, unusual pricing (bundles, pay-what-you-want), and during heavy promotional spikes. When confidence is low, we show it.
Edge cases, handled openly
- No reviews: lean on ranking and traffic data, flag lower confidence.
- Pay-what-you-want: use average transaction values from similar products, flag higher uncertainty.
- Multi-platform products: estimated per platform today, cross-platform aggregation is on the roadmap.
- Bundles and upsells: estimated on the primary product price, so figures may run conservative for sophisticated funnels.
The bottom line
No estimate is perfect. But ours are accurate enough to answer the questions that actually move a decision: is there money in this niche, what is the revenue ceiling for top products, is the market growing, and how does my product compare? For that last one, see our competitor analysis framework.
That is the information you need to choose where to invest your time. Without it, you are guessing. And guessing is how 95% of creators fail.
This analysis comes from InsightRaider, where you can enter any niche and get estimated monthly revenue, trend data, and competitor benchmarks for the top products in seconds.
๋ต๊ธ ๋จ๊ธฐ๊ธฐ