Calorie tracker loses €5,800 despite €22,000 gross revenue and 13,000 installs

A German solo developer's unit economics reveal how hidden API costs, platform fees, and misaligned ad attribution can turn a highly rated AI wrapper into a loss-making venture.

A German solo developer operating under the Reddit handle SuspiciousSoftware74 generated approximately €22,000 in gross revenue and achieved 13,300 installs for an AI-powered calorie tracking app. On paper, the business looks healthy, boasting a 4.7-star rating on Google Play and 678 active subscriptions yielding approximately $2,200 in monthly recurring revenue. Yet, after accounting for platform fees, ad spend, and API costs, the founder is €5,800 in the red.

The case exposes the structural fragility of consumer AI wrappers. While low-code tools and foundation APIs make it possible for a solo developer to build and launch quickly, the unit economics of customer acquisition and API-dependent features frequently collapse under scale.

The ad attribution illusion

The founder spent €12,870 on paid acquisition, split between Google Ads (€7,514) and Apple Search Ads (€5,356). The performance of these channels differed sharply in attribution accuracy and return. Apple Search Ads never achieved profitability. In June, a €2,000 spend on Apple's platform returned only 40% of its cost.

Google Ads presented a different trap. The Google campaign dashboard reported an acquisition cost of roughly €1 per user. However, this campaign optimized for app opens rather than purchases. When the founder cross-referenced Google click IDs with RevenueCat transaction logs, the actual cost to acquire a paying subscriber was €22. The campaign was optimizing for app opens, not for people who pay. The 22x discrepancy highlights the risk of relying on ad network optimization metrics without deep integration into down-funnel subscription events.

The compounding cost of grounding

The app uses photo recognition to estimate calories and macronutrients, relying on Google Cloud's Vertex AI and Gemini models. As user volume grew, Google Cloud costs rose to between €800 and €1,200 per month. Over 80% of this bill stems from model calls and Google Search grounding.

Grounding, which queries live search data to improve the accuracy of nutritional estimates, introduces a high variable cost for every user action. Because the app offered a free tier, non-paying users actively drained resources. Every logged meal incurred API costs, creating a direct link between unpaid user engagement and increasing operational losses.

A pricing pivot to save margins

To combat rising API expenses, the developer adjusted pricing in August. After testing price points between €2.99 and €5.99, the founder raised rates to €7.99 per month and €39.99 per year.

The change altered user behavior. While the yearly plan maintained steady conversion rates, the monthly plan experienced a near-total drop in sign-ups. Despite the death of the monthly option, overall revenue per paywall view increased. This shift to upfront annual cash flow helped offset immediate costs, though it did not fully resolve the historical acquisition deficit.

Introduce hard usage caps

Operating an un-capped free tier on top of a paid API is a structural error. The developer should implement a strict daily limit on free image scans, such as three per day, or restrict search grounding entirely to premium tiers. Grounding should only run when local or cheaper model heuristics fail.

Rebuild the attribution loop

Running Google Ads campaigns optimized for app opens is highly inefficient for subscription apps. The developer must pass RevenueCat purchase events back to Google Ads and Meta. Optimizing for trial starts or completed purchases would raise the upfront CPA on paper, but it would align ad spend with actual revenue generation.

Transition to local models

Using Gemini with search grounding for every single food item is unnecessary. A hybrid architecture would route common foods (like an apple or a slice of bread) to a lightweight, local, or cheaper on-device model. Only complex, multi-ingredient meals should trigger expensive cloud-based visual search pipelines.

The developer has turned off all paid advertising. With monthly revenue after app store fees hovering between €1,400 and €1,800, and cloud infrastructure costs consuming €800 to €1,200, the business should technically operate at a slight profit without ad spend. The critical test is whether organic discovery can sustain the app, or if the loss of paid traffic will trigger a slow decline in active users. For solo developers, this case serves as a warning that high ratings and strong initial download numbers mean little if the underlying infrastructure and acquisition funnels are not built for margin.

The investor read

This case illustrates why early-stage consumer AI wrappers are increasingly uninvestable without proprietary data or distribution advantages. The combination of platform fees (15-30%), high API costs driven by search grounding, and rising CAC on Apple Search Ads and Google Ads creates a structural margin squeeze. For investors, the critical takeaway is that high app store ratings and initial download velocity are lagging indicators of health. A product that cannot optimize its ad campaigns for down-funnel purchase events is effectively buying unprofitable traffic. To become viable, such utilities must transition from variable API cost structures to hybrid local/cloud models and enforce strict usage caps on non-paying users.

Pull quote: “The campaign was optimizing for app opens, not for people who pay.”