How we compute the Net App Satisfaction Score (NASS)

What NASS measures

NASS is a net score of App Store ratings: the share of 5-star ratings minus the share of 1-to-3-star ratings, on a scale from −100 to +100. It describes the users who rate an app, who are not a representative sample of all its users. It is calculated per company and per month from public App Store counts only: the number of ratings at each star level shown on every app listing, and the written reviews users publish. No language model, no sentiment classifier and no estimate enter the calculation.

NASS = ½ × score(ratings without a review) + ½ × score(written reviews), where score = % of 5-star − % of 1-to-3-star

Example: in a month, a company receives 80,000 ratings without a review (70% 5-star, 10% 4-star, 20% 1-to-3-star) and 2,000 written reviews (45% 5-star, 5% 4-star, 50% 1-to-3-star). Ratings score: 70 − 20 = +50. Reviews score: 45 − 50 = −5. NASS = ½ × 50 + ½ × (−5) = +22.5.

The scale

Each rating counts once, according to its stars.

The idea is close to the Net Promoter Score, but NASS does not ask whether users would recommend an app: it reads the ratings they actually left. Counting 3 stars as neutral instead of dissatisfied leaves the company ranking virtually unchanged (rank correlation 0.998 in August 2026).

Two channels, weighted equally

Ratings reach the App Store through two channels with opposite biases. Most ratings are left without text, often when the app itself asks for one, at a moment the developer chooses: they lean positive. A written review takes effort and is more often written by users who have a problem: it leans negative. NASS scores each channel separately and gives each half of the weight, whatever its volume. This convention limits the weight of solicited ratings; it does not cancel the biases of either channel. Both sub-scores are always published next to NASS, so anyone can recompute it with another weighting.

In August 2026, the 163 companies meeting the eligibility rules averaged +72.9 on ratings without a review and +25.6 on written reviews.

Ratings alone produce a ranking of their own, and that is the channel the developer controls. As soon as written reviews weigh at least 25%, the choice of weight barely changes the ranking. Equal weights sit in the middle of that range.

Data and period

Rating counts published on each App Store listing and written reviews, recorded every day by MWM Intelligence.

• Flow, not stock. NASS uses the ratings received during the month: the difference between the counts at the end of the month and at the end of the previous month, storefront by storefront. The cumulative total, which reflects years of history, is never used.
• Ratings without a review are the ratings of the month minus the written reviews of the month, star by star, so no rating is counted twice.
• A company is its operating perimeter: all its App Store accounts combined. A subsidiary run as an autonomous business stays on its own line.
• Version 1 covers iOS apps. Games are excluded. Two zones are published: worldwide and the United States.
• The monthly series starts in January 2025.

Precision and eligibility

Every NASS comes with a 90% confidence interval, computed from the exact variance of a difference between two shares of the same population. The interval measures statistical precision only. It does not cover who chooses to rate, rating requests made by apps, reviews missing from the collection or differences between countries.

SE(channel) = √((p5 + p1-3 − (p5 − p1-3)²) / n) × 100 · SE(NASS) = ½ × √(SE(ratings)² + SE(reviews)²) · 90% interval = ±1.645 × SE(NASS)

A company is ranked for a month only if:

• it received at least 5,000 ratings without a review;
• it received at least 120 written reviews: with fewer, a channel in which every review carries the same rating would show no margin of error at all;
• at least 90% of its ratings come from apps covered by our collection of written reviews;
• its 90% interval is no wider than ±2 points.

A change from one month to the next is reported as a move only when it exceeds the combined interval of the two months. The combined interval is √(m₁² + m₂²), where m₁ and m₂ are the two 90% margins; the same rule decides whether two apps or two companies can be told apart in the same month.

Exclusions and data quality

• Rating resets. When a developer resets an app’s ratings with a new version, the storefront count drops. That day is excluded for that app and storefront; the ratings received before and after still count.
• First observation. The first day an app is seen in a storefront has no previous count, so it has no flow. It is excluded.
• Source re-emissions. A day whose worldwide flow exceeds 10 times the median of the previous 28 days, with no missing day in the previous week, is a mass re-emission of counts, not ratings received that day. It is excluded. This happened once in 20 months, on 26 October 2025.
• Missing days. A day without data carries its changes over to the next available day. Each month and each week records how many days lacked usable data.

Reproducibility. Anyone who recomputes NASS from another source of App Store rating counts should find the same value, rounding aside. Because NASS uses the monthly flow, it does not depend on how a source handles rating resets: computed on all-time counts or on counts since the last reset, the monthly flow is the same outside reset days.

Limits

mwm.ai is operated by MWM, whose apps are included in NASS. The same rules apply to every company, MWM included. The weighting was set on methodological grounds; from January to August 2026, no weight between 25% and 100% on reviews moved MWM by more than eight places.

This is version 1 of the method (September 2026). Any change creates a new version, published with its date, and the history is recomputed.