Produktly’s 464-product dataset puts median SaaS tour completion at 29%

Produktly updated its 2026 SaaS Onboarding & In-App Engagement Benchmarks on August 20 with aggregate data from January through June: 464 companies and 15.8 million tracked in-app interactions. The report supplies percentile-based reference ranges across several product-adoption mechanics rather than relying on pooled averages. Among companies with at least 100 tour starters, median tour completion was 29% with a 15%–55% interquartile range. Across 256 individual tours, completion fell from a 73% median for 1–2 steps to 8% for 9+ steps. User-initiated tours completed at 69% versus 23% for auto-started tours, but Produktly explicitly warns that this is selection, not evidence that switching a tour to manual start will triple completion. The report also measures in-app NPS, announcement timing, changelog reading and tooltip opens.

Small and mid-size SaaS teams often make onboarding decisions against vendor anecdotes or generic conversion benchmarks that do not match the mechanism being measured. This dataset gives them a better starting point for asking whether a tour, NPS survey or in-app message is behaving unusually. Its more important lesson is methodological: benchmark against the right cohort and denominator. Auto-started and user-initiated tours serve different populations; NPS response rate per impression is not per user; announcements and changelogs have different attention curves. Builders can use these ranges to prioritize experiments and size samples, but should still tie onboarding changes to downstream activation, retention and revenue because the report measures engagement with widgets rather than business outcomes.

Across 88 SaaS companies with at least 100 distinct users starting a tour, median completion was 29%. The middle half of companies ranged from 15% to 55%, and the 10th-to-90th percentile spread ran from 4% to 71%. That wide distribution is useful in itself: a single industry-wide target would hide major differences in product, audience, tour design and trigger behavior.

Across 256 tours with at least 50 starters, median completion was 73% for 1–2-step tours, 38% for 3–5 steps, 25% for 6–8 steps and 8% for 9+ steps. Produktly correctly describes this as correlation rather than proof that deleting steps causes completion to rise; teams that create long tours may differ in product complexity and audience. The pattern is still strong enough to make tour length an obvious variable to test.

Tours where at least 90% of starts were user-initiated completed at a 69% median, while tours where at least 90% started automatically completed at 23%. Mixed-start tours were 32%. The report explicitly warns against treating the gap as causal: people who choose to open a tour have pre-selected for interest, while an automatic tour reaches users who never asked for it. Builders should therefore benchmark like against like rather than using the 69% number as a target for every onboarding flow.

Across 13 products with at least 20 standard 0–10 responses, median in-app NPS was +30, but the response-rate dataset is more operationally useful for many small SaaS teams. Across 122,595 NPS impressions, the median company converted 4.3% of impressions into a response; the middle half ranged from 2.2% to 7.7%. Only 11% of responses at the median company included a written comment. Teams using NPS for segment decisions therefore need to plan for slow sample accumulation and selection effects.

The median announcement received 43% of its total impressions within 48 hours, based on 159 announcements across 44 companies. By contrast, the median changelog item received 20% of its first-two-week reads in the first 48 hours and 54% in the first week. That supports treating announcements as push distribution and changelogs as a slower pull surface rather than substituting one for the other.

Across 354 smart tips with at least 1,000 impressions, the median tip was opened on 0.1% of impressions — about one in 1,000. The 90th percentile reached 7.3%. That makes tooltip opens a poor default channel for information every user must see, but a potentially appropriate low-friction mechanism for contextual help that only a small fraction of users need at a given moment.

Produktly excludes its own account, reports medians and quartiles to reduce heavy-tail distortion, publishes minimum sample thresholds and refuses to show buckets with fewer than eight companies. It also discloses the main limitation: its customer base skews toward small and mid-size SaaS, with enterprise underrepresented. More importantly, the dataset measures interaction with onboarding widgets, not whether those interactions caused activation, retention or expansion.