What 15.8 million in-app interactions across 464 SaaS products reveal about product tours, NPS surveys, announcements and smart tips. Aggregate, anonymized data from the Produktly platform, January to June 2026.
Last updated 2026-08-13 · n = 464 companies · Methodology
Across 88 SaaS companies that each had at least 100 users start a product tour between January and June 2026, the median completion rate was 29%. The 25th percentile was 15%, the 75th percentile 55%. Completion is user-based: of the distinct users who started a tour, the share who reached its final step.
| Percentile | Completion rate (per company) |
|---|---|
| p10 | 4% |
| p25 | 15% |
| Median | 29% |
| p75 | 55% |
| p90 | 71% |
n = 88 companies, Jan-Jun 2026.
Per individual tour the picture is similar: across 256 tours with at least 50 started users, the median tour completed at 26% (p25 8%, p75 58%).
Takeaway
If your tours complete above 55%, you are in the top quarter of this sample. Below 15%, the bottom quarter. A "good" rate depends heavily on tour length and how the tour starts, which is why the next two sections exist.
Shorter tours complete at far higher rates. Across 256 tours with at least 50 started users (Jan-Jun 2026), median completion by step count:
| Steps in tour | Median completion | Tours | Companies |
|---|---|---|---|
| 1-2 | 73% | 45 | 21 |
| 3-5 | 38% | 65 | 40 |
| 6-8 | 25% | 84 | 40 |
| 9+ | 8% | 62 | 35 |
One honest caveat: step counts reflect each tour's current shape. Tours edited during the window are counted at their current length.
Takeaway
This is a correlation: teams that build 12-step tours may differ from teams that build 2-step tours in other ways too. But the pattern is steep and consistent across buckets, and it matches what most onboarding teams learn the hard way: every extra step is a place to leave.
Tours that users open themselves complete at a median of 69% (34 tours across 21 companies where at least 90% of starts were user-initiated). Tours that start automatically complete at a median of 23% (182 tours across 72 companies). Tours with mixed start types sit in between at 32% (40 tours).
| Start type | Median completion | Tours | Companies |
|---|---|---|---|
| User-initiated (90%+ manual starts) | 69% | 34 | 21 |
| Mixed | 32% | 40 | 24 |
| Auto-started (90%+ auto starts) | 23% | 182 | 72 |
Read this carefully: it does not mean switching a tour to manual start will triple completion. Users who choose to open a tour have already told you they are interested. Auto-started tours reach everyone, including users who never asked for a tour, so their completion rate is diluted by design.
Takeaway
If you auto-start your tour, benchmark yourself against 23%, not 69%. And if reach matters more than completion rate, auto-start is still how you get in front of every new user.
Based on 3,248 responses collected on standard 0-10 in-app NPS surveys between January and June 2026, the median SaaS product scored +30 (p25 +11, p75 +44; range -12 to +64).
| Percentile | NPS (per product) |
|---|---|
| p25 | +11 |
| Median | +30 |
| p75 | +44 |
3,248 responses across 13 products with 20+ responses each, 0-10 scale surveys only, Jan-Jun 2026.
Pooled across all 3,248 responses: 56% promoters, 13% passives, 31% detractors.
Takeaway
Treat +30 as a reference point rather than a target. In-app NPS is a specific instrument: it surveys users while they are inside the product, which samples a different population than email NPS campaigns, and this sample skews toward small and mid-size SaaS.
Across 122,595 in-app NPS survey impressions that produced 5,757 responses (January to June 2026), the median company converted 4.3% of impressions into a response. The 25th percentile company converted 2.2%, the 75th percentile 7.7%.
| Percentile | Responses per impression |
|---|---|
| p25 | 2.2% |
| Median | 4.3% |
| p75 | 7.7% |
Sample: 19 companies with at least 500 impressions and 5 responses each.
Impressions count repeat displays to the same user, so per-user response rates run higher than per-impression rates.
A useful companion number: at the median company, 11% of NPS responses came with a written comment.
Announcements front-load hard. The median announcement collected 43% of its total impressions within 48 hours of first being shown (p25 9%, p75 85%; n = 159 announcements across 44 companies, created January to May 2026, minimum 100 impressions).
| Measure | Share of impressions in first 48h |
|---|---|
| Median announcement | 43% |
| p25 announcement | 9% |
| p75 announcement | 85% |
| Pooled (impression-weighted) | 17% |
n = 159 announcements across 44 companies, created Jan-May 2026, minimum 100 impressions.
Weighted by impressions the picture shifts: across all impressions in the sample, 17% happened in the first 48 hours, because a minority of long-running, high-traffic announcements accumulate most of the volume. Both numbers are true; the median describes the typical announcement, the pooled number describes the typical impression.
Takeaway
An announcement does most of its work in its first two days. If the message matters, make those days count, and take stale announcements down instead of letting them decay in place.
Changelog reading is a slow burn compared to announcements. Looking at reads a changelog item collects in its first two weeks: the median item got 20% of those reads within 48 hours of publish (p25 8%, p75 30%) and 54% within the first 7 days.
| Window after publish | Median share of first-two-week reads |
|---|---|
| First 48 hours | 20% |
| First 7 days | 54% |
n = 210 changelog items across 35 companies, published Jan 1 to Jun 14, 2026, minimum 30 readers each.
Reads here are measured for each item's first 14 days, deduplicated per user, so this describes early-life reading only.
Takeaway
The pairing with the announcement numbers above tells a clean story: announcements are push and spike immediately; changelogs are pull and get discovered gradually as users show up. They are complements, not substitutes.
Rarely, and that is fine. Across 354 smart tips with at least 1,000 impressions (57 companies, Jan-Jun 2026), the median tip was opened once per 1,000 impressions (0.1%). The 75th percentile tip reached 1%, the 90th percentile 7.3%. An impression is a page view where the tip was present and eligible; an open is a user actually opening it.
| Percentile | Opens per impression |
|---|---|
| Median | 0.1% (1 in 1,000) |
| p75 | 1.0% |
| p90 | 7.3% |
n = 354 smart tips across 57 companies, minimum 1,000 impressions each, Jan-Jun 2026.
Takeaway
Tooltips are ambient reference, not a delivery channel. Users open them at the moment they are confused about that specific control, which for any given user on any given page view is rare. Put content users must see into a tour or an announcement; keep smart tips for the "what does this field mean" moments where they quietly earn their keep.
Anonymized, aggregated usage data from the Produktly platform, January 1 to June 30, 2026. In that window, 464 companies produced tracked widget events, totaling 15.8 million tracked in-app interactions (step-by-step tour progress events excluded from that count). Produktly's own account is excluded from every number.
A tour start is a user seeing the first step of a tour, whether it opened automatically or the user opened it. A tour completion is a user reaching the end of the tour. Completion rates are user-based: distinct users who finished divided by distinct users who started, the same definition the Produktly dashboard uses. An impression is one display of a widget to a user; repeat displays count as separate impressions.
Per metric: tour benchmarks include companies with at least 100 distinct users starting tours, and individual tours with at least 50 started users. NPS scores include companies with at least 20 responses on a 0-10 scale widget; response rates include companies with at least 500 impressions and 5 responses. Announcements: created inside the window, at least 100 impressions. Smart tips: at least 1,000 impressions. Changelog items: at least 30 readers. No bucket smaller than 8 companies is published.
Medians and quartiles across companies or widgets, never pooled averages, because usage data is heavy-tailed and a single large product would otherwise dominate. Where a pooled number is shown, it is labeled pooled.
Produktly's customer base skews toward small and mid-size SaaS. Enterprise deployments are underrepresented. These numbers describe what real products in that population measure, not universal truths.
No individual customer or end user is identifiable in any figure on this page.
Across 88 SaaS companies on the Produktly platform (Jan-Jun 2026), the median product tour completion rate was 29% of users who started a tour, with the middle half of companies between 15% and 55%. We report medians because averages are distorted by outliers in heavy-tailed usage data.
Shorter tours complete at higher rates: tours with 1-2 steps completed at a median of 73%, 3-5 steps at 38%, 6-8 steps at 25%, and tours with 9 or more steps at 8% (256 tours, Jan-Jun 2026). This is a correlation, but it is steep and consistent.
Based on 3,248 responses collected on in-app 0-10 NPS surveys (Jan-Jun 2026), the median SaaS product scored +30, with the middle half of products between +11 and +44.
At the median company, 4.3% of in-app NPS survey impressions produced a response, measured across 122,595 impressions (Jan-Jun 2026). The middle half of companies fell between 2.2% and 7.7%.
Yes, but quickly: the median in-app announcement collected 43% of its total impressions within 48 hours of first being shown (159 announcements, Jan-Jun 2026).
Rarely: the median smart tip was opened once per 1,000 impressions (354 tips, Jan-Jun 2026). The top decile of tips reached 7.3%.
These benchmarks come from Produktly, the all-in-one onboarding and feedback platform: product tours, checklists, NPS, surveys, announcements, changelogs and roadmaps in one tool. See how your numbers compare.
Start a 14-day free trial