Public articles with ≥100 clicks; bubbles represent missed sessions.
Monthly snapshots only; Public article thumbs are customer-only. Verify source filters for customer-only clicks.
Public articles, clicks ≥100 · check source filter for customer-only
Public articles with ≥100 clicks
| # | Action list |
|---|
Test what a higher containment rate could mean in avoided contacts, time and cost. This is a planning estimate—not a forecast.
Selected period
See which knowledge-base topics create the most demand, missed sessions and improvement opportunity.
Lower is better. Topics are ranked by missed sessions in the selected period.
Monthly helped rate for the selected topic.
Latest month compared with the previous available month.
What moved, why it moved, and whether the change indicates better content performance.
Prioritised by impact, persistence and direction—not by movement alone.
September–August directional outlook.
Forecast date: 31 August
Actual months continue into a directional forecast.
The selected metric shows actual values through the latest month and estimates thereafter.
Portfolio sentiment, signal strength and articles worth investigating.
Positive reactions and reaction volume by month.
Low adjusted score with meaningful customer evidence
High adjusted score with meaningful customer evidence
Adjusted scores pull very small samples toward the portfolio average. They help prioritise investigation, but do not prove that an article caused the customer reaction.
A practical reference for the definitions and logic behind classifications, comparisons, forecasts and customer-reaction signals.
The shared language used throughout the dashboard.
A customer journey that starts when a customer opens a Veta article.
A session where the customer’s need is resolved through the article without moving to another support channel.
Contributes to SSC.
A session is helped when the customer does not seek additional support after reading a Veta article.
Examples
Outcome: counted as a Helped Session.
A session is not helped when the customer continues to seek support after reading a Veta article.
Examples
Outcome: counted as Missed Containment.
The page or source the customer came from before opening the article.
Entry points help explain traffic sources and customer intent.
How articles enter the matrix and receive a category.
The article must be marked Public, have at least 100 clicks for its language variant, and have matching session data.
High sessionsArticles at or above the current 75th-percentile session threshold. This represents roughly the highest-traffic quarter of eligible articles.
Low containmentContainment at or below the current portfolio median for the selected period.
RecalculationThresholds are recalculated when the selected period changes. An article can therefore change category even when its own result is unchanged.
| Category | Sessions | Containment | Suggested use |
|---|---|---|---|
| Prioritize | High | Low | Review first; this is normally the largest immediate opportunity. |
| Protect | High | High | Important, high-volume content that is currently performing well. |
| Optimize | Lower | Low | Improve when capacity allows. |
| No action | Lower | High | No immediate action; continue monitoring. |
Classification is a prioritisation aid, not an automatic decision. The priority queue itself is ordered by missed sessions.
Diagnosis: The selected end month is compared with the previous imported month. Changes in missed sessions are separated into traffic, containment-performance and portfolio-composition effects.
Weakness patterns: New Priority articles, persistent low containment and multi-month deterioration are shown separately so stable problems are not hidden by month-to-month rankings.
Action lists: Items are ranked using impact, persistence and direction. Improvements remain provisional until the signal continues.
Portfolio score: Positive reactions are weighted by reaction volume, so articles with more responses contribute more to the overall result.
Adjusted ranking: Very small samples are pulled toward the portfolio average using a six-reaction reference. This prevents one response from dominating the ranking.
Evidence: The default list threshold is four reactions. Rankings are investigation signals, not proof that the article caused the response.
A directional outlook from 1 September to 31 August.
The model uses up to six recent months with available KPI values, including recent months before the current FY when useful.
Recent months receive more weight than older months when establishing the monthly run rate.
When at least three observations exist, a deliberately damped trend is added so one unusual month does not create an extreme projection.
Actual months remain unchanged. Missing FY months are estimated through 31 August and refresh after every import.
A category is assigned to the article ID and is therefore shared by Norwegian and English variants. Ignore excludes the article from all Topic Health calculations while keeping it available in the category list.
Sessions, contained sessions and missed sessions are added across the topic. Helped rate is contained sessions divided by sessions. In Article categories, the trend chart repeats this calculation month by month for the selected filter.
Needs attention combines below-portfolio helped rate with above-median missed volume. Deteriorating marks a decline of at least three percentage points.
Controls article aggregation, classification and customer-reaction views.
Overall sessions, helped sessions, savings and contribution values come from monthly inputs under Collect data.
Imported data, report links and FY goals are stored in this browser using IndexedDB and are included in JSON backups.
Share of article sessions that returned to the web journey.
What is currently available in the browser database.
Customer-service page traffic from the GA4 export for the selected period.
Sorted by sessions, highest first.
Report links and reference notes, gathered in one place.
Choose a working month before importing data.
Not calculated from the available exports.
Data is stored in IndexedDB, which supports substantially larger datasets than localStorage. Keep a downloaded backup.
One file includes every imported month, KPI input, saved report, topic category and article assignment. Restore replaces current browser data after confirmation.
Permanently removes every imported month, KPI value, saved report and other stored dashboard data from this browser.
Choose which optional tools appear in your navigation.
Turn on the tools you want to test. You can change this at any time.
Classification is a simple guide to where attention may be most useful.
High sessions + low containment. Review these first.
High sessions + high containment. Important content that is working well.
Lower sessions + low containment. Improve when capacity allows.
Lower sessions + high containment. Keep monitoring.
Use the classification as a starting point, not an automatic decision.
See Logic explained for the detailed calculation.