Trends by DSI
The Trends by DSI dashboard provides visibility into cohort performance over time by plotting KPIs against days since install (DSI). This view highlights how ROAS, LTV, monetisation, and retention evolve during the user lifecycle, enabling comparisons across different cohort groups and how they mature over time.
What is in this dashboard
4 tiles: ROAS, LTV, Retention, ARPDAU
X-axis: Days since install (each point = same cohort, one day later)
Y-axis: KPI shown in the tile title
Solid line = actuals; Dashed line = predictions
How it works
KPIs are tracked at every DSI within the selected DSI window
Actuals are displayed where available (solid lines); predictions fill gaps where data has not yet fully actualized (dashed lines)
Example: a cohort may have up to D30 actuals, but the DSI window selected was D60, so D30-60 metrics are predicted and display as dashed lines to distinguish where forecasts begin
The dashboard is most useful for:
Understanding monetization velocity i.e. how quickly a cohort returns ad spend or generates LTV
Identifying engagement and retention decay curves across different cohorts
Comparing how recent cohorts differ from historical ones in terms of ROAS slope, ARPDAU drop-off, or retention stability
Diagnosing changes in UA quality or monetization effectiveness over time
Available filters
Paid vs Blended (include / exclude organic uplift)
Net vs Gross (with / without platform fees / taxes)
Pivot dimensions:
Default = Time (each line = one cohort group)
If Time is pivoted, no second pivot can be added
If Time is not pivoted, up to 2 dimensions may be selected (Campaign, Network, Geo, Platform) in any combination
In this case, all cohorts within the selected date range are aggregated into per-segment lines (showing average DSI trend)
Install groupings: Cohorts may be grouped by daily, weekly, or monthly installs
Date range selector: Quick ranges include Last 7, 14, 28, 60 days, or 3 months
Segment filters: Views can be narrowed by Geo, Platform, Network, or Campaign, either individually or in any combination
DSI window selector: Controls the x-axis limit of the chart (to D1, D7, D30, D60, D90)
Tips and warnings
✅ Use shorter windows (D1 - D7) for acquisition quality checks; longer windows (D30 - D90) for monetization insights
✅ With Time pivot, compare cohorts directly; with other pivots, identify systematic differences across platforms, networks, or geos
⚠️ Aggregated segment lines can mask intra-cohort variability - use Time pivot for cohort-level diagnosis
⚠️ Later DSIs rely more on predictions; dashed lines should be treated as directional guidance, not precise outcomes
⚠️ Very recent cohorts (small or incomplete) may distort averages when heavily segmented