Cohort Analysis for Deep Links: Tracking User Groups

Aggregate metrics hide important trends. "10,000 deep link clicks this month" tells you nothing about whether the users who clicked are retaining, converting, or churning. Cohort analysis solves this by grouping users who share a common trait (typically their acquisition date or source) and tracking their behavior over time.

This guide covers cohort analysis for deep link performance. For deep link analytics broadly, see deep link analytics: measuring what matters. For conversion funnel analysis, see conversion funnel analysis for deep links.

The analytics dashboard with date range selector, filters, charts, and breakdowns.

What Is Cohort Analysis

A cohort is a group of users who share a characteristic. The most common cohort type is an acquisition cohort: all users who installed in the same week or month.

Acquisition Cohort Example

Users who clicked a deep link and installed in Week 1 of July vs Week 2:

Cohort	Installs	D1 Retention	D7 Retention	D30 Retention	Revenue/User
Jul W1	2,400	45%	28%	15%	$3.20
Jul W2	2,800	42%	25%	12%	$2.80
Jul W3	3,100	48%	32%	18%	$4.10

Week 3 had the highest quality users despite not having the most installs. Something changed (maybe a new campaign, creative, or targeting) that improved quality.

Source Cohort Example

Group by the deep link source instead of date:

Source	Users	D7 Retention	D30 Retention	Conversion Rate	Revenue/User
Email campaign	1,200	38%	20%	8.5%	$4.50
Push notification	3,500	30%	14%	5.2%	$2.80
Social share	800	42%	25%	10.1%	$5.20
QR code	400	35%	18%	7.0%	$3.60

Social share users have the highest retention and revenue, even though push notifications drive the most volume.

Retention Cohort Heatmap

The retention heatmap is the most common cohort visualization. Each row is a cohort (week), each column is a time period after acquisition, and the cell value is the retention rate:

          Week 0   Week 1   Week 2   Week 3   Week 4   Week 5
Jun W1    100%     35%      22%      18%      15%      13%
Jun W2    100%     38%      25%      20%      17%      15%
Jun W3    100%     33%      20%      16%      13%      11%
Jun W4    100%     40%      28%      23%      19%      17%
Jul W1    100%     42%      30%      24%      -        -
Jul W2    100%     36%      24%      -        -        -

Reading this heatmap:

• Jun W4 has the best retention curve. Something improved that week.
• Jun W3 has the worst retention. Investigate what changed.
• Jul W1 and W2 show continued improvement.

Building the Heatmap

SELECT
  DATE_TRUNC('week', u.install_date) AS cohort_week,
  FLOOR(EXTRACT(EPOCH FROM (e.event_date - u.install_date)) / 86400 / 7) AS weeks_since_install,
  COUNT(DISTINCT u.user_id) AS cohort_size,
  COUNT(DISTINCT e.user_id) AS active_users,
  ROUND(COUNT(DISTINCT e.user_id)::DECIMAL / COUNT(DISTINCT u.user_id) * 100, 1) AS retention_pct
FROM users u
LEFT JOIN events e ON u.user_id = e.user_id
  AND e.event_type = 'app_open'
WHERE u.install_date >= '2026-06-01'
GROUP BY cohort_week, weeks_since_install
ORDER BY cohort_week, weeks_since_install;

Deep Link Cohort Dimensions

By Deep Link Route

Group by the route the user first entered through:

First Deep Link Route	Users	D7 Active	Conversion
/products/{id}	3,200	32%	6.5%
/offers/{id}	1,800	28%	12.0%
/referral/{code}	900	45%	9.0%
/ (home)	5,000	22%	3.0%

Users who enter through a specific product or offer page have higher conversion than those who land on the home screen. Referral users have the highest retention.

By Campaign

SELECT
  dl.utm_campaign,
  COUNT(DISTINCT dl.user_id) AS users,
  ROUND(AVG(CASE WHEN e.d7_active THEN 1 ELSE 0 END) * 100, 1) AS d7_retention,
  ROUND(AVG(CASE WHEN e.d30_active THEN 1 ELSE 0 END) * 100, 1) AS d30_retention,
  ROUND(SUM(r.revenue) / COUNT(DISTINCT dl.user_id), 2) AS revenue_per_user
FROM deep_link_clicks dl
LEFT JOIN user_engagement e ON dl.user_id = e.user_id
LEFT JOIN revenue r ON dl.user_id = r.user_id
  AND r.event_date <= dl.click_date + INTERVAL '90 days'
WHERE dl.click_date >= '2026-06-01'
GROUP BY dl.utm_campaign
ORDER BY revenue_per_user DESC;

By Device/Platform

Platform	Users	D7 Retention	Conversion	Revenue/User
iOS	4,500	35%	8.0%	$4.80
Android	6,200	28%	5.5%	$3.20
Web (fallback)	1,800	15%	2.0%	$1.50

iOS users from deep links tend to retain and convert at higher rates. Web fallback users (who did not have the app) convert at much lower rates, highlighting the importance of app installation.

Revenue Cohort Analysis

Track cumulative revenue per user cohort over time:

Cumulative Revenue Per User (by install week):

          Month 0   Month 1   Month 2   Month 3   Month 6   Month 12
Jan W1    $0.50     $2.10     $3.40     $4.50     $7.20     $11.00
Feb W1    $0.60     $2.30     $3.80     $5.00     $8.10     -
Mar W1    $0.45     $1.90     $3.10     $4.20     -         -
Apr W1    $0.70     $2.50     $4.00     -         -         -

This view shows whether user quality is improving or declining over time. Apr W1 has the highest early revenue, suggesting improving acquisition quality.

Implementing Cohort Analysis

Data Model

interface CohortConfig {
  dimension: 'install_date' | 'source' | 'campaign' | 'route' | 'platform';
  granularity: 'day' | 'week' | 'month';
  metrics: ('retention' | 'revenue' | 'conversion' | 'engagement')[];
  dateRange: { start: string; end: string };
  periods: number; // How many periods to track (e.g., 12 weeks)
}

async function buildCohortReport(config: CohortConfig): Promise<CohortReport> {
  const cohorts = await getCohorts(config.dimension, config.granularity, config.dateRange);

  for (const cohort of cohorts) {
    for (let period = 0; period < config.periods; period++) {
      for (const metric of config.metrics) {
        cohort.periods[period][metric] = await calculateMetric(
          cohort.userIds,
          metric,
          cohort.startDate,
          period,
          config.granularity
        );
      }
    }
  }

  return { config, cohorts };
}

Visualization

function renderRetentionHeatmap(report: CohortReport): string[][] {
  const rows: string[][] = [];

  // Header row
  const header = ['Cohort', ...Array.from({ length: report.config.periods }, (_, i) =>
    `${report.config.granularity === 'week' ? 'W' : 'M'}${i}`)];
  rows.push(header);

  // Data rows
  for (const cohort of report.cohorts) {
    const row = [cohort.label];
    for (let p = 0; p < report.config.periods; p++) {
      const value = cohort.periods[p]?.retention;
      row.push(value !== undefined ? `${value.toFixed(1)}%` : '-');
    }
    rows.push(row);
  }

  return rows;
}

Actionable Insights from Cohorts

Identifying Quality Changes

If D7 retention drops from 35% to 25% between consecutive cohorts, investigate:

• Did targeting change?
• Was a new creative introduced?
• Did onboarding change?
• Was there a bug in a new app version?

Optimizing Channels

If email deep link cohorts consistently outperform push notification cohorts, shift budget toward email or investigate why push users have lower quality.

Predicting LTV

Early cohort behavior (D1, D3, D7 retention and revenue) predicts long-term value. Build models that use early cohort data to project D90 and D365 LTV.

Tolinku for Cohort Analytics

Tolinku's analytics support cohort analysis through click and conversion tracking by source, campaign, and route. Build cohort charts and funnels in the Tolinku dashboard.

For deep link analytics, see deep link analytics: measuring what matters. For app growth metrics, see app growth metrics: the 15 KPIs that matter.