First Touch Application Rate

Definition

The application rate for each first-touch source group — the channel or medium that produced a person’s earliest recorded engagement touch. Groups with a lift above 1.0 are converting above the overall baseline; below 1.0 means underperforming.

Summary

Organic search is the highest-converting tracked channel at 28.7% (2.96x lift), and the second largest group by volume. Direct and unknown traffic is the largest group and also converts well at 21.3% — together these two groups account for the majority of high-intent first touches and point to strong unaided demand.

Paid and Carnegie-sourced channels consistently convert below the overall baseline from first touch, which is consistent with their role as awareness and top-of-funnel channels. Paid social in particular converts at near-zero (1.3%) — it reaches a large number of people who are not yet ready to apply.

Source Group	People	Applicants	App Rate	Lift
organic search	25,767	7,389	28.7%	2.96x
direct / unknown	38,783	8,264	21.3%	2.20x
ai referral	164	34	20.7%	2.14x
paid search	1,316	221	16.8%	1.73x
email	1,267	210	16.6%	1.71x
paid display	1,259	179	14.2%	1.47x
organic social	1,036	144	13.9%	1.43x
other	253	32	12.6%	1.30x
vendor / purchased list	36	4	11.1%	1.15x
paid / carnegie	2,022	135	6.7%	0.69x
paid social	844	11	1.3%	0.13x
no engagement touch	103,525	535	0.5%	0.05x

Note

Overall baseline rate is 9.7% across FY2023–2026. Lift is each group’s rate divided by the overall rate.

Key Insight

Organic search is the highest-converting tracked channel and the clearest signal of high-intent self-directed demand. Paid and Carnegie-sourced channels convert well below baseline from first touch — they are top-of-funnel awareness channels and should not be evaluated on first-touch application rate alone. AI referral (164 persons, 20.7% rate) is an emerging channel worth monitoring.

Why This Is Important

First-touch attribution reveals which channels are attracting prospects who eventually apply, regardless of what happens in between. It answers a different question than last-touch or multi-touch models: who showed up first, and did they end up applying?

This matters for channel investment decisions:

• Channels with high first-touch lift (organic search, direct) attract self-directed, high-intent prospects — protecting or growing those audiences has compounding value
• Channels with low first-touch lift (paid social, paid display, Carnegie) may still play an important role in the funnel as awareness drivers, but should not be held to a first-touch conversion standard
• paid / carnegie reaches 2,022 people but converts at only 6.7% from first touch — below baseline, suggesting Carnegie is effective at top-of-funnel reach but requires substantial nurturing before conversion
• The no engagement touch group (103,525 people, 0.5% rate) is a large population with almost no conversion — this is primarily list imports and name buys with no subsequent digital engagement

Methodology

The query assigns each person their first engagement touch using ROW_NUMBER() ordered by touched_at. That touch’s utm_source and utm_medium are used to classify the person into a source_group via a CASE statement. Persons with no engagement touch in person_touches fall into the no engagement touch group.

Application rate is calculated across all persons, not just applicants, so the denominator includes everyone in each source group regardless of whether they applied. Lift is the ratio of a group’s application rate to the overall application rate (9.7% across FY2023–2026).

Note

The other bucket will shrink as UTM classification rules are refined. Any utm_source / utm_medium combination not matched by the CASE statement lands here. Auditing high-volume unmatched values in person_touches is the fastest way to improve this analysis.

Query

WITH first_touch AS (
  SELECT
    person_id,
    touched_at,
    touch_type,
    touch_category,
    utm_source,
    utm_medium,
    utm_campaign,

    ROW_NUMBER() OVER (
      PARTITION BY person_id
      ORDER BY touched_at ASC, touch_type ASC
    ) AS rn

  FROM person_touches
  WHERE touch_category = 'engagement'
),

source_groups AS (
  SELECT
    person_id,
    touched_at AS first_touch_at,
    touch_type AS first_touch_type,
    utm_source AS first_touch_utm_source,
    utm_medium AS first_touch_utm_medium,
    utm_campaign AS first_touch_utm_campaign,

    CASE
      WHEN lower(utm_source) IN ('google', 'bing', 'yahoo', 'yahoo!', 'duckduckgo', 'ecosia', 'yandex', '360.cn', 'baidu')
        AND lower(utm_medium) IN ('organic', 'organic_search', 'search')
        THEN 'organic search'

      WHEN lower(utm_source) IN ('google', 'bing')
        AND lower(utm_medium) IN ('cpc', 'paid', 'paid_search', 'ppc')
        THEN 'paid search'

      WHEN lower(utm_source) IN ('googlead', 'mediately')
        OR lower(utm_medium) IN ('mediately', 'display', 'lightbox', 'banner')
        THEN 'paid display'

      WHEN lower(utm_source) IN ('carnegie-digital', 'carnegie', 'carnegie_digital')
        AND lower(utm_medium) IN ('ppc', 'paid-social', 'paid_social', 'cpc', 'display', 'lightbox', 'discovery', 'custom-content', 'video')
        THEN 'paid / carnegie'

      WHEN lower(utm_source) IN ('facebook', 'instagram', 'linkedin', 'youtube', 'tiktok', 'snapchat', 'twitter', 'reddit', 'meta')
        AND lower(utm_medium) IN ('paid', 'paid-social', 'paid_social', 'cpc')
        THEN 'paid social'

      WHEN lower(utm_source) IN ('facebook', 'instagram', 'linkedin', 'youtube', 'tiktok', 'snapchat', 'twitter', 'reddit')
        OR lower(utm_medium) IN ('social')
        THEN 'organic social'

      WHEN lower(utm_source) IN ('slate', 'gmail', 'miami-matters', 'mass_mailer', 'inbound-email', 'notifier', 'email', 'newsletter')
        OR lower(utm_medium) IN ('email')
        THEN 'email'

      WHEN lower(utm_source) IN ('nrccua', 'encoura', 'collegeboard', 'cappex', 'niche', 'niche.com', 'qs', 'usnews_studentconnect', 'hotcoursesinternational', 'idp', 'ziprecruiter')
        THEN 'vendor / purchased list'

      WHEN lower(utm_source) IN ('chatgpt.com', 'perplexity', 'claude', 'gemini', 'copilot.com')
        THEN 'ai referral'

      WHEN lower(utm_source) IN ('miami university calendar')
        THEN 'calendar widget'

      WHEN lower(utm_medium) IN ('referral')
        THEN 'referral'

      WHEN utm_source IS NULL
        OR utm_source = ''
        OR utm_source = '(none)'
        THEN 'direct / unknown'

      ELSE 'other'
    END AS source_group

  FROM first_touch
  WHERE rn = 1
),

touch_counts AS (
  SELECT
    person_id,

    COUNT(*) AS total_touches,
    COUNT(CASE WHEN touch_category = 'engagement' THEN 1 END) AS engagement_touches,
    COUNT(CASE WHEN touch_category = 'exposure' THEN 1 END) AS exposure_touches,
    COUNT(CASE WHEN touch_type = 'ping' THEN 1 END) AS pageview_count,
    COUNT(CASE WHEN touch_type = 'form_response' THEN 1 END) AS form_response_count

  FROM person_touches
  GROUP BY person_id
),

applied AS (
  SELECT DISTINCT
    person_id
  FROM person_tags
  WHERE tag_type = 'applied'
    AND tag_value = 'Yes'
),

labeled AS (
  SELECT
    p.id AS person_id,

    COALESCE(sg.source_group, 'no engagement touch') AS source_group,
    sg.first_touch_at,
    sg.first_touch_type,
    sg.first_touch_utm_source,
    sg.first_touch_utm_medium,
    sg.first_touch_utm_campaign,

    COALESCE(tc.total_touches, 0) AS total_touches,
    COALESCE(tc.engagement_touches, 0) AS engagement_touches,
    COALESCE(tc.exposure_touches, 0) AS exposure_touches,
    COALESCE(tc.pageview_count, 0) AS pageview_count,
    COALESCE(tc.form_response_count, 0) AS form_response_count,

    CASE
      WHEN a.person_id IS NOT NULL THEN 1
      ELSE 0
    END AS applied

  FROM persons p

  LEFT JOIN source_groups sg
    ON p.id = sg.person_id

  LEFT JOIN touch_counts tc
    ON p.id = tc.person_id

  LEFT JOIN applied a
    ON p.id = a.person_id
),

overall AS (
  SELECT
    ROUND(
      SUM(applied) * 1.0 / COUNT(*),
      3
    ) AS overall_application_rate
  FROM labeled
)

SELECT
  l.source_group,

  COUNT(*) AS people,
  SUM(l.applied) AS applicants,

  ROUND(AVG(l.total_touches), 1) AS avg_total_touches,
  ROUND(MEDIAN(l.total_touches), 1) AS median_total_touches,

  ROUND(AVG(l.engagement_touches), 1) AS avg_engagement_touches,
  ROUND(MEDIAN(l.engagement_touches), 1) AS median_engagement_touches,

  ROUND(AVG(l.exposure_touches), 1) AS avg_exposure_touches,
  ROUND(AVG(l.pageview_count), 1) AS avg_pageviews,
  ROUND(AVG(l.form_response_count), 1) AS avg_form_responses,

  ROUND(
    SUM(l.applied) * 1.0 / COUNT(*),
    3
  ) AS application_rate,

  o.overall_application_rate,

  ROUND(
    (
      SUM(l.applied) * 1.0 / COUNT(*)
    )
    /
    NULLIF(o.overall_application_rate, 0),
    2
  ) AS application_rate_lift

FROM labeled l
CROSS JOIN overall o

GROUP BY
  l.source_group,
  o.overall_application_rate

ORDER BY
  application_rate DESC;