Form Response Volume as a Predictor

Definition

Application rate by number of form responses submitted, controlling for total touch volume bucket. A form response is an active submission — a stronger signal of intent than a passive pageview. This page tests whether form volume predicts application independently of total engagement.

Summary

The most counterintuitive finding in this dataset: one form submission is below the overall baseline (12.9%, 0.76x lift), while zero form submissions is essentially at baseline (16.7%, 0.99x). The signal doesn’t emerge until two or more forms — where the rate jumps to 31.2% and climbs from there.

A single form submission appears to be an inquiry action, not an intent signal. It’s often the first touch in a funnel, not evidence of return engagement. Two or more submissions indicates a prospect came back — and that return is what predicts application behavior.

The cross-tab confirms this: within every touch bucket up to 50, having 1 form consistently underperforms having 0 forms. At high touch volumes (51+), the gap closes because engagement volume dominates.

Application Rate by Form Response Count

Form Responses	People	Applicants	Avg Engagement Touches	App Rate	Lift
0	160,511	26,757	19.3	16.7%	0.99x
1	41,796	5,372	43.2	12.9%	0.76x
2	7,188	2,242	112.8	31.2%	1.85x
3	2,033	871	192.1	42.8%	2.54x
4	772	396	270.8	51.3%	3.04x
5+	760	353	322.1	46.4%	2.75x

Cross-Tab: Form Responses × Touch Volume

Touch Bucket	Form Bucket	People	Applicants	App Rate
0	0 forms	126,843	12,916	10.2%
1–2	0 forms	442	296	67.0%
1–2	1 form	14,746	254	1.7%
1–2	2+ forms	768	56	7.3%
3–5	0 forms	706	317	44.9%
3–5	1 form	5,408	98	1.8%
3–5	2+ forms	1,009	55	5.5%
6–10	0 forms	2,064	296	14.3%
6–10	1 form	4,536	90	2.0%
6–10	2+ forms	970	39	4.0%
11–20	0 forms	5,840	469	8.0%
11–20	1 form	4,722	131	2.8%
11–20	2+ forms	1,167	44	3.8%
21–35	0 forms	5,887	786	13.4%
21–35	1 form	3,306	157	4.7%
21–35	2+ forms	1,013	75	7.4%
36–50	0 forms	3,563	936	26.3%
36–50	1 form	1,721	214	12.4%
36–50	2+ forms	711	98	13.8%
51+	0 forms	15,166	10,741	70.8%
51+	1 form	7,357	4,428	60.2%
51+	2+ forms	5,115	3,495	68.3%

Note

The 1–2 touch / 0 forms bucket (442 people, 67% rate) is an outlier worth investigating. These are likely high-intent direct applicants who applied through Common App or a direct channel without meaningful tracked web engagement — not a signal that 1–2 pageviews predicts applications.

Key Insight

A single form submission is not a strong intent signal — it converts below baseline and underperforms zero forms within every touch bucket. The signal emerges at two or more submissions, which indicates a prospect returned to engage again. At high touch volumes (51+), form count matters less because engagement volume is the dominant predictor. The actionable threshold is: two form submissions + sustained engagement is a meaningful qualification signal.

Why This Is Important

In the touch volume analysis, avg_form_responses is remarkably flat across all buckets (0.6–1.3), suggesting form submission count doesn’t scale with total touches the way pageviews do. This raises the question: is submitting even one form a disproportionately strong signal of intent, regardless of how many total touches a person has?

If form submission is a strong independent predictor, it has direct implications for:

• How forms are used in the funnel — as intent signals worth acting on immediately
• Whether low-touch persons who submitted a form deserve different treatment than high-touch persons who never did
• The value of optimizing for form completions vs. pageview volume in paid campaigns

Methodology

Persons are grouped by their total number of form responses (0, 1, 2, 3, 4, 5+). For each bucket, application rate is calculated across all persons in that group. A second view cross-tabulates form response count against touch volume bucket to isolate the independent effect.

Query

Application Rate by Form Response Count

WITH touch_counts AS (
  SELECT
    person_id,
    COUNT(CASE WHEN touch_type = 'form_response' THEN 1 END) AS form_responses,
    COUNT(CASE WHEN touch_category = 'engagement' THEN 1 END) AS engagement_touches
  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(tc.form_responses, 0) AS form_responses,
    COALESCE(tc.engagement_touches, 0) AS engagement_touches,
    CASE WHEN a.person_id IS NOT NULL THEN 1 ELSE 0 END AS applied
  FROM persons p
  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
  CASE
    WHEN form_responses = 0 THEN '0'
    WHEN form_responses = 1 THEN '1'
    WHEN form_responses = 2 THEN '2'
    WHEN form_responses = 3 THEN '3'
    WHEN form_responses = 4 THEN '4'
    ELSE '5+'
  END AS form_response_bucket,
  COUNT(*) AS people,
  SUM(applied) AS applicants,
  ROUND(AVG(engagement_touches), 1) AS avg_engagement_touches,
  ROUND(SUM(applied) * 1.0 / COUNT(*), 3) AS application_rate,
  o.overall_application_rate,
  ROUND((SUM(applied) * 1.0 / COUNT(*)) / NULLIF(o.overall_application_rate, 0), 2) AS application_rate_lift
FROM labeled l
CROSS JOIN overall o
GROUP BY 1, o.overall_application_rate
ORDER BY MIN(form_responses);

Cross-Tab: Form Responses × Touch Volume Bucket

WITH touch_counts AS (
  SELECT
    person_id,
    COUNT(CASE WHEN touch_type = 'form_response' THEN 1 END) AS form_responses,
    COUNT(CASE WHEN touch_category = 'engagement' THEN 1 END) AS engagement_touches
  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(tc.form_responses, 0) AS form_responses,
    COALESCE(tc.engagement_touches, 0) AS engagement_touches,
    CASE WHEN a.person_id IS NOT NULL THEN 1 ELSE 0 END AS applied
  FROM persons p
  LEFT JOIN touch_counts tc ON p.id = tc.person_id
  LEFT JOIN applied a ON p.id = a.person_id
)

SELECT
  CASE
    WHEN engagement_touches = 0         THEN '0'
    WHEN engagement_touches BETWEEN 1 AND 2   THEN '1-2'
    WHEN engagement_touches BETWEEN 3 AND 5   THEN '3-5'
    WHEN engagement_touches BETWEEN 6 AND 10  THEN '6-10'
    WHEN engagement_touches BETWEEN 11 AND 20 THEN '11-20'
    WHEN engagement_touches BETWEEN 21 AND 35 THEN '21-35'
    WHEN engagement_touches BETWEEN 36 AND 50 THEN '36-50'
    ELSE '51+'
  END AS touch_bucket,
  CASE
    WHEN form_responses = 0 THEN '0 forms'
    WHEN form_responses = 1 THEN '1 form'
    ELSE '2+ forms'
  END AS form_bucket,
  COUNT(*) AS people,
  SUM(applied) AS applicants,
  ROUND(SUM(applied) * 1.0 / COUNT(*), 3) AS application_rate
FROM labeled
GROUP BY 1, 2
ORDER BY
  CASE touch_bucket
    WHEN '0' THEN 1 WHEN '1-2' THEN 2 WHEN '3-5' THEN 3
    WHEN '6-10' THEN 4 WHEN '11-20' THEN 5 WHEN '21-35' THEN 6
    WHEN '36-50' THEN 7 ELSE 8
  END,
  form_bucket;