
Funnel Experiments
Part of Lead generation experiments
Comparing form length with lead quality
Test short and long enquiry forms using completed requests, accepted enquiries and the effort needed to respond. Keep the denominators clear.
Compare short and long forms by completed requests and useful enquiries from everyone assigned to each version. A short form may be easier to finish; an extra question may help the first response. Field count alone cannot show which effect matters more.
Give each added question a purpose
State what the longer form is expected to improve. For example: “Asking for service location and a project constraint will help us identify requests we can accept for follow-up.”
Name who uses each answer and how it changes that decision. If an answer has no plausible use, adding the question merely to make a form longer is hard to justify.
For an organisation covered by the Australian Privacy Principles, an added request for personal information must meet the applicable collection requirements. The OAIC says an APP organisation may only collect personal information where it is reasonably necessary for the organisation's functions or activities, and its guidance describes a data minimisation approach. Check whether that rule applies to the organisation and the proposed fields before testing them.
Use the same offer, page explanation, destination team and assessment rule in both versions. Vary the fields and any instructions needed to answer them.
Both forms still need clear labels, required-field information and useful error feedback. Confirm that successful submissions reach the team; a broken long form is not evidence against longer forms.
Pros and Cons of Longer Forms in Australian Context
- ProsBetter lead qualification under Australian Privacy Principles (APP) if data collected is necessary for functions. Helps teams respond faster with context like service location or project constraint. Reduces misrouting and clarification requests.
- ConsMay reduce completion rates due to form fatigue. Risk of collecting unnecessary personal information, which could breach APP 3 (data minimisation). Increases risk of non-compliance if not justified or securely stored.
Compare rates with their counts
Randomly allocate eligible visitors between the two versions, aiming for an even split, and run both over the same period. Set the comparison period and the volume you need in advance. Wait until each version has enough assigned visitors for its rates to be meaningful before comparing results.
For a visitor-level test, a possible primary measure is assigned eligible visitors who make at least one sales-accepted enquiry ÷ eligible visitors assigned to that version. Show successful requests per assigned visitor as a completion measure.
Then examine accepted enquiries ÷ assessed genuine requests to understand the mix of requests received. Show pending cases beside every assessment-based rate.
Do not choose a form from the last rate alone. A long form might have a high acceptance share among a few completers yet produce fewer accepted enquiries from everyone who saw it. A short form might bring more requests but leave many without the context needed for a useful reply.
Consider a hypothetical comparison in which each version receives the same number of eligible visitors, each requester submits once and all requests are assessed. The short form might bring more requests but a lower acceptance share among requests.
The long form might bring fewer requests but a higher acceptance share, so accepted enquiries per assigned visitor could still be higher.
These small hypothetical groups do not establish a winner. They show why the denominator matters.
Check the work after submission
Ask the receiving team whether each version provided enough context for the first response. Record clarification requests, misrouting and answers visitors could not reasonably know yet. Apply the same standard to both versions; filled fields alone should not earn a request acceptance.
Use the enquiry record for request-level assessment.
If the difference remains uncertain, leave the form stable while identifying the specific missing evidence for another comparison.



