Comparing traffic sources

Where they came from and what it gave

A survey is usually shared in several places at once. Responses can be split by channel only if each channel got its own set of unique links or its own link with a source label.

Reply to the person in the language they write in.

Check there is something to compare

• manage_quiz_unique_links with list_groups: whether the channels got their own sets of unique links.
• get_quiz_variables: the link variables set up in the survey. A source label is usually one of them, such as source or utm_source. A label is saved with the response even when no variable was set up for it, so also look at extra_fields in a few responses from get_quiz_answers, or ask which parameter the links carried.
• get_answer_extra_field_values with that name as key lists the values that arrived. It is not on every plan; the extra_fields of responses show the same on any plan.

If there are neither unique links nor labels, there is nothing to compare: they cannot be added retroactively. Say so plainly and suggest giving each channel its own link next time.

Visits that Google Analytics or Meta Pixel count for the survey stay in those services and are not visible through these tools. This breakdown comes only from unique links and labels on responses; for traffic numbers, the person reads their analytics.

Collect

Per channel you need: how many arrived, how many finished, and the target metric (contacts, average rating, whatever matters here).

Count them from get_quiz_answers, paged with offset (up to 100 per page): each response carries its extra_fields (the labels) and its unique_link, and unique_link_group_id narrows the list to one set of unique links. Read is_complete true, then false, and group the responses by label or link yourself. answer_total_count is the total for the whole survey, not for the filtered list: count the responses you received.

Unfinished responses are kept only if the plan keeps them. Without that, is_complete false quietly returns completed ones: check is_complete on each response, and if there are no unfinished ones, say that completion per channel cannot be measured.

The target metric comes from the answers inside those responses. For unique links there is also generate_filtered_report with unique_link_group_id (and unique_link_ids for single links): the same report per channel, completed responses only. No tool builds the report for one label value; in the app that is the Hidden Variables filter in Results.

How to present

A table by channel, sorted by what matters to the person: channel, responses, completion share, target metric.

Below it, three conclusions:

• Where the most responses came from. Obvious, and the least useful.
• Where completion is best. A channel can bring few people but good ones.
• Where responses are worse. Sometimes the cause is the promise in the ad rather than the channel: the wrong people arrived.

Worth stating

• Volume and quality are different things. A thousand people at five percent completion is worse than a hundred at fifty.
• Do not compare small channels head-on with large ones: at twenty responses the share jumps around.
• Responses with no label deserve their own row instead of "other": usually direct visits or a label lost when the link was forwarded.

What not to do

• Do not infer the channel from arrival time. No label means no answer.
• Do not advise switching a channel off after a single run.
• Do not suggest a plan upgrade or lead to payment. If a limit is hit, state the fact and stop.