Once you've finalized your survey, you should be ready to set up respondent sources for it. To create a respondent source, click Add respondent source on the Summary tab. This will take you to the Add respondent source wizard:

Different kinds of respondent source
Respondent sources fall into a few kinds: integrated providers, first-party data, external providers, and sources that bring in responses you already have or generate synthetically:
- The platform automatically sources respondents from integrated providers.
- First-party data, available on the Team or Enterprise plans, are used to survey your customer base.
- External respondent sources can be used on the Team or Enterprise plans, allowing custom integration with your sample provider. Dedicated setup guides cover InnovateMR and Dynata Signed or Signed+ links. You can also connect a panel through the generic bring your own sample flow.
- Third-party Import loads completed responses that were collected outside MX8 Labs so you can report on them alongside your other sources.
- Synthetic sources generate respondents rather than recruiting them — either Synthetic Twins trained on a previous survey, or Synthetic Profiles built from a prompt per respondent.
- Send personalized SMS invitations to a browser survey with Twilio Text to Web, using uploaded contacts and a validated invitation template.
- You can also field by phone with a Twilio Voice or Twilio Text source.
For a Basic / External Providers source, choose whether to keep separate live and test links or use one Survey URL. A single URL allows testing before the source first goes live and records live responses while it is open. After recorded live fieldwork, new arrivals while paused or complete are rejected instead of entering test data. See External-provider link controls.
Manage individual respondents
On the survey's Summary tab, select Respondents beside a respondent source. The action is unavailable when that source has no respondents. The dialog shows respondents from that source only; the actions available depend on your permissions.
Search by respondent ID or sequence number, filter by status, sort the table, and move through its pages. The table includes sequence, ID, status, language, first and last response times, duration, suspect score, duplicate finding and termination reason.
Select up to 50 respondents for one update, choose an action, and confirm:
| Action | Effect |
|---|---|
| Delete | Permanently removes the selected respondents and their transcripts. This cannot be undone. |
| Clear responses | Removes answers and warnings while retaining respondent identities. |
| Mark poor quality | Marks selected completed respondents as poor quality, retaining their answers. |
| Clear poor quality | Clears a poor-quality outcome and disables subsequent automatic quality screening for those respondents. |
Wait for processing to finish before making another update. The respondent list and reports update afterward. If processing fails, review the error and selected respondents before trying again.
Re-invite a previously terminated panelist
Before re-inviting a panelist who was previously terminated from this survey, delete their existing respondent record:
- On the survey's Summary tab, select Respondents beside their respondent source.
- Find and select the panelist, choose Delete, and confirm.
- Wait for deletion to finish, then re-invite the panelist.
Deletion permanently removes their existing responses and transcript and cannot be undone.
Read a transcript
Select Show (the eye icon) on a respondent's row to read their transcript. The dialog shows the respondent's status and the interviewer questions and answers.
The respondent ID in this example has been replaced, and personal answers have been redacted.
For a verified privacy-erasure request, follow Privacy Compliance; clearing answers is different from permanently deleting a respondent.
Understand source pricing
For panel respondents you source through MX8 Labs and for synthetic respondents, you pay your account’s quoted rates. Other respondent sources have no respondent charge through MX8 Labs; platform usage and any separate supplier charges still apply.
Check your quote before launching. Existing bookings keep their agreed charges. See Account plans and usage for pricing and currency limits.
Use a respondent source name in survey logic
Read s.respondent.source when survey logic needs to identify the respondent's source. It returns the source's display name when one was captured at launch and otherwise falls back to the technical source type:
from survey import Survey
s = Survey(**globals())
if s.respondent.source == "Customer panel":
s.show_message("Thanks for taking part in our customer study.")
The platform automatically adds a launch-time source name to reporting with the stable ID respondent_source and label Respondent source. Do not call s.store_value() just to make the source reportable.
For integrated providers, the respondent source definition will be taken from any quotas you have in the survey:

You can then edit the Demographic Targeting field to add additional targets available from the sample provider. Targeting is edited as structured JSON in a validating editor — describe the demographic quotas and targeting logic with a target percentage next to each value, and the editor will flag anything that isn't valid JSON before you can save:

Adding boost respondent sources
Sometimes, you might want to augment a specific survey with more respondents of a particular kind without skewing the overall results. To do this, you create a new respondent source with the required distribution and specify that it should be weighted to the main respondent source you have set up.
If you look at the example below, we have boosted the respondent source for African American respondents as follows:
- The respondent source definition specifies only African Americans and a lower percentage of Hispanics.
- The weighting respondent source is set to the primary respondent source, "US Genpop".
This will give us an additional 300 responses in the African American group, which will greatly increase the statistical significance of results within this group without impacting the overall distribution of results in the survey.

Reading live incidence rate while in field
Once a respondent source is live, the summary tab shows a live incidence rate (IR) for each source alongside the rest of the progress metrics. IR is calculated on the fly from respondents who have already entered the survey, so it reflects what is actually happening in field, not the estimate you used when pricing the study.
Hovering over the IR bar reveals the breakdown that makes up the number:

The breakdown separates four outcomes:
- Completed — respondents who finished the survey and passed quality checks. This is the numerator in the IR calculation.
- In progress — respondents who have entered the survey but not yet finished. A persistently high in-progress figure is usually a sign of a long survey or a confusing question.
- Terminated — respondents who were screened out by a termination question or a failed quality gate. This is normally the biggest driver of IR, and the one to watch when comparing sources.
- Poor quality — respondents flagged and removed by the fraud and quality systems. A spike here is a flag to investigate the source before continuing to spend.
Because IR is computed per source, you can quickly see whether a blended underperformance is caused by the audience as a whole or by a single source dragging the number down. If one source has a materially lower IR than the others, it is usually worth pausing or re-targeting that source rather than widening every source at once.
Live IR is recomputed as new respondents come in, so the number will move during the first few hundred completes. We would not recommend acting on the IR until a source has at least 50 to 100 entries, which is usually enough for the per-source figure to settle.
For a full breakdown of what is happening inside a source — where respondents are terminating, which devices are struggling, and how each demographic group is progressing — click the Report button next to the source to open the field report.
Reading response activity
The Activity value shows the period from the first completed response to the latest completed response. It appears on the survey list, the survey Summary, respondent-source cards, and the respondent-source list and detail views.

Dates use your browser's locale and are kept compact in the page. Hover over or focus the range to see the exact local date and time for the first and latest completed responses.
An open-ended range means one endpoint is unavailable:
- → latest date — the platform has a latest completed response but no recorded first response. This can occur for historical data created before first-response tracking was available.
- first date → — a first completed response is recorded but the latest endpoint is temporarily unavailable.
- No responses yet — neither completed-response timestamp is available.
Clicking the main area of a respondent-source card opens its field report when the report is available. If reporting is unavailable because the source is new, updating, or in an error state, the card opens Edit for users with edit access and View for read-only users. The separate Respondents, Report, View, Edit, and Delete actions remain available according to your permissions and the source state.
Recovering a failed creation or update
A respondent source changes to Error if its provider rejects the request after submission. The source shows the provider's explanation, such as demographic targets whose percentages do not total correctly.
If you have edit access:
- Open the errored source or select Edit.
- Read the failure description and correct the retained source settings.
- Save the source to try again.
The editor retains the submitted settings so you can correct the failed values. Do not copy credentials into support messages, screenshots or logs.
Select Delete when the failed source is no longer needed. You can delete a source whose initial creation did not complete. The normal restrictions still apply: sources with live fieldwork or response data cannot be removed through this recovery path.
Voice and text dispatch progress
For Twilio Voice, Twilio Text, Twilio Text to Web, and Twilio Text-to-Caller sources, use the source's progress bar on the Summary tab to monitor delivery across the uploaded contact list. The bar groups destinations into these states:
- Planned — the destination is in the campaign but has not been queued yet.
- Scheduled — the destination has been queued for delivery.
- Starting — the platform has handed the attempt to Twilio and is waiting for a terminal provider update. Voice calls remain here while they are initiated, ringing, answered, or in progress.
- Delivered — Twilio reported the terminal delivery milestone. For voice, this means the provider reported the call as completed; an answered or in-progress call is not yet Delivered.
- Failed — the attempt could not be delivered, for example because a call was busy or unanswered, or Twilio returned a delivery failure.
Do not treat Delivered as Complete. The progress bar describes what happened to the outbound message or call; the field report describes what the person did after entering the survey. A delivered destination can still be New, In Progress, Terminated, Poor Quality, or Complete.
To investigate an individual destination, open the source's field report and, if you have export permission, select Dispatch CSV. The export retains every retry attempt and marks the latest one for each destination. See Field reports.

