Termination Scripts (Optional)

Termination scripts allow you to automatically end a survey for respondents who meet certain criteria, such as being detected as duplicates or having low quality scores.

📌 Overview

Termination scripts are optional and should only be used if your research policy allows early termination. They help maintain data quality by automatically terminating surveys for respondents based on DQC Toolbox results.

Key Points:

Optional: Use only if your research policy permits early termination
How it works: Termination uses Decipher's <term> element to check values in the dqc_data holder. When conditions are met, the survey ends.
Placement: every termination block goes after your second question's <suspend/> — see Where termination blocks go for why.

Quick Reference

Type	Field	Value range	Recommended threshold range	Default Value
Duplicate	`dqc_data.dup.val`	True/False	True	True
Device Score	`dqc_data.dcs.val`	0-100	0-40	10
Data Trust Score	`dqc_data.dts.val`	0-1000	0-420	200
Persona	`dqc_data.per.val`	String	String	KEYBOARD MASHER

Lower scores = higher risk. Termination occurs when score <= threshold, so a higher threshold terminates more respondents.

One scale per score — and where to set your threshold

There is only one scale per score, and the holder stores the raw value on it:

dqc_data.dcs.val holds the device score on its full 0–100 scale.
dqc_data.dts.val holds the Data Trust Score on its full 0–1000 scale.

Write your <term> threshold on that same scale — float(dqc_data.dcs.val) le 10 means a device score of 10 or below, out of 100.

The narrower 0–40 and 0–420 ranges in the table are not a second scale. They are the part of each scale where terminating is defensible, which is why the generator's sliders stop there:

Score	Recommended ceiling	Why stop there
Device Score	40	41 and above is already a Good or Clean device — see Device Score. A threshold above 40 starts terminating devices DQC considers trustworthy.
Data Trust Score	420	Every participant starts from a neutral baseline of 600 — see Data Trust Score. A threshold above 420 starts terminating participants at or near neutral trust, including newcomers who simply have no history yet.

The generator clamps your threshold to the ceiling, so the only way past it is to edit the <term> condition by hand — and if you do, you are ending surveys for respondents DQC has not flagged as risky.

Where termination blocks go

Put every termination block after your second question's <suspend/>. Not a convention — it follows from when the data actually exists:

The toolbox runs client‑side on the first page and needs a round trip to DQC to resolve the participant's identity and scores.
save_dqc_data() runs in <exec when="submit">, so it copies whatever the toolbox has resolved into the dqc_data holder at each page submit — not before.
A <term> only reads that holder. It has nothing to compare against until a submit has filled it.

At the first question's <suspend/> that round trip may still be in flight — and what save_dqc_data() stores then is not a placeholder. The score fields fall back to '0':

dqc_data.dcs.val = getattr(p, 'client_dqc_device_score', '') or '0'
dqc_data.dts.val = getattr(p, 'client_dqc_data_trust_score', '') or '0'

'0' is a perfectly valid number, and it is at or below every threshold you could set. So a termination block placed before the data exists does not fail safe — it terminates everyone.

This is what the placeholder guard is for

What protects you is not a check on the score. It is the check on dqc_data.pid.val, which does get the placeholder (Submission too quick, data not processed) while identity is unresolved — so every generated condition requires a real participant id before it looks at a score.

This is the actual generated block, at the default threshold:

<term 
  label="DQC_Device_Score_Term" 
  cond="(dqc_data.pid.val and isinstance(dqc_data.pid.val, str) and dqc_data.dcs.val is not None and (float(str(dqc_data.dcs.val)) le 10) and dqc_data.pid.val not in ('Could not process','Submission too quick, data not processed','Deactivated-Key'))"
  markers="terminated, dqc_device_score_termination" 
  sst="0">
  DQC Device Score Termination
</term>

If you write your own <term>, keep that pid.val guard. Guarding the score field against placeholder strings does nothing, because the score is never a placeholder — it is '0'.

Which fields get what, while identity is unresolved:

Fields	Fallback
`rid`, `pid`, `cty`, `sub`, `sid`	`Submission too quick, data not processed`
`dcs`, `dts`	`'0'`
`per`	`'NONE'`
`dup`	`False`

By the second question's <suspend/> the real values are in the holder, which is why the termination blocks go there.

Prerequisites

Before implementing termination scripts, complete the General Setup steps:

✅ Added the DQC Toolbox initialization script after the <survey> tag (see Step 2.1)
✅ Added the dqc_data data holder at the end of your survey (included in Step 2.1)
✅ Placed the termination script after your second question's <suspend/> tag — see Where termination blocks go

Termination Types

Termination by Duplicate

Automatically ends the survey for respondents who have already participated in the same survey (based on survey ID).

Checks: dqc_data.dup.val field. Terminates when value is 'True'.

Implementation:

Add this block after your second question's <suspend/> (why):

<term 
  label="DQC_Duplicate_Term" 
  cond="dqc_data.dup.val == 'True'" 
  markers="terminated, dqc_duplicate_termination" 
  sst="0">
  DQC Duplicate Termination
</term>

Termination by Device Score

Ends the survey for respondents whose device score is at or below a specified threshold.

Score range: 0–100, lower scores indicate higher risk. Thresholds above 40 are not recommended — see One scale per score.

Checks: dqc_data.dcs.val field. Terminates when device_score <= threshold (excluding API placeholders).

Implementation:

Add this block after your second question's <suspend/> (why):

Tip: Use the slider below to adjust the threshold. The code block updates automatically.

Threshold:

(Higher means more aggressive. More participants will be terminated. Range: 0 - 40)

<term 
  label="DQC_Device_Score_Term" 
  cond="(dqc_data.pid.val and isinstance(dqc_data.pid.val, str) and dqc_data.dcs.val is not None and (float(str(dqc_data.dcs.val)) le 10) and dqc_data.pid.val not in ('Could not process','Submission too quick, data not processed','Deactivated-Key'))"
  markers="terminated, dqc_device_score_termination" 
  sst="0">
  DQC Device Score Termination
</term>

Termination by Data Trust Score

Ends the survey for respondents whose data trust score falls below a specified threshold.

Score range: 0–1000, lower scores indicate higher risk. Thresholds above 420 are not recommended — see One scale per score.

Checks: dqc_data.dts.val field. Terminates when data_trust_score <= threshold (excluding API placeholders).

Implementation:

Add this block after your second question's <suspend/> (why):

Tip: Use the slider below to adjust the threshold. The code block updates automatically.

Threshold:

(Higher means more aggressive. More participants will be terminated. Range: 0 - 420)

<term 
  label="DQC_Data_Trust_Score_Term" 
  cond="(dqc_data.pid.val and isinstance(dqc_data.pid.val, str) and dqc_data.dts.val is not None and (float(str(dqc_data.dts.val)) le 200) and dqc_data.pid.val not in ('Could not process','Submission too quick, data not processed','Deactivated-Key'))"
  markers="terminated, dqc_data_trust_score_termination" 
  sst="0">
  DQC Data Trust Score Termination
</term>

Termination by Persona

Ends the survey for respondents whose persona matches one of the specified high-risk personas.

Checks: dqc_data.per.val field. Terminates when the persona matches any value in the configured list. No additional validation is needed since the persona defaults to NONE when data is not available.

We only recommend terminating by the following personas. They are ordered from worst to less worst — the more personas you select, the more participants will be automatically terminated.

Available personas for termination (worst → less worst):

Order	Persona	Description
1 (worst)	`KEYBOARD MASHER`	Consistent pattern of poor in-survey behavior. High failure rates, inattentive responses, or rushed completion.
2	`INCOGNITO OPERATOR`	History of using devices that exhibit fraudulent characteristics. Commonly associated with automated traffic, bots, or deliberate fraud activity.
3	`LOSING TRUST`	Trending toward Keyboard Mashers or Incognito Operators. May use clean devices but exhibit neutral or negative historical behavior.

Default termination persona: KEYBOARD MASHER

For more details on all personas, see Data Trust Score — Personas.

Implementation:

Add this block after your second question's <suspend/> (why):

Tip: Use the checkboxes below to select which personas should trigger termination. The code block updates automatically.

Select personas to terminate (worst → less worst):(More personas selected = more participants automatically terminated)

1. KEYBOARD MASHER — Consistent pattern of poor in-survey behavior, high failure rates2. INCOGNITO OPERATOR — History of using devices with fraudulent characteristics (bots, fraud)3. LOSING TRUST — Trending toward Keyboard Mashers or Incognito Operators

<term
  label="DQC_Persona_Term"
  cond="dqc_data.per.val in ('KEYBOARD MASHER')"
  markers="terminated, dqc_persona_termination"
  sst="0">
  DQC Persona Termination
</term>

Threshold Configuration

Understanding Thresholds

Higher threshold = More aggressive: Terminates more respondents (e.g., threshold 30 terminates scores 0-30)
Lower threshold = More lenient: Terminates fewer respondents (e.g., threshold 10 terminates scores 0-10)
Default values: Device Score = 10, Data Trust Score = 200. Maximum recommended: Device Score = 40, Data Trust Score = 420 — why those ceilings.

Adjusting Thresholds

Device Score examples:

Threshold 10: Only terminates very low scores (0-10) - lenient
Threshold 25: Terminates moderate to low scores (0-25) - balanced
Threshold 40: Terminates all below-average scores (0-40) - aggressive

Data Trust Score examples:

Threshold 50: Only terminates very low scores (0-50) - lenient
Threshold 200: Terminates low to moderate scores (0-200) - balanced
Threshold 420: Terminates all below-average scores (0-420) - aggressive

Persona examples:

KEYBOARD MASHER only: Terminates only the worst respondents - recommended
KEYBOARD MASHER + INCOGNITO OPERATOR: Terminates the two worst persona groups - aggressive but useful in certain cases
KEYBOARD MASHER + INCOGNITO OPERATOR + LOSING TRUST: Terminates all low-trust respondents - extremely aggressive

Why the conditions say `le` instead of `<=`

Decipher parses the whole <term> as XML, so a literal < inside a cond would be read as the start of a tag and the survey would not compile — the same rule as no literal &, < or > in <exec> code. le is Decipher's word form for <=, so float(str(dqc_data.dcs.val)) le 10 reads "device score at or below 10".

Customization

The termination code automatically excludes API placeholder values (Could not process, Submission too quick, data not processed, Deactivated-Key) to prevent false terminations.

You can customize the termination message by editing the text inside the <term> element.

Using Multiple Termination Types

You can combine multiple termination types in the same survey. The survey will terminate if any condition is met.

Here's a complete example with all four termination types (using default recommended termination values):

View Complete XML with All Terminations

<?xml version="1.0" encoding="UTF-8"?>
<survey
  alt="Quality Tools Integration - Testing"
  autosave="0"
  builder:wizardCompleted="1"
  builderCompatible="1"
  compat="155"
  delphi="1"
  extraVariables="source,record,decLang,list,userAgent"
  fir="on"
  html:showNumber="0"
  mobile="compat"
  mobileDevices="smartphone,tablet,desktop"
  name="Survey"
  secure="1"
  setup="term,decLang,quota,time"
  ss:disableBackButton="1"
  ss:enableNavigation="1"
  ss:hideProgressBar="0"
  state="testing">

<samplesources default="0">
  <samplesource list="0">
    <title>Data Quality Co-op</title>
    <invalid>You are missing information in the URL. Please verify the URL with the original invite.</invalid>
    <completed>It seems you have already completed this survey.</completed>
    <exit cond="terminated">Thank you for taking our survey.</exit>
    <exit cond="qualified">Thank you for taking our survey. Your efforts are greatly appreciated!</exit>
    <exit cond="overquota">Thank you for taking our survey.</exit>
  </samplesource>
</samplesources>

<style name="respview.client.meta"><![CDATA[
  <link rel="preconnect" href="https://api.dqco-op.com" crossorigin="anonymous">
  <link rel="preconnect" href="https://fpmetrics.dqco-op.com" crossorigin="anonymous">
]]></style>

<style name="global.page.head" wrap="ready"><![CDATA[
(async () => {
  try {
    const { DQCToolBox } = await import('https://api.dqco-op.com/tools/toolbox/DQC_API_KEY');
    await DQCToolBox.getIdentity();
  } catch (error) {
    console.error('Error in client code:', error);
  }
})();
]]></style>
<suspend/>

<exec when="init">
# ===========================================================================
# DQC INTERNALS - GENERAL FUNCTIONS
# ===========================================================================
defaultAnswer = 'Submission too quick, data not processed'
def save_dqc_data():
    defaultDeviceFailuresAnswer = 'None' if getattr(p, 'client_dqc_participant_id', '') else defaultAnswer
    dqc_data.rid.val = getattr(p, 'client_dqc_request_id', '') or defaultAnswer
    dqc_data.pid.val = getattr(p, 'client_dqc_participant_id', '') or defaultAnswer
    dqc_data.per.val = getattr(p, 'client_dqc_persona', '') or 'NONE'
    dqc_data.dts.val = getattr(p, 'client_dqc_data_trust_score', '') or '0'
    dqc_data.dcs.val = getattr(p, 'client_dqc_device_score', '') or '0'
    dqc_data.cty.val = getattr(p, 'client_dqc_country_code', '') or defaultAnswer
    dqc_data.sub.val = getattr(p, 'client_dqc_subdivision_name', '') or defaultAnswer
    dqc_data.dup.val = getattr(p, 'client_dqc_is_duplicate', False)
    dqc_data.sid.val = getattr(p, 'client_dqc_survey_id', '') or defaultAnswer
    dqc_data.dfc.val = getattr(p, 'client_dqc_device_failures', '') or defaultDeviceFailuresAnswer
</exec>

<exec when="submit">
  save_dqc_data()
</exec>

<text
  cond="0"
  label="dqc_data"
  optional="0"
  size="10"
  translateable="0"
  where="execute,survey,report">
  <title>DQC Data Holder</title>
  <row label="rid">dqc-request-id</row>
  <row label="pid">dqc-participant-id</row>
  <row label="dts">dqc-data-trust-score</row>
  <row label="per">dqc-persona</row>
  <row label="dcs">dqc-device-score</row>
  <row label="dup">dqc-is-duplicate</row>
  <row label="cty">dqc-country</row>
  <row label="sub">dqc-subdivision</row>
  <row label="sid">dqc-survey-id</row>
  <row label="dfc">dqc-device-failures</row>
</text>

<suspend/>

<radio
  label="Q1">
  <title>Q1: Are you human?</title>
  <comment>Select one</comment>
  <row label="r1">Yes</row>
  <row label="r2">No</row>
</radio>

<suspend/>

<radio
  label="Q2">
  <title>Q2: Are you a duplicate?</title>
  <comment>Select one</comment>
  <row label="r1">Yes</row>
  <row label="r2">No</row>
</radio>

<suspend/>


<term 
  label="DQC_Duplicate_Term" 
  cond="dqc_data.dup.val == 'True'" 
  markers="terminated, dqc_duplicate_termination" 
  sst="0">
  DQC Duplicate Termination
</term>
<term 
  label="DQC_Device_Score_Term" 
  cond="(dqc_data.pid.val and isinstance(dqc_data.pid.val, str) and dqc_data.dcs.val is not None and (float(str(dqc_data.dcs.val)) le 10) and dqc_data.pid.val not in ('Could not process','Submission too quick, data not processed','Deactivated-Key'))"
  markers="terminated, dqc_device_score_termination" 
  sst="0">
  DQC Device Score Termination
</term>
<term 
  label="DQC_Data_Trust_Score_Term" 
  cond="(dqc_data.pid.val and isinstance(dqc_data.pid.val, str) and dqc_data.dts.val is not None and (float(str(dqc_data.dts.val)) le 200) and dqc_data.pid.val not in ('Could not process','Submission too quick, data not processed','Deactivated-Key'))"
  markers="terminated, dqc_data_trust_score_termination" 
  sst="0">
  DQC Data Trust Score Termination
</term>
<term
  label="DQC_Persona_Term"
  cond="dqc_data.per.val in ('KEYBOARD MASHER')"
  markers="terminated, dqc_persona_termination"
  sst="0">
  DQC Persona Termination
</term>


<radio
  label="Q3">
  <title>Q3: You made it to the end of the example survey</title>
  <comment>Select one</comment>
  <row label="r1">Yay</row>
  <row label="r2">Bummer</row>
</radio>

<suspend/>


</survey>

Custom (non‑DQC) terminations

Your own terminations — screeners, trap questions, quota logic, anything unrelated to DQC — are reported as Terminated too, with no extra work: Decipher marks every <term> with a term:<label> marker, and any term:‑prefixed marker counts as a termination.

<term label="Custom_Not_Human_Term" cond="Q1.r2" sst="0">
  Custom Termination — thank you for your time.
</term>

The complete marker list is sent in the payload's markers field, so DQC can see exactly which marker ended the interview. terminationReason stays DQC‑only — it is "" for a custom termination. See Sending Transaction Data — How status is derived.

✅ Summary

Termination scripts are optional - use only if your research policy allows early termination
Place termination blocks after your second question's <suspend/> tag — the holder is only filled at a page submit, so an earlier block reads a placeholder (why)
Higher thresholds = more aggressive termination (terminates more respondents)
Lower thresholds = more lenient termination (terminates fewer respondents)
Start with lower (more lenient) thresholds and adjust based on your data quality needs
Your own terminations are reported as Terminated too when they set terminated or a term:‑prefixed marker — see Custom (non‑DQC) terminations

Additional Resources

For complete, labeled survey examples with all integration components, see the Decipher XML Generator.

If a termination is firing on the wrong respondents, see Troubleshooting.

📌 Overview​

Quick Reference​

One scale per score — and where to set your threshold​

Where termination blocks go​

Prerequisites​

Termination Types​

Termination by Duplicate​

Termination by Device Score​

Termination by Data Trust Score​

Termination by Persona​

Threshold Configuration​

Understanding Thresholds​

Adjusting Thresholds​

Why the conditions say le instead of <=​

Customization​

Using Multiple Termination Types​

Custom (non‑DQC) terminations​

✅ Summary​

Additional Resources​

📌 Overview

Quick Reference

One scale per score — and where to set your threshold

Where termination blocks go

Prerequisites

Termination Types

Termination by Duplicate

Termination by Device Score

Termination by Data Trust Score

Termination by Persona

Threshold Configuration

Understanding Thresholds

Adjusting Thresholds

Why the conditions say `le` instead of `<=`

Customization

Using Multiple Termination Types

Custom (non‑DQC) terminations

✅ Summary

Additional Resources