Daily Data Quality Checks on SurveyLoopr with DataFlow

DataFlow is a desktop app that checks every interview on a SurveyLoopr ODK Central server as it syncs. SurveyLoopr collects the data; DataFlow adds the quality layer, while the team is still in the field.

On every sync it looks for duplicates, interviews that are too short or too long, outliers, values outside their range, broken logic rules, high "don't know" rates and GPS problems. It also compares back-checks with the original interviews and looks for statistical signs of made-up interviews, enumerator by enumerator. You get one ranked list of problems, a correction log that exports as a Stata do-file, and a Field Report your client opens in a browser.

DataFlow is made by ARCED International, the company behind SurveyLoopr, and it connects to SurveyLoopr directly. A SurveyLoopr account that pays for a hosted server includes one DataFlow seat (details below).

What does DataFlow check on SurveyLoopr data?

Check	What it flags
Duplicates	The same ID in more than one submission
Interview length	Interviews outside the minutes you allow
Outliers	Values far from the rest of the same question
Ranges	Numbers outside the allowed range
Missing answers	Required questions left blank
"Don't know" rate	Interviews where too many answers are "don't know" or "refused"
"Other, specify"	Free text that may belong to an existing option
Logic rules	Answers that shouldn't occur together
GPS	Where the interview was done, from the form's GPS question
Back-checks	Differences between the original interview and a re-interview
Enumerator signs	Statistical signs of made-up interviews, compared with the rest of the team

Rules use the XLSForm syntax your team already writes. To flag a household that owns land but reports no acreage, the rule is ${owns_land} = 1 and ${land_acres} = 0. Back-check differences are counted per enumerator, per back-checker and per question.

DataFlow's checks page explains each check and when it matters. The default thresholds are set out in the Field Data Quality Standard, an openly licensed checklist ARCED International published for field surveys.

What does a daily check produce?

Every sync ends with three things a data manager can act on the same day:

• One ranked list of problems. Problems are ordered by how likely each one is a real error and how much fixing it matters, so the morning's review starts at the top. DataFlow's local AI explains each flag and prepares a fix, and nothing changes until you approve it.
• A correction log. Every correction is recorded with who made it, when and why, including corrections later undone. The raw submissions are never modified. Each download includes dataflow.do, a Stata do-file that replays every correction, step and label from the raw data.
• A Field Report. Progress, team performance and every problem fixed, in one file. The client or principal investigator opens it in a browser, with no install and no login.

The checks run on the computer where DataFlow is installed. DataFlow talks only to your own survey server, which here is your SurveyLoopr server.

How does DataFlow fit with DataSnap?

They look at different evidence, so teams use both. DataSnap works from the audit log and other paradata in SurveyLoopr: time use, question friction, GPS and flagged interviews, to help supervisors choose records to review. DataFlow checks the answers themselves (duplicate IDs, ranges, outliers, logic rules, "don't know" rates and back-check agreement), and keeps every correction in a log that rebuilds the clean data from the raw export.

How does DataFlow connect to SurveyLoopr?

DataFlow signs in with your SurveyLoopr account, so nobody types an ODK password into it. SurveyLoopr reads ODK Central with the login it already keeps for the server and passes the answers back. The access is read-only: DataFlow can list projects and forms, download a form's XLSForm and read its submissions, and nothing else. It never changes a submission or its review state.

SurveyLoopr checks access on every request. If you stop sharing a server with someone, their DataFlow stops reading it at once.

How to connect DataFlow to a SurveyLoopr server

Step 1: Install DataFlow

Download DataFlow for Windows, macOS or Linux from letdataflow.com/download.

Step 2: Sign in with SurveyLoopr

Create a study. Under Connect your data, choose Continue with SurveyLoopr. Your browser opens SurveyLoopr: sign in with Google or your password, as usual, then choose Continue on the "Sign in to DataFlow on this computer" page. Go back to DataFlow; it updates by itself.

Step 3: Pick the server, project and form

DataFlow lists the SurveyLoopr servers you own and the ones shared with you. Pick a running server, then the ODK project and the form, and choose Connect & download. A paused server appears in the list but can't be picked until you resume it in SurveyLoopr.

DataFlow downloads the form's XLSForm, when the form was uploaded as one, and all its submissions.

Step 4: Turn on checks

DataFlow suggests checks from the form and the data, and you choose which to keep. Add rules for the study's own logic, set the interview length limits, and mark the questions that must never be blank.

Step 5: Schedule the sync

Set a schedule, for example every two hours during the working day, or use Sync now when you want fresh data. Each sync reads the new interviews and checks them, so a problem from this morning is on the list by the afternoon.

Step 6: Work the list every day

Go through the ranked problems with the field supervisor. Fix what is clearly an entry error, send the rest back to the field for an explanation or a revisit, and record the reason for every decision. Share the Field Report with the client at the end of the week.

DataFlow's guide to monitoring ODK Central fieldwork every day covers the same routine in more depth, including weekly back-check comparisons.

Is there a free DataFlow seat for SurveyLoopr customers?

Yes. A SurveyLoopr account that pays for at least one hosted ODK server includes one DataFlow seat, at no extra cost.

• Who qualifies: the account that owns the server and pays for it, through a server plan or a server and LooprAI Pro bundle. It is one seat per account, however many servers it pays for.
• How to claim it: there is nothing to redeem. The server owner holds the seat. Install DataFlow, choose Continue with SurveyLoopr and sign in as that account.
• Giving it to a teammate: in SurveyLoopr, open Settings and pick the person on the DataFlow seat card. The list shows the people you share a server with, so share a server with them first if they aren't on it. One person holds the seat at a time, and you can take it back the same way. DataFlow picks up the change at its next licence check, when it starts and every 12 hours.
• When it ends: the seat stays active while at least one of your servers is paid and not paused, and a late payment doesn't end it at once. Once every paid server is paused, cancelled or deleted, DataFlow becomes read-only: projects still open and export, but syncing and check runs stop until a server is paid for again.
• More seats: other people on the team need a DataFlow plan from DataFlow's pricing page. Someone with both a seat and a plan gets the better of the two.

Related guides

Plan which evidence your form should record with the ODK data quality toolkit, and use the ODK submission review checklist for the checks on the Central side. To compare enumerators, read how to monitor enumerator performance. For timing and answer-change evidence from the ODK audit log, see the ODK audit log explained.

Explore ODK hosting Download DataFlow