Choose one of three approaches. The steps for that option appear below.
Choose an option above. On a phone, swipe the tabs left or right to see all options →
Vizly profiles usage patterns with inspectable code; people alone classify findings into reasons, patterns and possible explanations.
Check, upload, check parse
Profile per exposure window
Build the findings ledger
Recompute, draft, check matrix
Use: You
Have ready: exit accounts, matched pre-exit usage histories, bounded question and permitted-use limits; the data owner's written confirmation that uploading this extract is permitted. Upload only the fields the question needs, as CSV or Excel, and check the parsed columns match your file.
The vendor states uploaded data is encrypted, isolated and deleted after one hour of inactivity on the free tier; decide with the data owner whether that matches the permission. This record does not establish that upload is permitted. Check row counts and key fields against the original.
Use: Vizly
Ask Vizly to summarise pre-exit usage patterns within each exposure window, generating the Python or R code so you can inspect it.
What to paste into the tool
Task: exit accounts and matched pre-exit usage histories; Bounded research question and permitted-use limits; Confirmed extract with a written upload-permission decision; Parsed dataset checked against the extract; Pattern summaries per exposure window with the generated code. Summarize pre-exit usage within each exposure window relative to the exit date. Separately list any post-exit observations. Show the code you ran. Keep the received originals and earlier results alongside this step output; do not treat carried items as newly generated or checked.
Use: You
Record each finding as stated reason, observed pattern or possible explanation, with the exposure window and product availability period noted for each observation.
Compare values, dates and identifiers against the originals; confirm consent and permission limits; avoid sensitive motive inference.
Use: The research lead
Recompute a sample of summaries; draft the matrix separating stated reasons, observed patterns and possible explanations; post-exit usage is not evidence of pre-exit cause.
Post-exit observations stay flagged; consent and temporal linkage checked before the brief is complete; unresolved findings stay visible.
Hex profiles the windows and drafts the matrix in a reviewer-only workspace; the reviewers trace every row before deciding.
Check permission, set up workspace
Profile, draft the matrix
Trace rows, output check
Use: You
Have ready: the stated exit reasons from your interviews or survey text, and the extract with exposure windows, exit dates and product availability periods. Set up a Hex workspace (30-day Team trial on the vendor page) with access limited to the named reviewers; record who has access.
The vendor page describes enterprise-grade permissions; the data owner decides whether the platform is permitted. This record does not establish that upload is permitted. Avoid sensitive motive inference. Check with the data owner whether a single-file upload without a data warehouse is supported; the vendor page emphasises warehouse connections and does not state it.
Use: Hex
Load the extract into a Hex notebook; profile pre-exit usage per exposure window, then draft the explanation matrix from the checked summaries and stated reasons.
What to paste into the tool
Task: exit accounts and matched pre-exit usage histories; Bounded research question and permitted-use limits; Confirmed extract and platform-permission decision; Workspace with reviewer-only access, documented; Notebook with pattern summaries per exposure window. Profile pre-exit usage within each exposure window relative to exit date. List post-exit observations separately. Show the underlying rows behind each summary. Keep the received originals and earlier results alongside this step output; do not treat carried items as newly generated or checked. Task: Draft the explanation matrix from the checked pattern summaries in this notebook and my stated-reasons list. One row per explanation: the stated reason, the observed pattern, the possible explanation, the exposure window, product availability and missing history. Keep stated reasons, observed patterns and possible explanations in separate columns. Flag any post-exit observation; do not use post-exit usage as a pre-exit cause. Mark rows unresolved where the data does not separate them. Do not infer sensitive motives. Do not send, publish or change anything. Stated-reasons list: [paste it].
Use: The research lead
Trace every row to the underlying rows; check linkage consent, type separation, product availability and missing history; post-exit usage is not used as pre-exit cause.
Unresolved findings remain visible for the named reviewer; confirm the check before the brief is treated as complete.
You prepare a minimal aggregate extract and ask targeted questions; classification into reasons, observations and explanations stays with you.
Prepare, upload, check extract
Ask about pre-exit usage
Record answers, classify
Build matrix, check it
Use: You
Prepare the aggregate extract with the fewest fields that answer the question and get the owner's written approval. Upload to DataGPT; check the field reading.
The vendor's page did not state retention or isolation terms when checked, so the data owner's written approval must accept that gap before anything sensitive is uploaded. This record does not establish that upload is permitted.
Use: DataGPT
Ask DataGPT about pre-exit usage within each exposure window; it answers supported by visualisations.
What to paste into the tool
Task: exit accounts and matched pre-exit usage histories; Bounded research question and permitted-use limits; Aggregate extract plus data terms and approval on file; Connected dataset checked against the extract; Answers with visualizations, each tied to its question. Using only rows inside each stated exposure window relative to exit date, describe pre-exit usage patterns. List post-exit observations separately. Keep the received originals and earlier results alongside this step output; do not treat carried items as newly generated or checked.
Use: You
Record each answer as an observed pattern with its question; classify stated reasons and possible explanations yourself; recompute a sample on the extract.
Compare values, dates and identifiers against the originals; confirm consent and permission limits. Define the exposure window explicitly in each question; the vendor page does not state the tool understands exposure windows itself.
Use: The research lead
Build the explanation matrix with exposure windows, product availability and missing history; post-exit observations stay flagged; post-exit usage is not evidence of pre-exit cause.
Check linkage consent; unresolved findings stay visible for the named reviewer before the brief is complete.
Best when: The owner permits a small extract and you want inspectable code
What it does: Vizly profiles; people classify and recompute
Strong at: Generated code can be read to check window handling
Limit: Vendor-stated retention needs the data owner's assessment
Free plan: Yes, limited (10 messages a month)
Best when: Several named reviewers work one extract in a permissioned workspace
What it does: Hex profiles and drafts; reviewers trace every row
Strong at: Enterprise-grade permissions limit the extract to named reviewers
Limit: Warehouse-focused; confirm single-file upload with the owner
Free plan: Yes, limited (Hex agent trial on Community plan)
Best when: The owner permits an aggregate extract; you want question-driven summaries
What it does: DataGPT answers with visualisations; you classify and re-check
Strong at: Answers come with visualisations tied to the question
Limit: Retention and isolation terms not stated; approval must accept that
Free plan: Paid plan needed
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Stated reasons, observed patterns and possible explanations stay separated.
Each explanation carries an explicit exposure window relative to exit.
Post-exit usage is not used as evidence of pre-exit cause.
Consent and temporal linkage are checked by the research lead.
Check the result against your original inputs. Approve it before sharing or acting.