Product updates
what's new.
A clear record of what has shipped, what we are planning next, and how Analysez AI is becoming more useful for researchers.
Current version
v0.12.0
Beta
01
Released.
v0.12.0· 13 October 2026· Every row counted, clearer results
- Large datasets are now read in full, up to 1,000,000 rows and 5,000 variables. A file above that limit gets a clear message instead of being analysed partially.
- Text predictors in regression models are handled exactly as before again; values are never converted silently. A column that mixes numbers and text is flagged so you can fix it in cleaning.
- 2×2 chi-square results now state that the continuity (Yates) correction is used and also show the uncorrected Pearson value, so they match SPSS line for line.
- If a planned test cannot run, you see it under its objective with a way to edit the plan, and results are grouped by objective again.
- The analysis step now shows a short summary of what was found, and blocked tests name the exact column to fix.
- Several Arabic labels in objectives and cleaning are now translated.
- New Data Explorer in the Explore step: draw histograms, box plots and scatter plots from your cleaned data, with journal style presets.
v0.11.1· 2 October 2026· Faster, steadier AI assistant
- The AI assistant now answers faster when drafting your objectives and reviewing your data.
- If one AI service is slow or unavailable, another takes over automatically, so you are no longer left waiting on "this took too long".
v0.11.0· 24 September 2026· Richer statistics, checked against R
- Table 1 now shows 95% confidence intervals for means and percentages, spread measures, outlier and zero flags, footnotes, and a decimals setting that the screen and report share.
- Significant group comparisons now include the post-hoc result directly in the saved result, and significant chi-square tests show which cells drive the association.
- Each result can show a chart, and a sample-size note appears when a study is under-powered, with how many participants a new study would need.
- New guided screens for prediction models, survival analysis, and adjusting a group comparison for other factors.
- Answers written as "None", "NA" or similar are now kept exactly as typed; only truly empty cells count as missing, and you decide what to do with placeholder words.
- Several smaller clarity improvements, including a copy-ready APA table button.
v0.10.0· 29 August 2026· Editable analysis plan + fixed menus
- You can now edit the suggested analysis plan — change the outcome, group, or test for any objective — before running it, instead of only running it as first proposed.
- Fixed the file menu (⋮) and the table download menu, which could silently do nothing when clicked.
- Report previews now render tables and chart images correctly — a rendering bug could show raw text where a table or figure should be.
- Reporting a problem now automatically attaches a screenshot of what you were looking at, with a preview before it sends, so issues are faster to diagnose.
- Clearer labels throughout the data-cleaning review (e.g. "Automatic checks" and "Human review").
- Several smaller clarity improvements across the cleaning, exploring, and analysis screens.
v0.9.0· 28 June 2026· Guided analysis + smarter cleaning
- Guided statistics: answer a few plain questions and we pick (and run) the right test — no jargon, no test-name menus.
- When 3+ groups differ, we now automatically show which groups differ (post-hoc), in plain language.
- Results come in two views: the line for your paper, plus the full numbers for statisticians.
- Coded survey answers are now counted as categories, never averaged — and Likert scales are detected.
- Cleaning detects more: transposed tables, survey/Google-Forms exports, repeated-measures columns, and re-identification risk.
- Descriptive stats lead with the normality check (two tests + Q-Q plot + histogram); pick figure colours; export as Markdown.
- Missing-data helper: it explains the type of missingness and recommends a method — it never fills values silently.
- Manuscript download in Word (opens in Word and Google Docs) and Markdown, short or full length, with figures and tables.
- After a test, the assistant reads your results and suggests sensible next steps (adjust for covariates, compare another way…).
what we are building next.
These are planned directions, not fixed promises. We will update this page as priorities become confirmed and features ship.
AI-assisted cleaning explanations
Add an optional AI layer that explains patterns and cleaning suggestions after the audited checks. Nothing will change without the researcher’s review and approval.
Broader guided inferential analysis
Make more of the validated inferential test families easier to choose, run, and understand inside the guided workflow.