Acutize product documentation

A practical guide to uploading datasets, preparing data, generating insights, testing model direction, exporting deliverables, and understanding how your data is handled in the workspace.

Product Overview

What Acutize helps you do

Acutize is a browser-based data workflow for people who need to inspect, clean, analyze, test, and export tabular datasets without moving between spreadsheets, notebooks, reporting tools, and separate ML scripts. It is best suited for structured business data such as customers, orders, sales, churn, revenue, inventory, demand, usage, and operations tables.

  • No-code dataset profiling
  • Guided cleaning and encoding
  • Business-friendly insights
  • Generated feature suggestions
  • Fast model preview and comparison
  • Forecasting and export packages

Best first workflow

  1. Upload a dataset
    Start with CSV, JSON, XLS, XLSX, or Parquet. Acutize creates a private dataset record and stores the file under your account.
  2. Analyze and review quality
    Open Analyze & Explore to profile columns, types, missing values, uniqueness, basic statistics, charts, and quality signals.
  3. Clean and encode
    Choose a target column, fill or drop missing values, encode categories, and scale numbers before downstream steps.
  4. Generate insights, features, and model previews
    Use business insights, generated feature suggestions, model ideas, instant training, model comparison, and forecasting as needed.
  5. Export your work
    Download cleaned data, generated-feature data, PDF reports, Python scripts, or a complete ZIP bundle.

Uploading and managing datasets

Private workspace

Datasets are linked to your account, listed in your dashboard, and accessible through authenticated workflows.

Supported files

Use CSV, JSON, XLS, XLSX, or Parquet files. Large files and active dataset counts are controlled by plan limits.

Delete when done

Use the dataset table delete action to permanently remove a dataset and its related artifacts from the workspace.

Analysis, cleaning, and transformation

Analyze & Explore

  • Profiles each column so you can review detected data types before changing the file.
  • Shows missing-value percentages, unique-value counts, numeric summaries, and chart-ready distributions.
  • Provides dataset quality signals that help you decide whether to clean, transform, or model next.

Clean & Encode

  • Select the target column for supervised machine-learning workflows.
  • Choose missing-value handling: none, mean, median, or row drop where appropriate.
  • Encode categorical fields with label or one-hot encoding.
  • Scale numeric columns with standard or min-max scaling.

Insights, features, models, and forecasting

Business Insights

  • Creates dashboard-style exploratory analysis for business users.
  • Summarizes patterns, trends, and recommendations in plain language.

Generate Features

  • Suggests new predictive columns from your cleaned dataset.
  • Each suggestion includes the formula, feature type, and expected impact.
  • You can add or remove suggestions before saving your selected feature set.

Model Ideas

  • Recommends algorithm directions based on the target column and dataset shape.
  • Detects whether the task is likely classification or regression.

Train Models and Model Comparison

  • Instant preview trains several models on a sample so you can compare likely performance quickly.
  • Model comparison displays saved preview results, charts, metrics, and recommendation context.

Forecasting

  • Forecasts dated business metrics such as revenue, sales, demand, usage, or volume.

Exports, reports, and handoff

What customers can export

  • Download cleaned CSV files after running Clean & Encode.
  • Download engineered CSV files after applying generated features.
  • Create PDF project reports with pipeline timeline, profiling, cleaning settings, insights, model results, and forecasts where available.
  • Export a ZIP bundle containing README instructions, report PDF, processed data, cleaning script, requirements, and full-training script when available.

Storage, access, deletion, and retention

Where dataset files are stored

Uploaded raw files, cleaned datasets, engineered datasets, and model artifacts are stored in a private Supabase Storage bucket named datasets.

How access is isolated

Database and storage policies are designed so authenticated users can access only records and files attached to their own user ID.

What metadata is stored

Acutize stores dataset metadata, file name, file extension, status, analysis summaries, cleaning settings, generated-feature settings, forecast results, and timestamps.

Manual deletion

When you delete a dataset, Acutize removes the dataset database record, related training runs and shared-report records.

Retention policy

Active datasets are retained while they remain in your workspace. Automated cleanup is enabled for inactive datasets after 30 days from their last update.

Legal note

This Help Center explains product behavior and should be read together with the separate Privacy Policy and Terms pages.

Account, settings, credits, and support

Account settings

Update your profile name, email, and password from Settings.

Usage & credits

Track current plan, credit balance, credit reset date, plan limits, action costs, and recent transaction history.

Support

Use Contact Support for technical help, billing questions, custom credit packages, onboarding, or bug reports.

FAQ

Dataset Questions

  • Which files can I upload?
    Acutize supports CSV, JSON, XLS, XLSX, and Parquet files.

  • Do I need to clean a dataset before using every feature?
    Not always. Analysis and business insights can start from the uploaded dataset, but cleaning is strongly recommended before generated features, model previews, model comparison, and Python training exports.

  • What is an active dataset?
    An active dataset is an uploaded dataset that still exists in your workspace.

Workflow Questions

  • What is the recommended order of work?
    Upload, Analyze & Explore, Clean & Encode, then use Business Insights, Generate Features, Model Ideas, Train Models, Model Comparison, Forecasting, and Export as needed.

  • How are Generate Features and Model Ideas different?
    Generate Features proposes new columns and can save them into an engineered dataset.

  • Does Model Comparison train new models?
    No. Model Comparison reads the latest saved instant-preview results.

Export Questions

  • What is included in the ZIP bundle?
    The ZIP can include a README, project PDF, cleaned CSV, engineered CSV, and cleaning pipeline script.

  • Does the PDF include my raw data?
    The report is designed as a project summary. It can include metadata, profiling, settings, insights, and forecasts, but raw rows should be exported separately through CSV or ZIP files.

Troubleshooting

  • Upload is blocked: Check the file extension, file size, and active dataset limit for your plan.

  • Cleaning fails: Confirm the target column exists, contains usable values, and is not completely missing.

  • Model preview gives weak results: Run cleaning first, review target-column quality, and remove columns that leak the answer.

  • Forecasting cannot run: Forecasting needs a date/time column and a numeric metric column.

  • Export button is disabled: Some exports unlock only after the related step exists.