From Garnet Hawk Solutions

gHawk Analytics

Choose the numbers your team needs to track and save them in a dashboard. Charts use refreshed source data when loaded. You can try it here with public data, before setting up an account.

Try dashboard setup

The demo

Try real public records or a clearly labelled fictional business. No account needed.

The business sample contains 490,800 records in a PostgreSQL database. Sales, invoices, projects and inventory each have their own views. Charts run against the full database; previews show a limited number of records.

The World Bank connection compares annual indicators across eight countries. Missing observations stay missing, and each source shows its publisher, units and update dates.

Choose a dataset. Start with the Canadian economy, weather or bike availability. Open a widget’s details to check its calculation and records.
Compare the results. “What changed?” compares two periods. “What if?” lets you adjust a value or workload and see how the result changes.
Edit the dashboard. Open the widget library to choose a template, review its fields and preview the result. Add the widgets you need, then arrange them on the dashboard.

Public dashboard changes stay in this browser. The data refreshes on the publishers’ schedules; some sources update daily or less often.

Features

Guided setup

Choose what the dashboard is for, select an available data source and review suggested templates. Check the fields and previews, then name the dashboard and create it. Public data needs no account; uploads and team data require approved access.

Analysis packs

Start with receivables: inspect your invoice fields, check balances against your reference totals, approve the measures and publish a dashboard. Save a report preview when ready. An 8,750-row fictional invoice history is included for trying it out. Partial payments count; balances from different dates are never added together.

Open analysis packs

Widget library

Choose totals, line and bar charts, pie and donut charts, scatter plots, distributions, heatmaps, timelines, waterfalls or pivot tables. Each template checks its required fields and previews the result with your selected data.

PostgreSQL sources currently support numbers, lines, bars, tables, pie and donut charts, stacked bars, heatmaps and histograms. The library shows which options work with each source.

Prepare your data

Fix inconsistent categories, convert dates and numbers, find duplicates, split columns or add calculations. Review the affected records before saving a prepared copy. The original dataset stays intact, and you can reuse the steps on another upload. Clean and deliberately messy sample files are available on the Data page.

Dashboards

Add totals, charts, maps, tables and notes. Drag widgets into place and resize them. Dashboard filters apply to widgets with matching fields. Choose a layout style, then pick colours and fonts separately.

Calculations and measures

Each widget shows its formula, filters and source records. Save calculations as measures to reuse them across dashboards. Measures record a definition, unit, owner and version history. The measure owner or team owner can approve a specific version; later edits need a new approval. Download results where the source licence allows.

Record what a measure means, what each record represents and any caveats. If the source fields or database view change, gHawk asks for a review before using that definition again.

Count distinct customers, calculate a ratio for a selected group or compare a closing balance. The same validated formulas run on files and approved database views. Reports record the source revision, reporting dates, definition and data-coverage limits.

Drafts and publishing

With an approved account, “Create dashboard” makes the reviewed initial layout available to people who have access. Later edits save as drafts until you select “Publish”. Published dashboards keep the measure definitions from that publication, while source data continues to refresh.

Comparisons and scenarios

Compare periods by category to see where a total changed. Test a percentage or fixed adjustment, or estimate how long a workload would take at a given capacity. Save the result to a dashboard or download it. These are calculations, with no AI request involved.

For flows such as sales over a period, approved PostgreSQL views support full-source comparisons of sums, averages and record counts. Category contributions apply to sums and record counts. Database scenarios and attribution of balances, ratios or distinct counts are not supported.

Teams and reports

Approved accounts can share dashboards and data with a team. Choose who can view or edit. Save a report now, or schedule one for later; each report keeps the figures from when it was created. Schedule daily, weekly or monthly reports in your time zone, choose the reporting period and send approved colleagues a sign-in link. Preview the result, pause a schedule or transfer ownership. An administrator must configure and test Microsoft 365, Gmail, secure SMTP or an approved relay before enabling email. Delivery history distinguishes accepted, failed and uncertain attempts; acceptance is not proof of inbox delivery.

Report Studio

Bring results from several dashboards into a report for your audience. Choose a layout, add your comments and review the figures. Download a PDF, an editable PowerPoint deck, CSV or JSON, or prepare a video briefing from the same report.

Save the setup for next time. Each result keeps its original dates, filters and calculation. AI can help with the draft; you decide what stays in the finished report.

Video briefings

Turn a saved report into a short video for your next meeting. Choose the figures, edit the script and pick a style. Add an approved voice, or keep it captions-only. Optional AI can help with the commentary; gHawk supplies the calculated figures. You review the words and voice cost before anything is made.

Your team can watch the finished briefing and check its source report, calculations and capture date. Videos and downloads require the same report access. An administrator sets up the render worker and optional voice connection.

Home and favourites

With an approved account, Home lists the dashboards, reports and open alerts you can access. Filter by team and use the stars to add dashboards to your favourites. Favourites are saved for your account.

Alerts

Check an approved measure against a threshold or a previous period. Set a minimum record count and decide how long a problem must persist. Track a problem through recovery, assign follow-up work and keep a due date. Acknowledging an alert does not pretend the underlying problem is fixed.

Measure checks work with uploaded data and approved database sources. Missing or stale figures are shown as omitted checks, never zero. Home includes a weekly review of checked measures, incidents and overdue tasks. Legacy record-level alerts remain available for loaded data.

Your data

You can use a sales export, service log or other business data. The first step is to check the fields and agree on what the numbers mean. Saved joins combine datasets using matching fields.

Public sources: The demo’s sources need no credentials. Website traffic uses Wikipedia page views; service requests come from Vancouver’s public 311 data.
File uploads: Request workspace access. We review requests before creating an account. Approved users can upload CSV or flat JSON files up to 10 MB, 10,000 rows and 80 columns. gHawk checks the structure before saving. Preview a replacement and review its field changes before accepting it. Retained versions keep earlier report evidence available to people who still have access. Replacing a file clears its AI approval.
Other connections: The PostgreSQL connector reads approved views using a restricted account. Calculations run in the database, and the browser receives summaries and a bounded record preview. A deployment administrator sets up and reviews each connection.
Other systems: Garnet Hawk can build connections for your APIs, databases and other systems. Authentication, field mapping and refresh schedules are agreed for each source. The public demo connects only to the sources listed here.

Request access or sign in

Using AI

AI can suggest charts, review field names or answer questions about the data. gHawk checks the proposed calculation and runs the arithmetic. Your team reviews the answer and decides what to do with it.

For data questions, the model chooses from facts that gHawk has already calculated. gHawk writes the numbers into the answer itself. An answer that refers to an unknown fact or an older source version is rejected. This checks the arithmetic and context; it cannot establish whether the original data is accurate. You can review a validated calculation and save it as a draft measure without another AI request.

Before sending a request, you’ll see the provider, model, reasoning setting, estimated cost and what will be sent. When the provider reports its usage, you can also see the cost calculated from those token counts. These are estimates in USD. Questions and file reviews keep a record of the inputs. Opening the widget library, browsing dashboards and refreshing data do not call AI.

Can we use our own model?

Yes. Connections support OpenAI, Anthropic and approved services with an OpenAI-compatible API. You specify the model and its supported reasoning setting. Customer-hosted or private-provider models need an approved secure connection. A hybrid setup can assign different connections to different datasets; gHawk won’t quietly switch providers.

What happens to private data?

Uploading a file does not give AI access. For dashboards and the assistant, the owner must approve a tested connection for that dataset. A saved prepared copy starts with no such saved connection, even though the source it came from may have one. One-off reviews, such as field checks or preparation step suggestions, work differently: each one needs you to choose a connection and give consent at the time, whether or not that dataset has a saved connection approved. Field reviews send structure and quality counts. Preparation advice can include a small sample of values only if you separately choose to share them. Questions use field details and limited summaries of the results. Labels, summaries and questions can still contain sensitive information, so provider terms and hosting locations matter.

Read about data, privacy and First Nations data governance

This is an evaluation environment. Before connecting sensitive data, we agree on access, hosting, retention and any community-specific requirements with you.