Work & Workforce Engineering · Worker

Demand Forecaster

Attach a CSV or JSON export of historical requisition/demand data in the upload chat and get back a monthly demand forecast, skill-enriched, backtested against your own history, and broken down by job profile and band, ready to explore or download once it's generated.

Documentation Pricing
WORKER Workers → Demand Forecaster
Input CSV or JSON, up to 500 MB Output Forecast report (.xlsx) + executive summary (.html) Status Available

What this worker does

Turns a plain historical requisition export into a monthly demand forecast, without you having to build or maintain a forecasting pipeline yourself.

You attach a file in the upload chat; the worker reads every row, works out a sensible training window, backtest length, and forecast horizon from your file's own date coverage, and shows you that recommendation before anything runs, so you can confirm it as-is or correct any one setting first. Once confirmed, each job profile in your file gets its own forecast: several models are trained and backtested against your real history, and the best-performing one is kept automatically. The result is a downloadable report plus a live, chartable view right inside the platform, no spreadsheet wrangling required.

How it works

StageWhat happensExample
1 · Upload & Validation Attach a CSV or JSON file of historical demand/requisition data in the upload chat. Its columns are checked on the spot, job title, job description, and date at minimum, before anything is queued for forecasting. job title ✓ · job description ✓ · date ✓
2 · Context & Skill Enrichment Each requisition's job description is analysed to extract its skills and build demand context, the same skill-intelligence signal used elsewhere on the platform, so the forecast is grounded in what's actually being hired for, not just raw monthly counts. skill extraction · demand context
3 · Forecasting Several forecasting models are run over this skill-enriched data and backtested against your own history. Each job profile self-evaluates its models on backtest accuracy and automatically keeps the best-tuned one; trend, seasonality, and hiring-demand signals all factor into the monthly forecast. multi-model backtest · best-model selection
4 · Report Once it's done, the report is ready to explore or download in the History and Visualize tabs. Credits are deducted only once the report is generated; nothing is charged for rejected files or if report generation fails. charged once ready

What you need to provide

Accepted file formats

CSV and JSON. A CSV should have one row per requisition with a header row. A JSON file can be a single object or an array of records with the same fields. Maximum size is 500 MB per file.

Columns

ColumnRequiredDescription
Job title required Column names can vary; "Position", "Role Title", and "Designation" are all recognised.
Job description required The text used for skill extraction and demand context in Stage 2.
Date required The requisition's start/open/posting date, used to derive the training window and monthly coverage.
Requisition ID optional Helps with per-requisition drill-down in the report's raw-data tables.
Band / grade optional If present, the forecast breaks results down by band, shown in a "lowest to highest" order you confirm before upload.
Location, skills, practice, capability optional Additional context columns that help the forecast when present.

Forecast settings

After a file is attached, the worker reads it in full and comes back with a recommendation for every setting below, grounded in your file's own date coverage, rather than a generic default. Confirm it as-is, or correct any one setting with a short typed reply before the upload actually runs.

SettingWhat it controls
Training start / end dateThe window of your own historical data the models are trained on.
BacktestHow many months of history are held back to test each model's accuracy before it's trusted for the real forecast.
ForecastHow many months ahead the forecast runs, past your training window's end.
Band orderOnly asked when your file has a band/grade column; the order bands are shown in on the report, lowest to highest.
Heads-up. The confirm screen also shows an estimated report time based on your data size; it's a rough approximation (large files in particular can vary a lot run to run), so treat it as a ballpark, not a guarantee.

What you get back

Once processing finishes, two files are generated per upload: the full forecast report (.xlsx) and a shorter executive summary (.html). Both are available from the History tab with a Download button.

Report tabWhat it shows
OverviewCombo list of job profiles, the forecast chart, and backtest/raw-data tables.
MAPE AnalysisPer-model forecast accuracy (Mean Absolute Percentage Error) across job profiles.
MASE AnalysisPer-model forecast accuracy relative to a naive baseline (Mean Absolute Scaled Error).

You don't have to download the report to see it: the Visualize tab picks any of your generated reports from a dropdown and charts it live in the platform, monthly demand curve, historical vs. forecast trends per job profile, forecast confidence, seasonality, and top skills, pulled straight from that report's Excel file.

Pricing

A flat number of credits is charged per forecast report, configurable per workspace; the live figure currently in effect for your organisation is always shown in the upload chat and Overview tab, rather than repeated here as a number that could drift out of date.

Only charged on success. Credits are deducted once, the moment your report is actually generated. A rejected file, or a report generation failure, is never charged for.

Data & retention

QuestionAnswer
Where does my file go? Uploaded securely to your organisation's private storage as soon as it's validated, then picked up from there for forecasting. Not shared outside your organisation.
How long is it kept? 30 days from the date you uploaded the file. After that, both the uploaded file and its forecast report are automatically deleted from History and Visualize, so download or review anything you need to keep before then.

FAQ

How long does the forecast take?

It depends on your data size, more job profiles and more months of history simply take longer to process. Once it's done, you'll find it ready to explore or download in the History and Visualize tabs. No need to keep the page open while it runs.

What's in the History tab?

Every file you've uploaded, with its status and, once ready, the generated forecast report(s) for that upload, each with a Download button. Use Refresh to check for a report that's still being generated.

What's in the Visualize tab?

Pick any of your forecast reports from the dropdown and it's charted right on the page: the monthly demand curve, historical vs. forecast trends per job profile, forecast confidence, seasonality, and top skills, all pulled live from that report's Excel file, no download needed.

What's the maximum file size?

Up to 500 MB per file.