Driver-based forecasting for commercial teams

Build a forecast you're proud to present.

Plainsong writes the story behind every number as it forecasts — one legible line connecting what's moving in your business to what's moving in the world. Unlock the insight your best analyst carries in their head: made explicit, auditable, and ready to present.

Built by a commercial finance professional, not a Silicon Valley coder.

Driver: Oil · Brent crude +5% ↳ cost base +1.8% · revenue −0.6%
Driver: UK · CPI +3.5% ↳ cost base +2.6% · revenue +1.2%

Known, but never written

Your best analyst understands intuitively which factors move the P&L. Their potential is wasted if they spend their time reconciling spreadsheets.

Up to the challenge?

When the CFO asks "why has the forecast moved?", the best answer is a clear narrative grounded in quantitative analysis, not a finger in the air.

Gone when they leave

The day that analyst walks out, years of hard-won commercial intuition walk out with them. Nothing transferable is left behind.

What is Plainsong

Where strategy meets accuracy. Have faith in your forecast.

A typical forecast looks like this:

  • ERPs reconcile the numbers.
  • After a lot of back and forward, analysts build a story retrospectively.
  • Then things change, and the process starts over.

Plainsong works differently: machine learning and modern visual analytics let your team bring what they know about the business straight into the forecast — one plain line, not a wall of dashboards.

Reconciliation Strategy Narrative Plainsong
The product

A forecasting instrument, not a dashboard.

  • External data libraryIndicators, FX, commodities, energy and weather from trusted sources like the IMF, Bank of England and FRED, including their forward projections.
  • Driver pipelineTake that data and transform it for the highest significance to your business, separating real signal from noise.
  • Scenario controlsSet your view on a driver, mark one-off events, and watch the forecast update, explanation and all.
  • Walk-forward backtestingAccuracy measured on data the model never saw, reported in plain English.
  • Export anywhereSend the forecast and its narrative to Excel or your board pack in a click.
Monthly revenue · driver-based forecast
3M6M1Y5Y
RETAIL SALES · LEADING BRENT CRUDE · LAGGING SEARCH INDEX · NO SIGNAL

Backtest: within 3.2% across the last 6 windows. Reliable to plan on.

Meet Cantor

The analyst's copilot.

// AI applied where it earns its place

Cantor explains what moved, drafts the commentary, and suggests drivers worth testing. The boring work, done. Ask Cantor in plain language or reach for a command.

The numbers come from Plainsong's deterministic, glass-box engine, not a language model guessing. Cantor never invents a figure, and never makes the call for you.
/explain /narrative /add driver /scenario
/explain the revenue jump in Q3
Q3 revenue is up 4.1% on the prior forecast. The move is carried by retail sales (leading, 2-quarter lag) turning up in July, partly offset by a softer CPI read. Brent crude had no material effect this quarter.
Add a new driver that reflects our EMEA expansion.
I'd suggest Eurozone retail sales (a demand proxy for the new region) or EUR/GBP (to catch the translation effect on reported revenue). Want me to test both?
The research

This isn't a hunch. It's what the forecasting research shows.

The largest review of modern forecasting finds that structured human judgement consistently improves accuracy, and that human-guided variable selection beats automated approaches when real domain knowledge is present.

Petropoulos et al. (2022) · International Journal of Forecasting

Combine the analyst's judgement and the machine's rigour, and you beat either on its own.

Zhang & Norman · 1994

The representational effect

How information is represented changes how well people reason. Plainsong lays the drivers out so your judgement has more to work with.

Amershi et al. · 2014

Humans in the loop

Humans directing a machine-learning system outperform purely automated ones. Plainsong is built to iterate with your expertise, not to replace it.

Hyndman & Koehler · 2006

Honest measurement

Accuracy only counts on data the model never saw. Plainsong backtests every forecast and never marks its own homework.

Why Plainsong

What Plainsong is not.

In finance, trust comes before everything. So we're clear about exactly where the line sits.

Not a reconciliation tool.Your ERP already ties out the actuals. Plainsong is about what happens next.
Not a replacement for analysts.Plainsong maximises what your analysts can do, not a substitute for them. The research is clear: human-guided forecasting is consistently more accurate.
Not a black box.Every number traces back to named drivers you can inspect, challenge, and override.
Not a replacement for judgement.Plainsong surfaces the highest-signal drivers and shows the trade-offs. The strategic call is, and stays, yours.
The best of both

Machine learning can surface deep, statistically significant relationships with your key commercial drivers, but never replaces human judgement and intuition.

Plainsong lets your team work to their full potential: fertile ground for a strategic conversation, not another spreadsheet to reconcile. You bring the context only you have; Plainsong does the rest.

Join the Waitlist

Quantify your instinct.

We're working with a small group of commercial teams ahead of launch. Tell us what you forecast and, once we're ready, we'll show you how Plainsong can help.

No spam. We'll only be in touch about early access to Plainsong.