Direct answer
Upstream portfolio forecasting
Portfolio forecasting means producing a forecast for every well in an asset with one consistent, declared method, then deciding which of those forecasts actually need review. PronoHorizon runs decline forecasts in batch across a multiwell portfolio, assigns each well a review status, scores the whole portfolio against a held-out period of history, and surfaces the low-confidence wells so engineering time goes where judgment changes the answer. The alternative — rebuilding several hundred wells by hand — does not fail because it is slow; it fails because nobody can review it.
At a glance
- Unit of work
- The portfolio, not the single well.
- Method consistency
- One declared decline method applied across the whole well set.
- Triage
- Per-well review status, so review is exception-based.
- Portfolio scoring
- Holdout backtesting with WMAPE, bias, cumulative error and P10–P90 coverage.
- Streams
- Oil, gas and water forecast through the same workflow.
- Handoff
- Reviewed forecast cases and EUR inputs for reserves, asset planning or economics.
Why well-by-well forecasting stops working
A hand-built forecast is defensible for one well and indefensible for four hundred. The problem is not arithmetic — spreadsheets do arithmetic. The problem is that method drifts.
When each well is fitted separately over months, by different people, with different defaults, the resulting portfolio forecast mixes reservoir behaviour with analyst habit. A well that declines faster than its neighbour may be depleting faster, or it may just have been fitted by someone who prefers a lower b-factor. Nothing in the workbook distinguishes the two.
Consistency is what makes a portfolio forecast interpretable. Once every well is fitted by the same declared method, the differences that remain are signal.
How to forecast hundreds of wells
The workflow is the same four steps regardless of portfolio size; what changes is that steps 2 and 3 stop being manual.
- Normalize once
- Import CSV, Excel or public well data and map oil, gas, water, days on, coordinates and well headers into a single project, so every well is described the same way before any fitting happens.
- Fit in batch
- Run the decline fit across the selected wells, producing P10, P50 and P90 cases per well rather than one curve at a time.
- Score against history
- Backtest the portfolio from a historical cutoff and read WMAPE, bias, cumulative error and P10–P90 coverage for the whole well set.
- Review by exception
- Work the wells flagged by review status and by holdout error, override where reservoir behaviour or operating context warrants it, and leave the rest as fitted.
Exception-based review
Reviewing every well equally is the same as reviewing none of them properly: attention is finite, and spreading it evenly guarantees that the wells that matter get the same few minutes as the wells that do not.
Exception-based review inverts that. Each well carries a status, holdout error identifies where the fitted curve disagrees with what actually happened, and the engineer works down that list. The wells with a clean fit and a clean backtest are left alone — not because nobody looked, but because the evidence says the automatic fit was adequate.
This is also what makes the portfolio number defensible in a review meeting. You can say which wells were examined and why those wells rather than others.
From curves to decisions
A portfolio forecast is an input, not an output. Once flagged wells are resolved, the reviewed forecast cases and their EUR inputs move on to reserves work, asset planning or economic evaluation, carrying with them the assumptions and engineer edits that produced them.
That handoff is where most spreadsheet workflows lose their audit trail. Keeping the exception record and the assumptions attached to the reviewed case is what lets the next technical decision inherit the reasoning rather than just the number.
Frequently asked questions
How do I forecast hundreds of oil wells?
Normalize the production history for every well into one project, run a single declared decline method across all of them in batch, score the resulting forecasts against a held-out period of known production, then review only the wells the backtest and the review status flag. Forecasting every well is automatable; deciding which forecasts to trust is not.
How is portfolio forecasting different from single-well DCA?
The fitting mathematics is the same. What differs is everything around it: method consistency across the whole well set, per-well review status so review can be exception-based, portfolio-level error measurement on a holdout window, and a handoff that carries assumptions forward instead of leaving them in individual workbooks.
What is exception-based forecast review?
It is reviewing only the wells where the evidence says review is needed — those flagged by review status or by holdout error — rather than examining every well equally. It concentrates limited engineering attention on the forecasts most likely to be wrong.
Does batch forecasting remove engineering judgment?
No. The batch run produces a proposal for every well with visible Di, b-factor and terminal decline. An engineer can override any curve, and the assumptions and edits stay attached to the reviewed case, so judgment is applied where it matters instead of being spent re-deriving routine fits.