Forecasting
How AskEdi projects a metric forward using a real trend model, and why it refuses to guess when your data cannot support a forecast.
Forecasting is AskEdi's response to questions that ask what a metric will look like ahead. AskEdi fits a real trend model to your metric's history, tests whether that trend is statistically reliable, and only then projects it forward. When the data cannot support a reliable projection, AskEdi says so instead of guessing.
When Forecasting Activates
AskEdi classifies the intent of your question automatically. Questions that typically activate Forecasting include:
- "What will sales look like next quarter?"
- "What will on time delivery rate be next month if this trend continues?"
- "Forecast revenue for every region over the next three quarters."
- "What will units sold be over the next year?"
You do not need a special keyword or command. There is no separate mode to select.
What Data This Needs
Forecasting gets more reliable the more history your metric has behind it:
- At least a handful of historical periods, and ideally 12 or more. Both the significance test and seasonal detection improve with more data. Short history is not refused outright, but it is far more likely to produce a "no reliable trend" result, which is the honest outcome, not a failure.
- A consistent time grain. Data bucketed by month works best for a monthly forecast, by week for a weekly one. Mixed or irregular granularity weakens the trend AskEdi can detect.
- A category column, if you want a breakdown. Asking for the forecast per region or per product only works when that category exists in your table with more than one value.
AskEdi checks the metric's available history before answering. A question that asks to forecast periods well beyond what your data can support is explained back to you with guidance on what to ask instead, no credit is consumed for that response.
What a Forecast Response Includes
A Forecasting response is five sections. The exact shape of the third section, Forecast Projection, depends on whether a reliable trend exists and whether you asked for a single projection or a breakdown by category.
1. Historical Trend Analysis
States the aggregate trend across your metric's available history: the direction and the timeframe covered. This section always covers the overall pattern, even if you asked for a category breakdown, the category-level detail comes later.
2. Driver & Segment Context
Names which segment of your data contributes most to the current trend. If your table has no category to break the metric down by, this section says so plainly rather than inventing one.
3. Forecast Projection
When AskEdi finds a statistically reliable trend, the projected value is shown along with the method used to produce it: a straight line trend, a trend built around a detected seasonal cycle, or a trend estimated to resist a small number of unusual periods skewing the result.
When AskEdi does not find a statistically reliable trend, the projected number is withheld. The section states plainly that no reliable trend exists in the data, so any number would give false precision to what is effectively noise.
If your question asked for a span of time ahead (for example, "next quarter" or "the next six months"), the projection covers that full span, not just the next single period. The section names the first period's projected value directly and describes the overall trajectory across the rest of the span in plain language.
When your question asks for the forecast per category (for example, "forecast sales for every region"), this section renders as a table instead of prose:
| Segment | Current | Forecast | % Change | Confidence |
|---|
Each row is a category from your table, forecasted independently, so one category can use a seasonal model while another uses a straight line trend if that fits its own history better. Rows are sorted by projected percentage change. A category with no statistically reliable trend of its own shows "No reliable trend" in the Forecast column instead of a number.
If your table has a large number of distinct categories, AskEdi forecasts the top ones by volume and states in the response how many were included and how many were left out.
4. Confidence & Risk Bounds
States how much weight to put on the forecast, and what would cause the actual result to depart from the projection. For a category breakdown, this section summarizes how many of the categories showed a statistically reliable trend, rather than repeating one number for the whole table.
5. Next-Step Recommendation
For a single projection, this is forward-looking monitoring guidance, what to watch for and when to revisit the forecast, rather than a directive action. For a category breakdown, this section names the largest projected gainer and decliner directly.
A forecast is a projection from historical patterns, not a guarantee. AskEdi withholds the projected number entirely when the underlying trend is not statistically reliable, this is a deliberate safeguard against presenting a noisy signal with false confidence.
What This Looks Like
A shortened example, for a question like "what will on time delivery rate look like next month?":
Historical Trend Analysis On time delivery rate has been tracked from December through September, showing an upward trajectory over this window. The fit quality of the trend stands at 7%, and the trend is NOT statistically distinguishable from noise.
Driver & Segment Context The top contributing segment is the high street channel, accounting for 39% of the on time delivery signal.
Forecast Projection No statistically reliable trend was found (p=0.44, only 7% of variation explained). Any projected value should be treated as a low-information extrapolation, not a forecast to act on.
Next-Step Recommendation No reliable projection is available. Monitor the current run rate and revisit the forecast once the next period's actual value is available.
The real response is longer and grounded in your own column names and numbers. This shows the shape, not the content.
Verifying the Numbers Behind the Answer
The trend model, the significance test, and the projected values are all computed server-side before the response is written. To inspect the work directly:
- Analysis view: shows the exact query that ran against your data source.
- Methodology Notes: shows which projection method was used and why, the sample size, and the statistical reliability of the trend, independent of what the response text says.
Frequently Asked Questions
Next Steps
Root Cause Analysis
Understand why a metric moved before projecting where it goes next.
What-If Analysis
Model the impact of a hypothetical change instead of projecting the current trend forward.
Methodology Notes
Inspect the exact projection method, sample size, and reliability test behind every forecast.
AskEdi
Back to the AskEdi overview: how a chat works, response types, and how to get started.
Need help? Email support@edilitics.com with your workspace, job ID, and context. We reply within one business day.
Last updated on
What-If Analysis
How AskEdi models a hypothetical change instantly, using a sensitivity measured from your own data whenever there is enough history to measure it.
Category Comparison
How AskEdi tests whether two categories are really related, or whether a category mix genuinely shifted, using real statistical tests instead of a visual read.