Skip to the page

Solutions

Next month, with a range. And how far off past forecasts have been.

Ask what next month’s orders will be. nexibi forecasts them from your own history, gives a likely range, and tells you how far off its past forecasts were. When your history is too thin, it says so instead of guessing.

A range, not one number

Orders forecastTarnwick Supply Co. · Assistant and boardSample data
What will our order value be in December?
Assistant

About $498k, likely between $452k and $541k (an 80% range). Past forecasts were off by 6% on average, and beat ‘the same month last year’.

From the saved forecast ‘Orders by month’ 1Its own track record
Orders forecast · monthly, next 3 monthsUpdated today 06:00
Share
ActualForecastLikely range (80%)
400450500550Spring promotionJanMarMayJulSepNovJan$498k
400450500550Spring promotionJanAprJulOctJan$498k
2Your promotions marked
  1. 1Its own track record
  2. 2Your promotions marked

Sounds familiar?

In the words owners and finance teams use:

  • Next month’s number is a spreadsheet someone updates when they have time.
  • Nobody knows how far off last month’s guess was.
  • A holiday moves the month, and the plan does not move with it.

How it works at Tarnwick.

On Tarnwick Supply Co., sample data.

  1. 1

    Ask in plain words. The assistant hands the question to the forecast engine. It is told to use the engine for every future figure, and never to estimate one itself.

    AI Assistant
  2. 2

    It checks the data first, and says what it found: missing months, a month not yet finished, duplicates. Fewer than 4 points, or mixed units, are refused.

    Analysis & prediction
  3. 3

    It picks the method that did best on your own history, and drops any method that does worse than a simple benchmark.

    Analysis & prediction
  4. 4

    It gives a likely range, built from its own past errors, and corrects itself as actual figures arrive. It says how right it has been: off by how much on average, running high or low, and whether it beats “the same month last year”.

    Analysis & prediction
  5. 5

    Your calendar is part of it. Public holidays, Ramadan and Eid in their phases, from 33 country calendars, and your own promotions and price changes. Branches in different countries each keep their own holidays and time zone.

    Analysis & prediction
  6. 6

    Split it by product line or branch, up to two columns. The parts always add up to the total.

    Analysis & prediction
  7. 7

    Keep it. A saved forecast refreshes itself, daily at 06:00 company time unless you choose otherwise, and says when it fell behind and why. Draw it on a board: actuals solid, forecast dashed, the likely range shaded.

    Analysis & prediction

6Split it

Orders forecast · by product lineBoardSample data
TotalDec $498k
400500JanFebMarAprMayJunJulAugSepOctNovDecJan$498k
400500JanAprJulOctJan$498k
Power toolsDec $142k
Hand toolsDec $96k
Fixings and fastenersDec $118k
Safety and workwearDec $64k
Electrical and lightingDec $78k
The parts add up to the total
Split by product line

What if, within what your history has seen.

  • Ask what a week’s promotion or a price rise would do. You get the figure with and without the change, and a wider range.
  • If your history has never seen anything like it, a deeper or longer promotion, or a bigger price change, it says so instead of guessing.
  • Override a month with a reason (“raise January: a known event”). The model’s own forecast is kept beside yours, and accuracy later compares the benchmark, the model and the final figure.

It refuses to guess

AssistantOrders forecast · what ifSample data
What if we ran 15% off Power tools for a week in January?
Assistant

January would be about $27k higher, likely between $12k and $41k more.

And 40% off for a month?
Assistant Your history has no promotion deeper than 20% or longer than two weeks, so I will not forecast that.

How far to trust it.

  • An accuracy page gives one sentence per forecast on how right it has been, and how often its ranges held.
  • Steps far ahead that do not beat the simple benchmark are hidden; an owner can show them.
  • A money forecast is kept in one currency.
  • Unusual actual figures and level shifts are said in words.
  • A question that matches a saved forecast is answered from it, and the forecast is named.

Share it, and act on it.

Analysis & prediction
  • Share a forecast with people, teams or roles. They see the totals, never the rows underneath.
  • Sharing a board that shows a forecast says first what each person would see of it.
  • A flow step can add each forecast point, with its range, to the rows a flow works on.
  • A forecast can be made straight from a list kept by the history engine.

What you get.

  • Next month’s figure with a likely range, and its track record in one line.

  • A board that redraws itself each morning, with your holidays and promotions marked.

  • A what-if that stops where your history stops.

Before you start.

  • Forecasts are set up through the assistant. A saved forecast cannot be reshaped (horizon, split or period), and none can be deleted yet.
  • The forecasts page shows the company total; splits show on boards.
  • It needs history: at least 4 points, and two years of weekly figures for a yearly pattern.
  • “So far this month, and where it will land” works once you choose “when new data arrives”.

Bring the figure you forecast by hand.

In the demo we connect one of your systems, and you ask the questions.