Piwaka is the change layer for your forecast. It keeps a live record of every change your team makes, in any system, helps at the moment each one is made, and hands your executive team a monthly account, in dollars, of what those changes earned and what they cost.
Modern planning suites measure your changes after the fact: Forecast Value Add tells you, once the actuals land, whether an adjustment helped or hurt. A fair question, a month too late. Piwaka works at the other end of the same timeline, while the change is still being made, on top of the planning system you already have.
Keep the forecast tool you have: SAP, o9, Kinaxis, Oracle, NetSuite, Anaplan or your own model. Piwaka improves the changes you make to it.
A beach-town store, eight products, six weeks of demand forecast. Click the festival button below, and watch what Piwaka shows you before you commit: your forecast already has half that event in it, and adding the manager's +40% on top would double-count it. Or click any cell and type your own number. Take one up, then take one down, and notice that Piwaka treats them differently. That isn't arbitrary.
The forecast your company acts on is not the one your model produced. It is the one your people made afterwards, change by change, between the model run and the moment someone bought stock or booked capacity. Most forecasts are changed by a person before anyone acts on them, and across 147,000 real forecasts, about half of those changes made the forecast worse (Fildes, Goodwin and De Baets, International Journal of Forecasting, 2025). Half your changes are earning money and half are burning it, and nobody can tell you which are which until the money is gone. Every other process that moves this much money has an owner, a record and a review. The month of changes has none of the three.
Your platform records what the number became. The reasoning that moved it, who owned it, how sure they were and what else was on the table, is a free-text note at best. So the same debate starts again every month, and it tends to be settled by seniority rather than by track record.
Forecast Value Add grades the change once the actuals are in, which is a number a planner reads the following month. By then the stock is bought and the sale is missed. Nothing in that loop reaches the person at the moment they make the call.
For a mid-size retailer, the systematic and correctable part of this runs into millions a year, often a meaningful share of net profit. We size it on your own numbers before you commit to anything.
Once the actuals land, Piwaka turns the month’s record into a page an executive team can read in four minutes. What the month’s changes made. What they cost. Which reasons for changing the number are reliably worth acting on, and which are an expensive habit that keeps repeating. Not accuracy percentages: dollars, ranked, with the reasons attached.
It reads in the direction your business runs. We start from the goals the month was meant to serve, margin, waste, service and cash, and work back to the changes that moved them. Forecasting tools grade the maths and leave you to find the money. The change ledger starts with the money.
Piwaka has no forecast of its own and never will. Some months your planners beat the machine. Some months the machine was right and the override cost you. Piwaka scores both directions by the same rules, set before anyone made a change, and reports the result to you. A referee cannot play for one of the sides. We don’t.
A good change is easier to make, open for anyone to contribute to, traceable, informed before you commit, and over time less biased. Those are five things a scorecard read the following month cannot give you.
Paste the supplier email, mention the festival, or type over a number. Piwaka works out what you meant and recomputes everything it affects. The afternoon spent un-editing cells goes away.
Store managers, reps, designers and country managers hear the news first but rarely sit in the planning system, and never will. In Piwaka they contribute in ninety seconds, free, forever. We ask them what they’re seeing, never what the number should be: the calibration does the sizing. Their knowledge arrives as a named, scoped input instead of an email at six o’clock.
Who changed it, why, how sure they were and what else they considered, all attached to the number and written back to your system of record. When someone asks in a meeting why it moved, the answer is already there.
Before you commit, Piwaka shows what is already in the baseline so you do not double-count the festival, how big your move is against the line's normal swing, and what it does to stock, service and budget. You still choose.
When the actuals land, each change is scored against them, and Piwaka learns which kinds of call, from which people, on which products, tend to be worth making. It then offers the debiased option. There is no leaderboard.
Piwaka watches every change but speaks only when the money says it should: when a move is large against the line’s normal swing, thin on evidence, or about to tip a real decision. Then it helps first and asks second, and “gut call” is always an acceptable answer, in one tap. There is a budget on how often it may interrupt each person, and it keeps to it. Nobody wants a tool that argues about everything. Neither do we.
We make no promise about your forecast accuracy. We forecast the change, not your demand. Your baseline stays the truth, because you keep the modern, AI-based forecast you already run, whether that is o9, Blue Yonder or something else. If you do not have one yet, we can help you get one. If you do, we leave it alone and work on the changes you make to it. We target the systematic, repeating part of the error, the part that can be corrected, and we measure it on your own data. That is the Forecast Change Ledger.
A change made through Piwaka arrives with its explanation attached. So the month-end assembly job, the variance commentary, the review pack, the slide for the board, stops being a job: it is generated from the ledger, with every number traceable back to who changed it, why, and what it turned out to be worth. The reason you give at the moment of the change is not admin. It is prepayment, and the refund is your month end.
Changes happen in the planning system, in a spreadsheet, in a meeting that ended with “make it 12,000.” Piwaka reconciles what the forecast was with what it became, so the ledger holds the complete month, whatever surface each change was made on. A large move with no owner and no reason isn’t a gap in the record. It is the first thing the record shows you.
The promotion that flopped, the container that did not sail, the competitor who opened across the road, the six weeks of aging stock in a warehouse in Auckland. None of it is in your sales history, and all of it lands on your forecast anyway.
The job was never to defend the theory. It is to reshape it, again and again, against what is actually happening, and to still know at the end what you did and whether it worked.
The shock arrives in any format. Piwaka takes it in plain language and works out what it affects, before you have finished reading the email.
Everything downstream, reconciled, with what is already in the baseline flagged and a range of options costed against stock, service and budget.
Who asked, what they knew, how sure they were and what it did, so when the actuals land the change still has a name on it and everyone gets a little better at this.
We keep the record to learn from it, not to blame anyone with it.
There is no leaderboard, and there never will be. It is for the slow work of finding out which kinds of call, from which people, on which products, tend to turn out right.
Disruption happens. News lands on a Tuesday. Different parts of the business genuinely see different things. Sometimes it is simply a business call that lands on your forecast anyway. Every planner in every company does this, all the time, and always has.
The research is clear that changes made well improve the forecast, and that some make it worse, and that until now the reasoning behind each one was never written down, so no one could tell which was which.
Piwaka helps you make more of the good changes and gives you a moment's pause on the ones that history says tend to go wrong. It will not stop you. It makes the change quick, keeps the reasoning, and tells you honestly how it turned out.
Everyone sells you the forecast. We work on the change. Here is the research we built it on.
How Piwaka handles a cut versus a raise, what it shows you, and what it leaves off the screen all came out of the published research into what happens when a person changes a forecast. It is the same evidence Forecast Value Add rests on. We act on it earlier.
forecasts get changed by a person before anyone uses them. Adjusting the forecast is the job, and always has been.
of those changes make the forecast worse, across 147,000 forecasts from six studies (Fildes, Goodwin and De Baets, 2025). No company involved could say which half, because the reasoning behind each change was never written down.
that the bias in human forecasts has been documented, and shown to persist even when people know about it. Awareness is not the fix. Correction is.
So we make no accuracy promise. We target the systematic, repeating part of the error, which is the correctable part, and we measure it on your data before you commit to anything.
Start your Forecast Change Ledger Read the case files
Fildes, Goodwin & De Baets, IJF 2025. Oliva & Watson, POM 2009. Flyvbjerg, Holm & Buhl, Transport Reviews 2003. Kahneman, Sibony & Sunstein, Noise 2021. Baecke, De Baets & Vanderheyden, 2017.
A supplier email. A headline. Someone walking over to your desk. None of it is in your sales history, and all of it lands on your forecast.
Every change keeps its reason, its author and its confidence. When the actuals arrive, it gets scored. Not the number in the abstract , the reasoning behind it, the options weighed against it, and whether it was worth making.
All three land on you, and all three land on your forecast. Piwaka handles every one of them the same way: catch it, work out what it touches, write down why, and tell you later whether it was worth doing.
Changes propagate as soon as you make them. When the promo gets cancelled, that's one action rather than an afternoon of un-editing cells. You can branch a scenario before you commit to it, and every number carries a receipt, which tends to end the argument in the meeting rather than start it. Keep your spreadsheets for thinking things through. Piwaka is there for the bit where what you know has to become a defensible change in the system of record.
Every movement since the last review, itemised by the assumption that caused it, with the variance explanation written by the ledger rather than by someone reconstructing it afterwards. Approvals route by blast radius. The audit trail is produced by the work itself, so nobody has to go back and document what they did last quarter.
"We'll land the deal in Q3" stops being forecast-by-shouting and becomes a scoped input with a confidence level on it, credited to you when it turns out you were right. Consensus becomes a negotiation between tracked positions instead of a number everyone quietly surrenders to.
You start read-only. Write-back is scenario-staged and goes through your platform's own documented APIs. Everything is attributed and everything is reversible, and there's no new system of record to defend: your platform keeps the number and Piwaka keeps the reasoning behind it. The security review is usually one meeting. Request the integration note →
The change, the reasoning behind it and the approval all happen in one motion instead of three. Your forecast stops being a quarterly PDF that everyone quietly disbelieves, and starts being a document people trust, because any number in it can show you where it came from.
Most of this site is written for the people who change the forecast. This part is for the team who will be asked to let us in. You start read-only, nothing new becomes a system of record, and the first step needs no integration at all.
You export your forecast, actuals, stock and targets into a cloud tenant you own. We read them there. Nothing is written back until you choose to go live, and even then it is scenario-staged for your approval.
We never hold the number. Your planning system keeps it. We keep the reasoning attached to each change, and hand any change back through the platform's own documented interfaces.
Every change we write is attributed to a person and can be undone. There is no hidden state, and turning us off leaves your system exactly as it was.
SAP, o9, Kinaxis, Blue Yonder, Oracle, NetSuite, Anaplan, Dynamics 365, in-house models, and Excel.
The first engagement, the Forecast Change Ledger, runs on the same export you already produce for finance. No connector, no new system of record, no data leaving your control. We can send your IT team a short technical brief covering data handling, hosting, access and write-back, so the review is done before it starts.
A fixed-fee first month that one manager can approve. No integration and no IT project: we work from the forecast reports your planners already run, plus about an afternoon of their time to tell us what we’re looking at.
What month one can measure depends on what your history holds, and we tell you which before you pay. If your systems kept their forecast versions, as most do for their own scorecards, we rebuild last year’s changes from them and measure the systematic, repeating part of what those changes cost, priced with your own financials and every assumption on show. If your history is thinner, we say so, start your record properly from day one, and measure your first live month instead. Either way the record starts now, and it cannot be started retrospectively later.
You finish the month holding three things that are yours to keep: your change ledger, the priced workbook with its assumptions visible, and a one-page readout of where your changes earn and where they leak. Your data stays yours throughout: held for the engagement, exportable in full, deleted on request.
Month one is measurement. The live help while changes are being made comes after, and only where your own numbers show it will pay.
The Forecast Change Ledger puts a number on what your forecast changes are worth to you, measured on your own data. Our price is a fraction of that number, set so it always sits well below the value we can demonstrate. If we cannot show the value, there is nothing to charge for.
Start with the free estimate, then confirm it in the Ledger’s first month on your real data. You know the number before you decide anything.
We take a share of the value we can show, always well below it, as a fixed monthly licence so your bill is predictable. Our running costs, including the AI behind every change, are built into that.
You never pay by the seat. Bring every country manager, brand lead and salesperson who touches the number. The more of your people in the ledger, the more value there is to share.
Keep the modern forecast you already run, from o9, Blue Yonder, SAP or anyone else, and there is no engine fee. If you do not have one, we can provide the RabbitHawk forecast as your baseline. Either way, we take the baseline as the truth and work on the changes.
See a first estimate yourself in two minutes, or have us confirm it on your own data. The Forecast Change Ledger runs weightless from an export you own, it is a fixed fee, and it credits toward your first months of licence if you go ahead. You keep the findings either way.
Estimate it yourself, free Start your Forecast Change Ledger
We estimate the value up front and confirm it in the Ledger’s first month, once we know what we are dealing with. The exact fraction and terms are set with you then.
If this reads well, your next step is the right one: pass it to your data science team and ask them to pull it apart. We wrote them their own page, with the scoring method in full, the claims we deliberately don’t make, and the test we agree with you before any engine is trusted. We’d rather be examined than believed.
The method, for data science teams Start your Forecast Change Ledger
We didn't start with research. We started by meeting the same person over and over: a planner, in a real company, changing the forecast again, in a spreadsheet, for the fourth time that month, with no way of showing anyone why.
A scorecard that tells you next month is not enough. Make every change easier, more open, traceable and less biased, and find out what they are costing you today.