You just left a demo. Revenue lines split into base, downside, and upside cases. A slider labeled “input cost inflation” moved and every number moved with it. A natural-language box answered “what happens to gross margin if Asia volume drops 12%?” in seven seconds, with a chart and the five line items most affected. The word on the booth — Anaplan, Oracle EPM, Workday Adaptive, Pigment, Board, whichever one you saw — was predictive planning. And the pitch was clean: buy this, and your planning function is modernized. Here is the problem the demo does not mention. Predictive planning is not software. It is a discipline, and the software is what makes the discipline useful — not the other way around.
Start with why the word is on every booth. “Predictive planning” was sitting unclaimed in the analyst vocabulary around 2019, and vendors that already sold financial planning and analysis software needed a story to refresh their existing platforms. AI features were arriving inside those products anyway. Wrapping them in a category name that sounded forward-looking and CFO-friendly was the obvious play, and by 2023 every major vendor had a “predictive planning” product page. Be fair to them — they are not lying. The features are real, the AI is real, and the platforms genuinely do things their predecessors could not. What they are doing is what software companies have always done: claiming the vocabulary that travels with their roadmap. The category error is not theirs. It belongs to the buyer who confuses the vocabulary with the discipline behind it.
Strip the marketing off and the platforms cluster around five real capabilities. Data plumbing — pulling actuals from the ERP, headcount from the HRIS, pipeline from the CRM, and reconciling them into one consistent model. Model management — building, versioning, and governing a plan more than three people can touch without breaking it. Scenario modeling — spinning up a what-if cheaply and comparing it to base. Collaboration and workflow — multi-user editing, approval flows, comments tied to cells. And the newer AI-assisted modeling and anomaly detectionlayer — pattern recognition, forecast suggestions, outlier flags, natural-language queries. Every one of these is genuinely worth paying for. An organization that has them is meaningfully better off than one that doesn’t. The mistake is not in buying them. The mistake is in believing that buying them is the same as installing the discipline.
Because here is what the software cannot give you. It cannot give you judgment under ambiguity — the platform can model a 12% Asia volume drop, but it cannot tell you whether three customer conversations and one regulatory filing add up to that being the scenario worth modeling. It cannot give you scenarios that mean something — a scenario is not a slider position, it is a coherent narrative the leadership team can rehearse against, and the platform computes the consequences of that story only once a human writes it. It cannot give you the nerve to make a stake, the political will to reallocate $15M from the regional president who just lost the argument, or the leadership accountability that makes any of it run. Software supports a discipline. It cannot install one. That distinction is the whole conversation.
So define the thing the software is meant to support. Predictive planning is the continuous discipline of converting weak signals into strategic decisions — using AI to decide faster than your operating environment changes. It is the descendant of the scenario planning Pierre Wack built at Shell in the early 1970s, the discipline that let Shell’s leaders move faster than every peer when the 1973 oil shock hit because they had already rehearsed it. What changed since then is the operating environment: the cost of running a scenario has fallen by an order of magnitude, the velocity of meaningful change has gone up, and the audience broadened from a small planning priesthood to CFOs who have to operate the discipline themselves. Predictive planning is what scenario planning becomes when it has to be cheap, fast, and broad. The methodology updates. The lineage holds.
The discipline runs as a four-phase continuous loop — Woodring’s Loop. Scan asks what’s emerging and maintains a curated signal layer. Story asks what it could mean and turns signals into three to five pressure-tested scenarios. Stake asks what you’ll commit and converts scenarios into resource commitments under uncertainty. Steerasks what to adjust and watches reality unfold against the scenarios, moving money as some converge and others collapse. Notice where the software fits. Scan gets real help — AI can triage filings, news, and data anomalies at a volume no human team can match. Steer is mostly software plumbing — refreshing actuals against forecast, flagging variances, surfacing convergence. But Story and Stake, the middle of the Loop where the discipline actually lives, get almost none. Software is useful at the edges and beside the point in the center. That is the honest accounting.
This is not a new mistake, which is why it is worth taking seriously. Between 1998 and 2008, Fortune 1000 companies spent an estimated $200 billion on ERP systems, sold on the same logic driving predictive planning purchases today: a single system of record, real-time data, integrated planning across functions. Hershey compressed a 48-month SAP deployment into 30, went live weeks before Halloween in 1999, and took a 12.4% sales drop — roughly $100 million in undelivered product. HP’s 2004 SAP migration cost about $400 million in revenue when 20% of orders failed to move cleanly. The companies that succeeded with ERP did so because they used the system to run a discipline they already had. The ones that bought the system hoping it would give them a discipline mostly failed, and the failures carried nine- and ten-figure write-downs. The technology was never the problem. The assumption that the technology was the discipline was the problem. Predictive planning software is on the same track.
Understand why the buyer keeps making this trade, because naming it is half the defense. Procurement is a familiar shape and leadership change is not. A $4M software contract with implementation services, a steering committee, a Gantt chart, and a go-live date is a known motion — every CFO has run one, and at the end there is a system in production and a checkbox on the strategic agenda. A discipline has none of those things. No go-live date, no vendor on the hook for the outcome, no system to show the board — only a different way of working whose sole evidence is that decisions look different than they used to. So the temptation is to translate the discipline back into the shape procurement understands: buy the platform, stand up the model, declare predictive planning installed. Eighteen months later the dashboards are gorgeous, the data is clean, and the plan is still wrong by April. Software gives the appearance of motion without the substance, and that appearance buys another year of believing the model is sound.
So run a real diagnostic before the next license renewal — the Software/Discipline Gap Audit. For each phase of the Loop — Scan, Story, Stake, Steer — score two numbers from 0 to 10 with your leadership team in the room. The first is what your current software does for that phase. The second is what your discipline does for it. A high software score paired with a low discipline score is the most common pattern in companies that have already bought a planning platform, and it is the reading that tells you the next dollar belongs to the discipline, not the next seat. The rarer pattern — high discipline, low software — is easier to fix, because the platform can simply be bought. Run it honestly, with the executive team, not as a procurement exercise. The point is not to grade the vendor. The point is to see the gap clearly enough to act on it.
One more distinction, because the vocabulary keeps collapsing it. Predictive planning is not forecasting — forecasting predicts a number, the discipline prepares for a range. It is not scenario planning alone — that is one phase, Story, and organizations that adopt it without Stake and Steer produce excellent scenarios and act on none of them. And it is not predictive analytics — a category of software that projects churn, demand, or fraud probability with a confidence interval. Predictive planning gives you a discipline for deciding under uncertainty when no confidence interval will save you. All three are useful inputs. None of them is the thing. Conflating them is exactly the category error vendors are happy to encourage and executives should refuse.
Here is the whole argument in one line: buying capability is not the same as installing discipline. The CFO who buys Anaplan and declares planning solved in 2027 is the same CFO who bought SAP and declared it solved in 2002. Buy the platform — the data plumbing alone often justifies it, and Scan and Steer genuinely need the plumbing. But buy it knowing what you are buying: a tool that supports a discipline you still have to install with your own leadership team. Software does what software does. The discipline is what makes any of it useful. Get the order right, and the platform earns its keep. Get it backwards, and you have bought an expensive way to keep doing exactly what you were already doing.