How CFOs Can Build a Continuous Inflation Forecasting Model

By Andrew Blair, Head of Global Presales and Value Advisory, Kyriba. Originally published on the Kyriba blog.

Inflation ranks first globally in Kyriba's CFO Risk Radar, and it is the only risk CFOs place in the top three in every one of the nine markets surveyed.

CFOs already know why inflation matters. It hits procurement costs, wages, debt service and FX-driven input pricing, and does so at different speeds in different markets. The harder problem is that most forecasting processes were not built for a risk that never actually resolves.

Persistent risks need a different forecasting method

Risks that spike and resolve fit neatly into a calendar-based review cycle. The spike itself is the trigger: something happens, the model gets updated, the team moves on. Inflation does not work that way. Part of the reason is that it never has just one driver: rate decisions, currency moves, unpredictable tariffs and political instability all feed into it at once, each on its own schedule. Inflation risk sits at an elevated, moving level indefinitely and never produces its own “update the model now” moment the way a credit rating downgrade or a sanctions announcement does. A risk that never announces its own recalibration point is easy to leave running on last quarter's assumptions.

That is why other economic data matters here: it changes the inflation forecast directly, and quickly. Consider what played out in the US last summer over the July jobs report. A weaker labour market gave the Fed more reason to pause on additional rate hikes, but it also limited how aggressively policymakers could use tighter rates to push inflation lower without worsening growth and employment risks.

Days after the jobs report was released, the consumer price inflation (CPI) report added another layer: headline inflation cooled slightly to 3.4 percent, the second straight month of easing. Yet energy prices are still predicted to keep swinging because of political instability in the Middle East, and economists continue to warn that higher oil and fertiliser costs will keep working through the economy in the months to come.

One cooling number does not cancel out that volatility; a calm headline and a calm outlook are not the same thing. That is the speed this kind of risk moves at: within days, traders lifted the odds of a September rate hold to about 64 percent, up from roughly 50/50 before the inflation data. You need that kind of data flowing into your model as an input, on the same timeline it arrives.

Build a continuous inflation model

Only 26 percent of finance teams run continuous or automated scenario modelling for external risks that behave this way. 34 percent review monthly, 35 percent quarterly and 3 percent only annually. For a risk like this, a quarterly cycle is a lot of runway for a stale assumption to sit in your forecast unnoticed. Two habits solve that problem:

1. Map your inflation channels to their leading indicators. Procurement, wages, debt service and FX are each moved by a different set of external data, and that data does not all arrive on the same schedule. Interest-rate and currency signals tend to update on a short cycle; wage and supplier-cost data moves on a slower, more irregular schedule and takes more hands-on work to translate into a model assumption. Either way, the goal is the same: know which indicator moves each channel, and put each one on the update cadence its own data actually supports, rather than letting all four wait for the same quarterly review.
2. Treat directly related economic data as a live input, not background noise. When a rate decision, a currency move or a labour report changes the outlook for one of these channels, treat that as new information for your model. Build the habit of asking, the day that kind of data lands, whether it changes one of your inflation assumptions, and updating it on that timeline rather than saving it for the next scheduled cycle.

Calibrate inflation risk by market

A single blended inflation assumption erases exactly the variation that determines how much margin inflation actually costs. Take import exposure: in a market that sources heavily from abroad, currency swings compound the local inflation rate directly, so the FX-to-input-cost linkage deserves its own explicit line in the model. Labour markets vary just as much: where retention pressure is high, wage growth shows up in comp costs faster than a quarterly cycle can catch, so that assumption needs its own, tighter update schedule. Debt follows the same logic: a market with significant floating-rate exposure ties its discount rate to the same policy signals driving local inflation, so the two need to move together in the model.

The real test is whether your model reflects which channel is under the most pressure in each of your markets right now, and how fast it gets updated the moment the inputs to that channel change.

The inflation forecasting checklist

Answer these for your five largest markets. If you cannot answer all three with confidence, your model is running on assumptions older than the data behind them.

1. Do you have a continuous or rolling process for updating inflation assumptions, or does it wait for the next scheduled model refresh?
2. Can you name, for each of your top markets, which channel (procurement, wages, debt service or FX) is currently doing the most damage to margin?
3. When a data point changes the outlook for rates or labour costs in one of your markets, does that update your inflation assumptions right away, or does it sit until the next cycle?

A model that fails any of these questions is already running on outdated inflation data, sometimes by a full quarter or more.

Forecast as fast as the risk moves

Inflation will not resolve, and it will not wait for the next quarterly review to matter less. Build a model that runs continuous, channel-by-channel scenario modelling, one that moves as fast as the data does. That is the edge.

This article was originally published on the Kyriba blog.