One print is weather. Three is a climate.
Month-to-month moves are mostly seasonal noise; the trend moves on a slower clock. The minimum unit of meaning in economic data, and how to see a turning point early without fooling yourself.
The charts on this page draw the consumer price index on its 2012 base, the series that ran to December 2025. From January 2026 the NSO publishes the index on a new base, 2024 = 100. The idea on this page works the same on either ruler; the latest figures are in the Consumer prices guide.
The minimum unit of meaning
First principlesEvery month brings a new number, and every month someone reads too much into it. Prices jump half a per cent: alarm. They dip the next month: relief. Both readings were weather. The climate of the economy moves on a slower clock, and telling the two apart is the most useful skill in reading data.
The opposite of noise is not silence; it is persistence. What survives after the spikes have cancelled each other out is the trend, and persistence is the whole of its definition.
Why three? Because noise in Indian price data is seasonal at heart: mangoes in summer, onions after the monsoon, ease in winter. A single odd month proves nothing, and two can still be coincidence. Three moves in one direction, surviving a change of season, is where signal begins.
Five years of weather, drawn beside the climate
m/m %, y/y %The same inflation, measured two ways. The thin grey line compares each month with the one before; the thick blue line compares it with a year earlier. One jumps; the other judges.
The spike was weather: the line that matters barely noticed
Watch how violently the grey line jumps and how calmly the blue one responds. The grey line is not wrong; it answers a faster, noisier question. For direction, trust the thick line. For early warning, watch the thin one, but demand three months of agreement before believing it.
In July 2023 the monthly reading spiked to –, the largest in this window, on a tomato-driven surge in food prices. The very next month the grey line printed –, and the month after, –. The surge had unwound.
The thick line rose and fell back over the same months, from – to – and back to – within a quarter. That is what weather looks like: loud, brief, and gone.
How to see a turning point early
An honest methodThere is an honest way to be early. When the fast line turns, note it and wait. If a second month agrees, lean. If a third agrees, the slow line will follow, and you have read the turning point a quarter before the headline admits it. The cost of the method is real: sometimes you must sit with a suspected turn and say nothing.
The discipline cuts both ways. A trend that has persisted can also end, and refusing to see a genuine turn is as expensive as inventing a false one. The rule that guards against both errors is the same rule: three months, one direction, a change of season survived. Anything less is a guess with a chart attached.
How to misread this concept
- i.
“This month's number proves it.”
One print is weather. The July 2023 spike looked like runaway inflation and unwound within two months. Proof needs three readings in the same direction; before that you hold a hypothesis, not a finding.
- ii.
“The trend never changes.”
It changes rarely, but it changes. Inflation fell through 2025 in a genuine break, visible precisely because month after month agreed. A method that dismisses single spikes must also admit sustained runs, or it becomes stubbornness with a formula attached.
- iii.
“More data points always mean more meaning.”
More points help only when they are allowed to disagree with you. Collecting readings until one confirms the story is not analysis. Meaning comes from consistency across time, never from the size of the pile.