Almost every traded product has a buying calendar. Reading a single period against no baseline produces confident, wrong conclusions.
A buyer whose imports fell forty percent last month may be in trouble, or may simply have finished stocking for a season. Without a baseline you cannot tell, and the difference matters enormously if you are about to approach them.
| Driver | Typical pattern |
|---|---|
| Retail seasons | Buying concentrates months ahead of the selling season, not during it |
| Harvest | Agricultural origins ship in a window and are quiet outside it |
| Festival demand | Regional festivals move consumer imports in a predictable annual shape |
| Fiscal year end | Buyers clear budgets or defer purchases depending on their incentives |
| Weather and monsoon | Construction, agri inputs and logistics all shift with it |
| New year shutdowns | Manufacturing origins pause, and the pause shows up as a volume hole |
A recent period that looks weak is often just incomplete. Customs authorities publish on cycles of thirty to sixty days, so the last month or two of any series will fill in after you look at it. Treating an incomplete period as a decline is the most common false signal in trade data.
The most recent one or two periods are usually still filling in. Reading them as a decline is the most common false signal in trade data.
This quarter against the same quarter last year, not against last quarter.
Two points make a line; three years make a pattern.
Drop the most recent period until the market's lag has passed.
Value can move because the commodity moved, not because anyone bought differently.
Three different things move a monthly series and they require different responses. Seasonality is the repeating annual shape — the same months are high or low every year, and the correct treatment is to compare a month against the same month in prior years rather than against the month before it. Trend is the underlying direction once seasonality is removed, and it is the only thing that should change a strategy. Noise is everything else: a bulk parcel, a delayed filing, a one-off project, a reclassification.
The failure that produces most bad decisions is treating noise as trend. A single unusual month is almost always noise, and the discipline of requiring a movement to persist for two or three periods before acting on it will prevent more wasted effort than any analytical technique.
| What you observe | Most likely explanation | What to do |
|---|---|---|
| One month far above the norm | A bulk consignment or an advance buy | Wait for the next period |
| The same month high every year | Seasonality | Compare year on year, not month on month |
| Three months of steady decline | Possibly trend | Check against the prior year before concluding |
| Sharp fall in the newest period | Reporting lag, almost always | Drop the tail entirely |
| Value up, quantity flat | Price movement, not demand | Read the two series separately |
| A gap then a resumption | Late filings catching up | Check whether the gap filled retrospectively |
Once you know a category’s buying window, it stops being a nuisance and becomes a planning tool. Outreach timed to arrive when budgets are being set, samples that land before a range review, capacity offered ahead of a peak rather than during it — all of these are worth more than the same activity performed at a random time. The calendar is visible in the data at no cost, and almost nobody uses it deliberately.
You cannot separate seasonality from trend with one year of data, because you have no second observation of the same month. Three years is where the pattern becomes reliable.
Everything above is a framework, and a framework is only worth what it survives contact with. The useful discipline is to test each assumption against what consignments actually did, because customs data is one of the few commercial sources where the underlying event — goods crossing a border — physically happened and was documented under legal obligation at the time.
Fix the tariff line before anything else. Every filter, every duty figure and every comparison downstream depends on it.
Learn more ›A single period is a snapshot. Three years separate a trend from seasonality, and let you discount the incomplete recent periods.
Learn more ›Frequency and consistency beat size. A steady mid-scale counterparty is usually a better prospect than an occasional large one.
Learn more ›Declared unit values tell you the range you are entering before you quote into it.
Learn more ›Two failure modes account for most wrong conclusions drawn from trade data, and both are easy to avoid once named. The first is reading the incomplete tail of a series as a decline — authorities publish on a lag and revise afterwards, so the last one or two periods will fill in after you look. The second is reading a value movement as a demand movement, when declared value can move because volume moved, because unit price moved, or because the product mix inside a tariff line changed.
Customs data covers goods that crossed a border. It does not cover services, domestic trade, margin, contract terms or intent. Treat it as a dated, quantified observation to corroborate — not as a conclusion that arrives finished.
The difference between teams that get value out of trade data and teams that ran one interesting project is almost never analytical sophistication. It is whether the work became a routine. A saved query reviewed weekly, a short written note against each counterparty you assessed, and a standing habit of checking the period stamp before quoting a figure will out-perform an elaborate one-off study within a quarter, because markets move and a study does not.
The second habit worth building is writing down not just what you concluded but why and when. Records get revised, prices move, and counterparties change behaviour. Six months later nobody remembers whether a supplier was rejected on volume, on price band or on timing, and without that note the assessment simply gets repeated from scratch. A one-line rationale is what converts a list into institutional knowledge, and it costs seconds at the point where the thinking has already been done.
Finally, be explicit with colleagues about the confidence attached to any figure you circulate. A declared value from a complete period, controlled for origin and unit, is strong evidence. The same figure pulled from an incomplete recent period, averaged across a whole chapter, is barely evidence at all — and the two look identical once they are in a slide. Saying which one you have is what keeps trade data credible inside an organisation over time.
At least two years, and preferably three. With one year every month is simultaneously a seasonal observation and a trend observation, and there is no way to separate them.
Because it is incomplete. Customs authorities compile and publish on a lag, and late filings continue to arrive after first publication. Excluding the final one or two periods removes the problem entirely.
Only within a seasonally stable stretch, and even then carefully. Year-on-year comparison of the same month is almost always the safer read.
Frequently, yes — peak demand and thin supply windows both move declared unit values. Benchmark within the same period of the year rather than against an annual average.
Markets refresh on their customs authority's own release cycle — monthly for most, 45 to 60 days for a few. The most recent one or two periods are always still filling in, so exclude them when you are reading a trend rather than treating the gap as a decline.
Yes. Give us the HS code or a product description and the market you care about, and we will return a sample of live customs records filed against it.
Keep reading
The next questions this one usually raises are covered in How often trade data is updated, Competitor analysis with trade data and Using trade data for market research. Each picks up where this article stops, and together they cover the sequence a consignment actually goes through — classification and duty before anything moves, documentation and payment while it moves, and verification of the counterparty before any of it is committed to. Reading them in that order is usually more useful than reading them by topic.
Refresh cycles are set by customs authorities, not by data providers.
Learn more ›Once you can search by company name, competitive research stops being inference and starts being observation.
Learn more ›A customs record answers questions a survey cannot, and cannot answer questions a survey can.
Learn more ›