The right question is not “why is it down?” but “down compared with what?”

A drop in revenue in August, after a sales period, or in a B2B business tied to school holidays does not mean the same thing as a decline that persists outside the season, hits several products, and remains despite sales actions. Eurostat notes that seasonality refers to regular influences distinct from calendar effects, and that seasonal adjustment exists precisely to make series comparable. Insee also shows that sales periods, holidays, and working-day effects blur retail readings.

The SME trap is comparing a “weak” month with a “strong” month without restoring context. A real commercial warning usually shows up as a broken trend, a lower sustained level, or a deterioration visible across channels. A seasonal dip, by contrast, tends to recur at the same time and in a fairly recognizable shape.

  • Always compare with the same month last year, not only with the previous month.
  • Look at the 6-to-12-month trend, not a single point.
  • Separate revenue into price, volume, and calendar effects.

A practical four-filter method: calendar, season, trend, commercial proof

First filter: calendar. At minimum, correct for working days, holidays, and exceptional closures. Insee notes that holidays and working-day effects change monthly sales. Eurostat explicitly distinguishes calendar adjustment from seasonal adjustment; a seasonally adjusted series is also calendar-adjusted, but not the other way around.

Second filter: season. Compare the current month with the same month over three years, then use a median rather than a simple average if your activity is noisy. Third filter: trend. If the last three months sit below their usual level and the underlying slope is negative, the signal deserves attention. The NIST reminds us that a series can combine trend and seasonality; in that case, season must be built into the model, or the movement is misread.

Fourth filter: commercial proof. Check whether the decline also appears in average basket size, order count, conversion rate, qualified traffic, won quotes, and follow-up activity. A decline that is truly commercial rarely survives scrutiny at every layer. Banque de France shows in its work that one should distinguish observed shocks from a counterfactual and take account of the activity’s own seasonality.

  • Calendar corrected?
  • Recurring seasonal effect?
  • Underlying trend broken?
  • Do several sales indicators confirm the drop?

Fictional numerical example: don’t confuse a soft patch with a structural break

Imagine a B2B services SME. Revenue is €120k in June, €92k in July, and €78k in August. Seen in isolation, August looks alarming. But the history shows that July-August are usually weaker, with an average drop of 22% from June to July and 15% from July to August. This year, the decline is 23% and 15%. The level is low, but the pattern matches the season. That is seasonal variation, not yet a structural decline.

Now change one thing: in September, activity does not rebound. It stays at €79k, while the usual return is around €105k. If, in addition, won quotes fall by 18%, qualified traffic is flat, and conversion slips by 2 points, we are no longer in weather-report territory. You are probably facing a real commercial decline, or at least a pipeline problem.

  • One weak month is not proof.
  • A weak month followed by a missing rebound is far more informative.
  • The real signal shows up in repetition and indicator coherence.

Method limits: when history is too short or the market changes too fast

Seasonal adjustment is not magic. Insee notes that seasonal profiles evolve when new promotional events appear or rules change, and that correction techniques only absorb those changes gradually. Eurostat also indicates that adjustments rely on models and revisions, so they are estimates, not carved-in-stone truth.

In plain English: if you have less than 24 months of history, if your offer changes sharply, or if a competitor enters the market, your “normal season” may be obsolete. In that case, favor simple, explicit, recent comparisons, then add qualitative reading: lost key accounts, stock, response times, pricing policy, or promotion effects. No statistic can replace a solid end-of-month conversation with the sales team.

  • Short history: maximum caution.
  • Offer change: reset your benchmarks.
  • New market context: combine data with field feedback.

A 7-day action plan: decide quickly without deciding blind

Day 1: export 24 months of monthly sales. Owner: finance control. Observable result: one clean table with revenue, orders, basket, and margin by month.

Day 2: add working days, holidays, closures, and promotions. Owner: finance or data. Observable result: atypical months are flagged.

Day 3: compare each month with the median of the same month over three years. Owner: finance control. Observable result: a “deviation from normal season” column.

  • Day 4 — Owner: sales director. Observable result: conversion rate and won quotes added.
  • Day 5 — Owner: marketing. Observable result: qualified traffic and lead source matched against sales.
  • Day 6 — Owner: CEO. Observable result: a written decision: likely seasonality, likely alert, or mixed.
  • Day 7 — Owner: CEO + sales. Observable result: one priority action and a 30-day follow-up indicator.