Customer demand trends: spotting seasonal demand and shifting enquiries

Short answer

To see seasonal demand and shifting enquiries, track the share of conversation fields (topic, product of interest, objection) over time and compare every change against the right baseline: the same period last year as well as last week, weeks with and without campaigns, holidays and working days. Share, not absolute count, should be the main measure.

The biggest risk is a false trend: random variation in small numbers, analysis delay or one campaign's effect read as real change. Below are checking rules for reading customer demand trends.

Why conversation trends show up before sales trends

Customers ask before they buy. So a shift in demand can start in conversations days or weeks before it shows up in a sales report: questions about children's products before the school year, air-conditioner servicing before the heat, gift packs before a holiday. This is early warning for stock, staffing and advertising.

But it is not an automatic rule: sometimes more questions do not turn into sales because customers check the price and buy elsewhere. So track the conversation trend together with sales outcomes.

Share, not count

If all enquiries double in December, every topic's count rises too. That is a change in volume, not demand. The real signal is a change in share: if "gift wrapping" rises from 3% to 11%, the structure of customer interest has changed. Track count for volume planning and share for priority.

Choosing the right baseline

1. 1Previous periodComparing with last week or month shows short-term change but does not separate season.
2. 2Same period last yearThe best baseline for separating seasonal effects. Needs at least a year of history.
3. 3Same weekdaysWeekend and weekday question mixes differ. Compare Monday with Monday.
4. 4A campaign-free periodTo separate a campaign's effect, compare with a similar period without ads.

The campaign effect

When an ad starts, the enquiry mix changes: questions about the advertised product rise, a new audience asks different questions, and sometimes the ad's promise creates false expectations. That change is not seasonal demand. Mark campaign dates on the trend chart and look at conversations from the campaign channel separately with a cross-tab.

Five sources of false trends

• Small numbers: a value going from 6 to 12 has "doubled" but may be chance. Look for consistent change over at least 3–4 periods.
• Analysis delay: if the last days' conversations are not analysed yet, the last point drops.
• Field changes: if the value list changed and old conversations were not re-analysed, the trend breaks.
• Channel changes: when a new channel is connected, its audience shifts the mix.
• Calendar boundaries: if "August" is computed in another time zone rather than local time, edge days get mixed.

A checklist for reading a trend

1. 1Did the share change?Share, not count. Separate out volume changes.
2. 2Is it consistent?At least 3 periods in the same direction. A one-period jump is not yet a trend.
3. 3Is coverage complete?Do the periods being compared have similar analysis coverage?
4. 4Is the baseline right?Last year, same weekday, campaign-free period.
5. 5Are campaigns and events marked?Check against the event log.
6. 6Was the evidence read?Read 10 conversations for the rising value: are customers really asking the same thing?

Decisions that rest on trends

• Stock: which product to stock up on before the season.
• Staffing: in which weeks an extra shift or temporary staff is needed.
• Knowledge base and bot: preparing answers to seasonal questions in advance.
• Advertising: shifting budget before demand rises, or cutting ads if growth is happening anyway.

Illustrative example

This is an illustrative example. An air-conditioner sales and installation company sees the share of the "installation date" question rise for three weeks in a row in the second half of April. Last year's event log shows the same rise began in mid-May and coincided with the first heat.

Evidence conversations confirm customers are asking "can you install before June?". The company opens its installation crews' schedule two weeks earlier and starts showing the next available dates on its website.

A seasonal calendar template

Trends collected over a year should become a simple seasonal calendar, so next year's decisions are made in advance. Write one line per month:

• Month and week: the week demand started to shift.
• Rising topics and products: the two or three values whose share rose most, in percentage points.
• New questions: values that moved from "other" into the field.
• Campaigns and events: what happened in that period.
• Action taken and result: stock, staff, bot answers, ads — and whether it worked.

The calendar becomes a shared document for sales, marketing and operations and is reviewed at the start of each quarter.

Keeping fields stable

A trend only makes sense between periods measured the same way. When you change a field's definition or value list, record it in the log and re-analyse old conversations; otherwise a jump in the chart will reflect your change, not your customers. Not changing fields during a season is the safest course.

Trends and comparisons in Vexvon

In Vexvon the panel's AI assistant gives a trend by day, week or month and a two-period comparison for every field value; the result is shown as a line chart. Periods are computed by local calendar day in Baku time, and a fact is dated by when the customer said it, not when it was analysed — an objection raised on the last day of a month does not slip into the next. The report also shows the share of the period analysed. More: Vexvon analytics.

Next step

Start an event log today: each week, note campaigns, holidays, price changes and outages in one line. If you have not built a topic list yet, start with customer inquiry analysis. We can show the trend panel on your own data in a demo.