Question 1

How do you forecast demand and handle seasonality?

Accepted Answer

By separating what is **repeating**, what is **trending** and what is **noise** — because the commonest forecasting error is reacting to normal variation as though it were a change in direction.

**The three components of most demand series:**

**Trend**, the underlying direction over time.

**Seasonality**, repeating patterns tied to the calendar — annual, monthly, weekly and daily. **Weekly seasonality is the most underestimated**: many businesses have strong day-of-week patterns and compare Monday to Sunday without adjusting.

**Residual noise**, which is random and should not be explained.

**How to separate them practically:** compare **like periods** — this March against last March rather than against February; use rolling averages over a full seasonal cycle to see trend without seasonal wobble; and calculate seasonal indices, so you know that December is typically 140% of an average month and can judge whether this December is good.

**Why year-on-year comparison is the workhorse.** It controls for seasonality automatically, at the cost of being slow to detect change and sensitive to whatever happened in the base period — **a distorted base year produces misleading comparisons for a full year afterwards**, which is the trap after any unusual event.

**What else shifts demand:** moving holidays, which fall in different months between years; the number of trading days and weekends in a month; weather, which dominates some categories; pay dates; promotions, which **borrow** demand from the following period rather than creating it, and which must be accounted for or you will forecast a decline that is simply payback; competitor activity; and one-off events.

**Methods, in ascending complexity:** naive forecasts such as "the same as last year plus growth", which are a legitimate baseline and surprisingly hard to beat; exponential smoothing with seasonality; regression including known drivers; and machine-learning approaches, which require substantial history and frequently do not outperform simpler methods.

**Always compare against the naive baseline.** A sophisticated model that cannot beat "last year plus 5%" is not adding value, and this comparison is skipped remarkably often.

**Forecast ranges, not points**, and state the assumptions.

Question 2

Why do marketing tests so often produce misleading results?

Accepted Answer

Because most tests are underpowered, stopped early, or measuring the wrong thing — and each of those produces a confident number that does not survive contact with reality. Stopping early is the

Question 3

Are trade shows and events worth it?

Accepted Answer

Sometimes, and rarely for the reason budgets are justified — the value is almost never in leads scanned at the stand, and almost always in conversations that would not otherwise happen. Where the

Question 4

What is marketing mix modelling and why is it coming back?

Accepted Answer

Marketing mix modelling (MMM) is a statistical technique estimating how different marketing activities contribute to sales, using aggregate historical data rather than tracking individuals. How it

Question 5

What is a media plan, and what does flighting mean?

Accepted Answer

A media plan sets out where advertising will run, when, for how much, and to whom. Flighting is one of the scheduling patterns it can use — concentrating spend into bursts rather than spreading it

Question 6

What is churn and why does it matter more than growth?

Accepted Answer

Churn is the rate at which customers stop paying you over a period. It is the counterweight to acquisition, and it determines whether growth compounds or leaks away. Two measures, and the difference

Question 7

What are the best practices for Google Ads campaigns?

Accepted Answer

Implementing best practices for Google Ads campaigns, such as keyword research and ad copy optimization, is essential for maximizing ROI and improving performance.

Related Questions