Marketing Statistics
I learned the hard way that marketing statistics are useless unless you choose the right ones and check their significance before acting.
Start here
- Define one KPI per campaign before you collect any data.
- Set up a Google Analytics dashboard that shows conversion rate and revenue per visit.
- Always calculate confidence intervals for samples under 100 before reporting a result.
- Use median instead of mean when your data has outliers from viral spikes or seasonal drops.
- Compare every number against a benchmark from your own past performance or industry averages.
Plain-English take
If you are tracking everything and understanding nothing, marketing statistics are where you fix that. I used to think collecting any number I could find counted as insight. It does not. Statistics are just numbers that describe what is happening in your marketing, but only if you pick the right gauges.
Take descriptive statistics. Your mean time on page is 2 minutes 30 seconds. That sounds healthy. But the median might be 1 minute — half your audience bails early. The range tells you the spread: from 10 seconds to 10 minutes. Each number shows a different story. I use these to [benchmark seo](/benchmark-seo/) performance against past campaigns and competitors.
Correlation and regression are where most people go wrong. You see a 0.8 correlation between blog posts and sales. Does that mean publishing more posts causes sales? Not necessarily. Maybe both are driven by a third factor like a seasonal trend. I learned to run a regression first and check the p-value before making budget decisions.
Marketing statistics are not the truth. They are a language for what your data says. Learn the grammar — mean, median, standard deviation, significance — before you speak to a client.
When it actually matters
Before a campaign, market statistics tell you which audience to target and which channel to use. A client in Bangkok real estate wanted to spend on print. I looked at market data showing 70% of buyers are under 35. Print reaches 45-plus. We switched to Instagram and the cost per lead dropped from £15 to £4.
During a campaign, live metrics like conversion rate and click-through rate keep you honest. I set up a [google analytics](/google-analytics/) dashboard that refreshes every hour. When conversion rate dropped below 2%, I paused ads and tweaked the landing page. That single action saved £1,200 in wasted spend that week.
After a campaign, calculate ROI properly. One campaign cost £5,000 and generated 30 leads with a 10% close rate. That is three customers at £1,667 each. I compare that to [seo roi](/seo-roi/) from the same period to decide where to invest next month.
Edge case: small samples lie. Five conversions from 100 visitors is 5%, but if it rained during your promotion, the data might not repeat. I now require at least 100 events in a segment before I report a percentage. Without that rule, you are guessing.
When reporting to a director, they want numbers that prove impact. I use [seo reporting](/seo-reporting/) to show month-over-month trends in organic sessions and revenue. Statistics without context are just noise.
What I got wrong
First, I confused statistics with analytics. I spent a year collecting page views without asking why they moved. Analytics is the 'why' — statistics is the 'what'. Now I never look at a number without a hypothesis. If time on page drops, I ask: did the page load slower? Did the content change?
Second, I treated one result as universal truth. A guest post strategy doubled traffic for a client in finance. I assumed it would work for every client. It bombed for a hospitality client because their audience does not read long-form content. Now I treat every result as a hypothesis to test again.
Third, I ignored sample size. I once reported a 50% increase in email open rates based on 20 sends. The sample was too small. A follow-up with 500 subscribers showed 12%. I now calculate confidence intervals before presenting anything to a client. [Seo metrics](/seo-metrics/) like conversion rate mean nothing if your sample is 30 visitors.
My biggest mistake? Trusting a trend line without checking the r-squared. A straight line through three data points is not a trend. It is a wish.
Next step
Quick answers
What is the difference between a metric and a statistic?
A metric is a specific measurement you track, like conversion rate or cost per click. A statistic is a calculation applied to data, such as the average or median of those metrics. In practice, marketers use statistics to summarise metrics over time or across segments.
How many data points do I need before a number is reliable?
It depends on your goal. For a conversion rate, I want at least 100 conversions in the period before I report it. For a correlation, a sample size below 30 makes the result unreliable. Use a sample size calculator or a rule of thumb: 100 events per segment you plan to compare.
Should I use mean or median for marketing data?
Use median when your data has outliers — a viral spike that skews average traffic, or a single big customer that distorts average revenue per user. The median gives you the middle value, so it better represents typical behaviour. I use median for session duration and revenue per user, and mean for things like page load time where outliers are rare.
Sources
Primary documentation is linked directly. Anything commercial is marked nofollow.
- Google Search Central — Backs the claim that search performance data must be interpreted with statistical rigour.
- Optimizely — Supports the explanation of descriptive statistics and their role in summarising marketing data.
- SAS — Provides the distinction between statistics as raw data and analytics as interpretation.
- Quirk's — Confirms the definition of market statistics including size, demographics, and purchasing patterns.
Notes from Callum Bennett.