Every advertising platform reports its own success, and each one has an incentive to claim as much credit as possible. Measuring marketing honestly means starting from the business outcome, keeping your own records, and treating any single report with polite suspicion.
Start from the outcome, not the channel
Write down what a successful month looks like in business terms: qualified enquiries, booked calls, signed proposals, orders. Then work backwards to the marketing measures that plausibly lead there. Clicks and impressions belong on the list as diagnostics, not as goals.
Key takeaway
If a number can rise while the business stands still, it’s a diagnostic, not a goal.
Record a baseline before you change anything
Before a new campaign, note the last few months of your outcome measures, by week, along with anything unusual: holidays, price changes, a mention in the press. Seasonality and one-off events explain more swings than most creative changes do, and without a baseline you can’t tell them apart.
Label every link you control
UTM parameters are small labels added to a link so your analytics can tell where a visit came from. Use them on every link you place: emails, social posts, partner sites, QR codes on print.
| Parameter | Answers | Example |
|---|---|---|
utm_source | Who sent the visit? | newsletter, linkedin |
utm_medium | What kind of channel? | email, social, print |
utm_campaign | Which campaign or offer? | autumn-audit |
Consistency matters more than detail. “LinkedIn”, “linkedin” and “li” will appear as three separate sources in most reports.
Read attribution as a model, not a fact
Attribution assigns credit for a conversion to the touchpoints before it. Every model is a set of assumptions: last-click favours whatever came last, platform reports favour themselves, and data-driven models depend on how much data you have and what they are allowed to see. Privacy controls, blocked cookies and conversations that happen offline all leave gaps.
Attribution tells you where to look. It rarely tells you what caused what.
Two cheap habits help. Ask every new enquiry how they heard about you, in their own words, and record it next to the analytics source. And compare totals: if platforms together claim more conversions than you actually received, the overlap is the size of your uncertainty.
Test one thing at a time
The most trustworthy evidence comes from changing something on purpose and comparing against what would otherwise have happened.
- Write the prediction first. What should change, by roughly how much, and by when?
- Change one thing. A new offer, a new audience or a new landing page, not all three at once.
- Leave a comparison. A region, period or audience segment where nothing changed.
- Wait for enough data. A few days of results is usually noise, especially at small volumes.
- Record the result, including the tests that showed no effect. They stop the same idea being tried again next year.
Report what you can stand behind
A useful monthly report fits on a page: outcome measures against the baseline, what you changed, what you think happened and how confident you are, and what you’ll try next. It is less impressive than a dashboard with forty charts, and far more useful.
Related service
Market research and data analytics
Measurement plans, dashboards and analysis that support decisions.
