Campaign tracking usually starts after the campaign, which is why so many results are unreadable. Proper marketing campaign tracking is set up before launch, because the most important measurement is the one taken while nothing is happening.
The baseline is the whole thing
Record the preceding four weeks before anything launches. Without it, every post-campaign figure is a number without a comparison, and it can be presented as a success or a failure with equal ease.
Baselines also expose seasonality. A campaign in a naturally busy period will look effective whatever it does, and only a comparison against the same period previously will show whether it actually contributed.
Setting up tracking
- Record the baseline period before launch.
- Agree which metrics will be judged, in advance.
- Use consistent naming so results can be found later.
- Separate paid from organic from the start.
- Note anything else happening that could affect the figures.
- Decide the measurement window before seeing any data.
- Write down what would count as success, and what would not.
Deciding the window in advance prevents the common failure of extending measurement until the numbers look better, which is a decision people make without realising they are making it.
Be honest about attribution
Social activity rarely produces cleanly attributable outcomes. People see something, do nothing, and act later through a different route entirely. Any model claiming precise attribution is making assumptions, and those assumptions deserve to be stated rather than hidden inside a figure.
Report what you can observe and label what you inferred. Confident attribution that later proves wrong costs more credibility than an honest "this moved, and here is what else was happening at the time". Review structure sits in the campaign performance review workflow, and targets in the KPI planning guide.
Common questions
How long should the baseline be?
At least as long as the campaign, ideally longer, so normal variation is visible.
What if several campaigns overlap?
Attribution becomes unreliable. Either stagger them or accept that the results are combined and say so.
Should everything be tracked?
Track what could change a decision. Additional metrics mostly add noise and reporting effort.
How soon can results be judged?
After the agreed window, not before. Early readings mostly capture novelty.
What if results are inconclusive?
Report that plainly. Inconclusive is a legitimate and common outcome, and inventing a conclusion is worse.
Should the criteria be written down?
Always, before launch. Success defined afterwards is defined to match the result. See contact.
Should tracking be shared with the client?
Share the criteria before launch. A client who agreed what success looks like beforehand reads the result very differently from one shown a conclusion afterwards.