Measuring what actually works
Watch cost per result, conversion rate and return on spend, and ignore almost everything else. Clicks, impressions and follower counts move without telling you anything about money. And a number means nothing on its own — it needs last month, and the same month last year, to compare against.
The four numbers that matter
| Number | What it says | Watch for |
|---|---|---|
| Cost per result | What one enquiry or sale cost you | Rising while volume is flat |
| Conversion rate | How many visitors act | A drop after a website change |
| Return on ad spend | Revenue for each dollar spent | Only useful with real revenue values |
| Total enquiries | Whether the phone is ringing | The sanity check on all of the above |
If you track nothing else, track cost per result and total enquiries. Those two catch nearly every problem worth catching.
What to stop watching
- Impressions — you can buy a million of them and sell nothing.
- Click-through rate on its own — useful for diagnosing an ad, useless as a goal.
- Followers — slow, and only loosely related to revenue.
- Bounce rate — too easily misread to base decisions on.
A number needs something to sit against
"A$41 per enquiry" is not good or bad until you know it was A$33 last month and A$58 in the same month last year. Most tools keep a short window by default, so the comparison you want is often the one that has just expired.
- This period against the one before it — catches sudden change.
- This period against the same one last year — catches seasonality, which matters a great deal for trades and retail.
- Each account against the others — catches one account quietly carrying or dragging the total.
Atlas stores thirteen months of daily figures per account for exactly this reason, so the year-on-year comparison is there when you need it rather than gone.
A review routine that survives a busy week
- 01Daily, one minute: is spend pacing where it should be, and has anything stopped?
- 02Weekly, ten minutes: read the search terms, add negatives, note anything that converted.
- 03Monthly, half an hour: cost per result against last month and last year, then one deliberate change.
- 04Write the change down with the date, so next month you can tell whether it helped.
Record the changes, not just the results
The most common reason a business cannot say what worked is that nobody wrote down what they did. Keep a log — what changed, when, and by whom — and the next comparison becomes an answer instead of a guess. Atlas keeps this automatically for every edit it makes to an ad account, and marks each one better, worse or too early to tell.
Common questions
- What is a good cost per lead?
- One you can pay and still make money. Compare it with the profit on a job and your close rate, not with someone else's benchmark.
- How often should I check my ads?
- A minute a day for anything broken, ten minutes a week on search terms, half an hour a month for decisions. Daily tinkering usually makes results worse.
- How much history do I need?
- At least thirteen months, so you can always compare this month with the same month a year ago.
- Is return on ad spend better than cost per lead?
- Only if you record real revenue per sale. Without that, cost per lead is the more honest number.