We said it back in Module 1: averages lie. A CPA of €40 can be €25 on desktop and €80 on mobile; a CTR of 4% can be 6% on Google and 1% on search partners. The control that breaks any table into its components is called Segment, and it is the most powerful analysis tool the interface has without leaving it.
How it works
The Segment icon (above the table) → you pick a dimension → each row splits into sub-rows, one per value of that dimension, with every column recalculated. It can be combined with filters and with period comparison, and it downloads segmented.
The segments people use most
| Segment | Splits by | Question it answers |
|---|---|---|
| Device | Computer / mobile / tablet / TV | Does mobile convert the same? Does TV deserve a −100%? |
| Network (with search partners) | Google / partners / Display | Are partners performing? Has Display expansion crept in? |
| Hour of day · Day of week | 24 hours · 7 days | Are there time slots that spend without converting? (mind the conversion delay) |
| Click type | Headline, sitelink, call, image... | Which part of the ad are people clicking? |
| Conversions → Conversion action | Each action separately | Of those 40 conversions, how many are forms and how many "see phone number"? |
| Top vs. other | Above the results / the rest | What is being at the top worth, in CTR and in CPC? |
| Search terms match type | Exact / phrase / broad of the real term | Which match type brings the conversions? |
| Time (day, week, month) | One row per period | Trends without opening charts |
Five typical findings
1. Mobile converts at half the rate and costs the same
Segment campaigns by device. If the mobile CPA is double the desktop one, the cause is almost always the page (slow, long form) — not the bids. Fix the landing page before touching settings.
2. Search partners
Segment by network. It is common to see partners with a very low CTR and a high CPA compared with Google. If that happens, untick the box (Module 2). If they perform just as well, leave them: they are extra traffic at the same price.
3. Those conversions were clicks on "see phone number"
Segment by conversion action. Discovering that 80% of your "conversions" are a secondary action someone set as primary is the most frequent finding in inherited accounts (Module 6).
4. The small hours are spending
Segment by hour of day. If between 1 and 6 am there are clicks with no conversions steadily for weeks (not one day), an ad schedule or a −100% adjustment for that slot saves money without losing anything. With two warnings: Smart Bidding already bids less there; and the conversion delay affects hours too.
5. Being at the top is worth it (or not)
Segment by "top vs. other". Normally the top has a much higher CTR and CPC; if the conversion rate is the same, the CPA tells you whether the position pays off. That is the figure that should decide an impression share strategy.
Careful with the statistics
When you segment, the data gets split and each slice has less volume. An hour with 3 clicks and 0 conversions is not a finding; 300 clicks and 0 conversions over two months is. Segment with long date ranges and only decide with volume (the 2-3 × CPA with no conversion rule from Module 3 applies here too).
💡 Ninja trick: segments are snapshots: you look at them today and tomorrow you have to look again. The Suite scripts make these cuts every night and keep the history — the Agent watches the split by device and network, Smart Bidding (SBNS) works with daily stats per campaign — and warn you when a segment drifts. You segment to learn; the script, to watch.
What you should remember
- Segmenting = breaking the average apart. Device, network, hour/day, conversion action and top vs. other are the top five.
- A high mobile CPA is usually the page, not the bid.
- "Conversions" segmented by action expose badly configured measurement.
- Only decide with volume; every segment has less data than the total.