Reading the traffic sources honestly
Direct is not brand recognition, it is everything the tool could not identify. And the useful view is landing pages sorted by outcome, not by visits.
The traffic sources report is the most useful and most misread screen in analytics. Misread because the categories are not what their names suggest, and because the largest one is usually a bucket for everything the tool could not identify.
What the categories actually mean
- Organic search: somebody arrived from a search engine's unpaid results. Reasonably reliable.
- Direct: the tool could not tell. It includes people typing your address, and also links from documents, apps, messaging, email clients and anything that stripped the referrer. Treating it as brand recognition is optimistic.
- Referral: another website linked to you. Reliable and often the most interesting, because it tells you who is talking about you.
- Social: a platform, when the platform passed the information, which it often does not.
- Paid: only correct if your campaign parameters are right, which is why they matter.
- Email: only correct if you tagged your links, otherwise it is in direct.
The practical consequence: a large direct number is usually a measurement gap rather than a finding, and the fix is tagging rather than interpretation.
The question to ask of the report
Not where traffic came from. Which sources produced the outcome. A source sending many visits and no orders and a source sending few visits and several are both important findings, and the total-visits view shows neither.
Illustratively: one channel sends 2,000 visits and 4 orders, another sends 200 and 6. The first dominates every traffic chart and the second is the business. Looked at by volume you would invest in the wrong one, and most dashboards default to volume.
Landing pages, which is the more useful view
People arrive on a page, not on a site. Sorting by landing page and outcome tells you which specific pages do the work, and it is usually a short list with a long tail of pages that bring nothing.
- Sort landing pages by outcome rather than by visits, and look at the top ten.
- For each, check where its traffic comes from. A page that works for search traffic and not for social traffic is telling you something about both.
- Then look at the pages with traffic and no outcome. Either they are answering a question with no commercial intent, which is fine and should be linked onward, or they are failing at something fixable.
- Write down the two pages that matter most. On most small sites, a handful of pages produce nearly all the result, and knowing which is the whole point of the report.
The referral report is the one to read for pleasure
Referrals are a small number of visits and the most informative screen in the tool, because each line is a human being who linked to you somewhere.
- Open each referring domain and find the actual link. What they said about you is free positioning research, and it is frequently better phrased than your own website.
- Note which of your pages they linked to. The pages other people choose to cite are the ones worth developing.
- Look for forums, communities and question sites. A link from a discussion means somebody recommended you in answer to a question, which tells you the question.
- Thank the ones that matter, or at least notice them. A person who linked once will often link again if they know you exist.
Two things to filter out first. Referral spam, which is traffic generated purely so the domain appears in your report and should be ignored rather than investigated. And your own domains, including a separate shop or a staging site, which register as referrals from yourself and distort everything.
This report will never be a large share of your traffic, and it is the one that most often changes what you write next.
Work to do
- Check the size of your direct bucket. If it is large, tag your email and social links before drawing any conclusion from it.
- Open your referral report and read what three of those sites actually said about you.
- Sort landing pages by outcome, not visits, and write down the top ten.
- Find one source with high volume and low outcome, and decide whether to fix it or stop feeding it.