For one stretch last month, more than half of the sessions on our own website came from "Direct" traffic, and fewer than one in five of those sessions showed any engagement at all. Read at face value, Direct was our biggest channel. In reality it was mostly bots, and the visitors who mattered were hiding in two smaller rows further down the report.
Spotting that took about a minute. Not spotting it would have pointed every decision that month in the wrong direction. That gap, between reading a report and believing it, is most of what separates an expert from someone who knows the tools.
Becoming a digital marketing expert is less about learning more channels than about learning to catch your own numbers lying, quickly, before you act on them.
Here is a concession first, because it matters. Platform certifications are worth doing. They teach the vocabulary, the settings and where the buttons are, and nobody can skip that. They will not teach you anything below, because everything below only shows up in a live account.
1. Judge a channel by its quality, not its volume
In that same 28-day window, visitors arriving from AI assistants such as ChatGPT were engaged in 61% of their sessions. Visitors from Google organic search were engaged in 31%. The two channels sent almost the same number of sessions. One was twice as engaged as the other per visit, and neither looked like much next to the inflated Direct row.
The skill is to stop reading the sessions column first. Engagement rate, events per session and conversions per channel tell you where attention is real. The trade-off is that quality metrics are noisy at small volumes, so you hold the conclusion loosely until the numbers grow.
2. Test the conversion path, not the page
A landing page can load in a third of a second, return a healthy status code and pass every uptime check while the form on it is broken. Uptime monitors request the page; they do not submit the form. If the form depends on a database, a CRM or an email service, any one of those can fail while the page itself looks perfect, and nobody submits briefs to their own site.
An expert tests the path a customer takes, not the page they land on: fill in the form, place the test order, click the email link, on a schedule. Build forms so a failure downstream degrades rather than drops the lead, for example by emailing it when the database is unreachable. A missing database row is recoverable. A lost lead is not.
3. Treat forms and email as infrastructure
An automatic "thanks, we got your brief" email is a small courtesy with a known weakness. Bots can submit a form using real strangers' email addresses, and every submission then sends a message from your domain to someone who never contacted you. Enough of those and your sending reputation suffers, which puts every email from that domain at risk, including the ones to clients.
Nothing about that looks like a marketing problem until it is one. Lead forms, auto-replies and sending domains are part of the marketing system, so filter submissions before any automatic email goes out, and watch the sending domain's reputation as closely as you watch conversion rate.
4. Make filters fail loudly
A common spam defense is a hidden field that only bots fill in. It has a quiet failure mode: if the field has a meaningful name, such as "company website", browser autofill can complete it for real people, and a filter that deletes anything with that field filled will throw away genuine enquiries without a trace.
Give hidden fields meaningless names, flag suspicious submissions instead of deleting them, and drop only the ones that are unmistakably machine-made. The general lesson matters more than the specific fix. Any rule that silently discards data, whether a spam filter, a tracking exclusion or a deduplication step, needs a way to show you what it threw away.
5. Read the reason, not the total
After we fixed how the www version of our domain redirected, Search Console's "not indexed" count jumped. On a dashboard that looks like a sudden problem. The breakdown showed the opposite: the new entries were redirecting duplicate addresses that Google was correctly declining to index, which was the fix working.
Every aggregate number hides a mix of causes. The expert habit is to open the breakdown before reacting to the total, in Search Console, in GA4 and in every ad platform.
6. Match the page to the search, not the keyword
The costliest version of this is a redirect. When a page that ranks is moved to a new address that shares its name but does a different job, Google re-ranks it for the job it now does, and the position can halve with every technical check still passing. We have written that pattern up in full in our SEO strategy piece, because it changes how every URL change should be planned.
What this does not cover
This is not a curriculum. It will not teach you to build a campaign, write ad copy, set up a feed or run a bid strategy, and those skills are real and worth learning properly. It also assumes you have data to read. On a brand-new account, the skill is patience until there is enough signal to catch anything at all.
It is, though, the part that is hardest to get from a course: we hire for it in senior people, and it is how we approach the digital marketing retainers we manage for clients, and the reason behind the free tools we publish.
Questions to test yourself with
- What share of your sessions last month showed no engagement, and do you know where they came from?
- When did someone last submit your own lead form, end to end, and check every place the lead should arrive?
- Which of your filters, exclusions or automations can silently discard a real customer?
- The last time a total moved sharply, did you open the breakdown before you acted?
If this is how you already work and you want to do it alongside a senior team, our jobs page is the place to start.