Ellis Agency. ANALYTICS · TRAINING GUIDE
GOOGLE ANALYTICS 4 · THE BASICS

Reading your own
traffic.

A working tour of Guilford's Google Analytics — what the plumbing does, what the words mean, which five screens answer real questions, and a guided lap you can do in ten minutes. Every number in this guide is your number, pulled from your own property.

PROPERTY359308554
STREAMG-09SSBH3WCE
CONTAINERGTM-WZHR49S4
FORMarketing & Comms
Ellis Agency.How the plumbing works
A

How the plumbing works

Two different products, and almost everyone mixes them up. Tag Manager decides what gets sent. Analytics decides what you can see. If a number is missing, the answer is almost always that nothing is sending it yet — not that the report is broken.

WhatIDThink of it asWhat it actually does
GA4 property359308554The warehouse.Where all the data lands and where you build reports. One property covers guilford.edu.
Measurement IDG-09SSBH3WCEThe address on the parcel.The web data stream inside that property. Anything tagged with this ID reports into 359308554.
GTM containerGTM-WZHR49S4The delivery van.Google Tag Manager. It sits on every page, watches what visitors do, and sends the events to GA4.
RetiredGTM-5BZ4942 · UA-9584163-1The old van, and a dead one.The previous container plus a Universal Analytics tag that stopped processing in 2023. Both removed.

The one thing to understand before you open a report

Guilford's four most valuable actions do not happen on guilford.edu. Applying goes to Slate (apply.guilford.edu). Requesting information and booking a visit go to Slate. Giving goes to giving.guilford.edu / GiveCampus. Your website's job is to get someone to those systems — so what GA4 measures on your site is intent, and the finish line lives somewhere else until those systems are tagged too.

WHY YOUR GIVING NUMBERS ARE ZERO

They are not broken and they are not hiding. In the 90-day baseline there were 7,044 conversions and every one of them was enrollment-flavored — applications, information requests, email and phone clicks. Giving was not tracked at all. Standing that up is the point of the external-systems rollout, not something you can fix inside a report.

Two habits that will save you an argument later

  1. Data is not retroactive. GA4 only knows what was being sent at the time. Turning on a new event today gives you data from today forward — never backfilled. This is why we tagged before the redesign launched rather than after.
  2. Give it a day. Standard reports settle over roughly 24–48 hours. A number you check at 4pm will not match the same number tomorrow morning, and that is normal, not a bug.
Guilford College · Using GA4Ellis Agency · Aug 202602 / 09
Ellis Agency.The words GA4 uses
B

The words, and what they cost you if you get them wrong

Nearly every reporting mistake is a vocabulary mistake. Learn these eight and most of GA4 stops being mysterious. The right-hand column is Guilford's own 90-day baseline, so you have a feel for what normal looks like on your site.

TermWhat it meansYours, in 90 days
UserOne person — really, one browser on one device.78,234 in 90 days. The same person on a phone and a laptop counts twice.
SessionOne visit. Ends after 30 minutes of inactivity, then a new one starts.120,036 — so the average person came back about 1.5 times.
EventAnything that happened. In GA4 everything is an event, including a page view.A page view, a click on Apply, a form submit, a search, a file download.
Key eventAn event you have flagged as mattering. (Google renamed "conversion" to this.)7,044 in 90 days. You choose which events get the flag — it is a switch, not a calculation.
Engaged sessionA visit that lasted 10+ seconds, or saw 2+ pages, or fired a key event.The bar is genuinely low. It is the opposite of a bounce.
Engagement rateEngaged sessions ÷ all sessions.45.0%. Our post-launch target is ≥ 47%.
DimensionA what or where — always text you can group by.Page path, channel, device, country, event name.
MetricA how many — always a number.Sessions, users, event count, engagement rate, average duration.
THE TRAP: EVERY ROW IS "OF SOMETHING"

A dimension and a metric only mean something together. 45% engagement rate is not a fact about Guilford — it is the average of every page, device and channel mixed together. Split it by device and desktop and mobile behave differently. Split it by channel and Direct behaves nothing like Display. The number at the top of a report is the least interesting version of that number.

A worked example, in your data

Sessions are 120,036. Split by channel, that is 51.8% Organic Search, 31.2% Direct, 5.1% Display, 4.0% Referral, 3.1% Organic Social, 2.6% Paid Other. Split by device it is 63.9% desktop and 35.3% mobile. Now the same total tells you three different stories — a search-led site, with a very large group of people typing you in from memory, mostly on a laptop. That last group is largely current students and staff, which is why /current-students (15,946 views) sits second only to the homepage.

Guilford College · Using GA4Ellis Agency · Aug 202603 / 09
Ellis Agency.The five screens
C

Five screens answer almost every question

GA4's menu is large and most of it you will never open. These five carry the work. Each one below is written as the question it answers, because that is how you will actually go looking for it.

1

Is the thing I just published actually working?

Reports → Realtime

The last 30 minutes. Use it to prove a tag fires or a page is live — never to judge performance. Open your site in another tab, click something, and watch yourself appear.

2

Where did these people come from?

Reports → Acquisition → Traffic acquisition

Sessions by channel. This is where you see Organic 51.8% and Direct 31.2%, and where paid campaigns show up. Set the dimension to Session source / medium for detail — that is where encoura / display (6,113) lives.

3

What are they actually reading?

Reports → Engagement → Pages and screens

Your top-pages list. / 44,537 · /current-students 15,946 · /faculty-and-staff 8,193 · /admissions/apply 7,840 · /program-finder 6,179. Sort by engagement rate rather than views to find pages punching above their weight.

4

What did they do?

Reports → Engagement → Events

Every event by name and count. This is where Guilford's own events live — cta_click, apply_start, form_submit, site_search. Click any event name to see its parameters.

5

On what, and from where?

Reports → Tech / User → Demographics

Device, browser, screen size, city. Desktop 63.9% / mobile 35.3% / tablet 1.1%. Useful when a page looks fine to you and broken to a third of your audience.

The date picker is the most important control on the screen. It sits top-right and it defaults to the last 28 days. Change it once and it follows you from report to report. Tick Compare and pick "preceding period" or "same period last year" — on an admissions site, year-over-year is nearly always the honest comparison, because your traffic has a season and last month does not.

Guilford College · Using GA4Ellis Agency · Aug 202604 / 09
Ellis Agency.A guided lap
D

A guided lap — do this one together

Ten minutes, start to finish, in your own property. Nothing here changes any setting, so nobody can break anything. The point is to end up somewhere you did not already know.

  1. Open the property and check you are in the right one. Top-left selector should read the account, then property 359308554. If you see a different number, every figure in this guide will disagree with your screen.
  2. Set the date range to the last 28 days and turn on Compare → preceding period. Everything from here on shows a percentage change next to it. That change is the story; the raw number rarely is.
  3. Reports → Realtime. Open guilford.edu in another tab and click an Apply button. Find yourself in the event list. You have just watched the container fire — that is the whole pipeline in one motion.
  4. Reports → Acquisition → Traffic acquisition. Read the channel mix. Ask out loud: which of these did we pay for, and which did we earn? Then switch the dimension to Session source / medium and find encoura / display.
  5. Reports → Engagement → Pages and screens. Sort by views. Then sort by engagement rate instead. The pages that move up are the ones doing more work than their traffic suggests — that is a content decision you can act on this week.
  6. Reports → Engagement → Events. Find cta_click. Click it. Scroll to the parameter cards and look at cta_type — apply, give, visit, rfi, learn. You are now looking at which calls to action people actually press.
  7. Search for a page. In Pages and screens use the search box above the table and type /visit. You have just filtered the whole report to one section. This is how you answer "how is my page doing?" without building anything.
  8. Change the date range to last year, same period. Watch the numbers move. Admissions traffic has a shape — this is you meeting it for the first time.
IF SOMETHING LOOKS WRONG

Check the date range first — it is the cause about eight times in ten. Then check whether the event existed that far back. Then check whether you are looking at users where you meant sessions. Only after all three is it worth calling it a problem.

Guilford College · Using GA4Ellis Agency · Aug 202605 / 09
Ellis Agency.Guilford's own events
E

The events we built for you

GA4 gives every site the same generic events. These are the ones written specifically for Guilford, fired from the container on every template. They are the difference between "3,000 people clicked something" and "3,000 people clicked Apply, from the hero, on mobile."

EventFires whenThe question it answers
cta_clickEvery call to action on the site, classified.Which CTA earns the click — and where on the page it was.
apply_startSomeone left for the application portal.The top of the enrollment funnel. Replaces three overlapping legacy events.
form_start / form_submitA form was begun, and finished.Carries form_type, so RFI is finally separable from every other form.
site_searchSomeone used the site search.What people cannot find. Read the search terms weekly — it is the cheapest content research there is.
card_clickA card in any grid was clicked.Which stories, programs and profiles pull.
program_interestA program page or the program finder was engaged.The leading indicator for applications, by program.
filter_use / load_moreA listing was filtered or paged.Whether your listings are usable or people give up.
donation_start / donation_completeThe giving flow began and finished.Built and waiting — it turns on when GiveCampus is tagged.

The eleven labels that make them useful

Each event carries parameters — registered in GA4 as custom dimensions, which means you can group and filter by them like any built-in field. This is where the real answers live.

cta_typetemplatepage_typepositionform_typecard_typeapply_levelaudience_prioritylabelgift_typegift_designation
WHAT THIS REPLACED

The old setup fired three different events that all meant "someone tried to apply" — guilford_application_button (1,889), guilford_app_click_to_apply_site (1,068) and application_page_visit (183) — which overlap and cannot be added together. Forms were worse: 15,767 starts and 12,708 submits, with no way to tell an information request from a help ticket. One clean event per action, with labels, is what makes a funnel possible at all.

Give new parameters 24–48 hours. A newly registered custom dimension takes a day or two to appear in the field list. If one is missing, that is nearly always the lag rather than a mistake.

Guilford College · Using GA4Ellis Agency · Aug 202606 / 09
Ellis Agency.Not lying to yourself
F

How not to lie to yourself with this data

Analytics rarely gets caught being wrong. It gets caught being read wrong. These are the five that will come up in your first month.

1

Consent changes the denominator

The site runs a cookie consent banner with Google Consent Mode. Visitors who decline are measured in a reduced, modelled way. If consent rates move — a banner redesign, a new regulation — your traffic will appear to move with it. Always ask whether the site changed or the measurement changed.

2

GA4 will never match Slate

GA4 counts browsers; Slate counts applicants. One person applying from a phone and a laptop is two users and one application. GA4 tells you how many started; Slate tells you how many finished. When the two disagree, Slate is right about applications and GA4 is right about intent. Neither is broken.

3

"(not set)" and "Direct" are the same confession

Both mean GA4 could not tell. Direct is 31.2% of your sessions — some of it is genuinely people typing you in, and some is an untagged link, an email client, or a QR code. Treat a rising Direct share as a question about link tagging before you treat it as brand strength.

4

A cross-domain hop can fragment one visit into two

When someone crosses to the apply portal, that hop has to be configured or GA4 records a new session referred by your own systems. You could see this in the baseline: portal.mist.com appeared as a referral source with 1,834 sessions — your own applicants, counted as strangers arriving from outside.

5

Annotate anything that moves the line

A launch, a tag change, an email send, a snow day. Six months from now nobody will remember why the third week of March spiked, and someone will build a theory on it. GA4 has annotations — use them the day it happens, not the day you are asked.

Guilford College · Using GA4Ellis Agency · Aug 202607 / 09
Ellis Agency.Funnels and what comes next
G

The one advanced thing worth learning

Standard reports tell you what happened. Explore tells you where people stopped — and for an admissions site that is the whole game. It is one menu item down the left, and a funnel takes about three minutes to build.

  1. Explore → Funnel exploration. Start blank rather than from a template.
  2. Add your steps down the right-hand panel — for the enrollment funnel: a program or hub page view, then cta_click where cta_type = apply, then apply_start.
  3. Set it to open funnel so people can join partway through. Closed funnels flatter you by hiding everyone who did not start at the top.
  4. Drop a breakdown on it — device, or template. This is the moment you find out that mobile drops out somewhere desktop does not.
  5. Right-click any step → view users. That is your remarketing audience: everyone who started an application and never finished.

Where this is heading

NextWhat it gives you
Looker Studio dashboardThe same data as a one-page read, organised by your four audience priorities — enrollment, giving, reach, wayfinding — so nobody has to learn GA4 to get the answer.
Scheduled emailWeekly to marketing, daily to us. The report finds you rather than the other way round.
External systems taggedSlate and GiveCampus carrying the same container — the point at which apply_submit and donation_complete start reporting, and the funnel finally has a bottom.
Post-launch comparisonThe 90-day baseline in this guide is the "before". Annotate the launch date and the before/after is honest by construction.
THE HONEST STATE OF PLAY

Enrollment intent is well measured today. Giving is not measured at all, and application submits are inferred rather than confirmed, because both finish on systems that are not yet tagged. Anyone who tells you the giving number is zero is reading the tracking, not the donations.

Guilford College · Using GA4Ellis Agency · Aug 202608 / 09
Ellis Agency.Your first month
H

Eight questions you can now answer without us

This is the real test of the training. Every one of these is answerable from the five screens in section C plus one funnel from section G — no new tagging, no engineering, no ticket.

  1. Which channel brought us the most engaged visitors last month, and is that different from last year?
  2. Which program pages get read but never lead to an apply click?
  3. What are people typing into site search, and are they finding it?
  4. Does the homepage Give CTA get pressed, and from which position on the page?
  5. Where do mobile visitors drop out of the application path that desktop visitors do not?
  6. Which news stories actually pull, and which just get published?
  7. Are current students and faculty still finding their pages as fast as before the redesign?
  8. Did last week's email send move anything, and for how long?

A sane weekly habit

  1. Monday, ten minutes. Traffic acquisition with compare-to-last-year on. Note anything that moved more than about 10% and write down why you think it moved.
  2. Read the search terms. Engagement → Events → site_search. It is the shortest route to knowing what your site fails to say.
  3. Check the top of the funnel. cta_click where cta_type = apply, and apply_start. Those two are your leading indicators, weeks ahead of Slate.
  4. Annotate anything you did. Sends, launches, press, outages. Thirty seconds now, hours saved later.

A closing word on judgement. Analytics is very good at telling you what people did and completely silent on why. Every number in here is the beginning of a conversation with someone who knows the audience — it is not the end of one. The colleges that get value from this are the ones that let the data pick the question, and let people answer it.

Ellis Agency. GUILFORD COLLEGE · GA4 TRAINING · AUGUST 2026 · EA-2026-0002
Guilford College · Using GA4Ellis Agency · Aug 202609 / 09