DEV Community: Robert Breen

Yoya: Stop Spending the First Ten Minutes of Every Intro Call Finding Out What the Client Wants

Robert Breen — Sat, 31 Oct 2026 18:35:14 +0000

Most booking links hand you two facts: a name and a time.

So every intro call opens the same way. "Tell me a bit about what you're looking for." Ten minutes go by before you know what the client wants, and some of those calls turn out not to be a fit. If you take five intro calls a week, that's most of an hour spent on questions you could have had answered in advance. It's also ten minutes you could have spent sounding like you'd done your homework.

Forms have their own problem. Long intake forms are a chore to fill in, and short ones tell you nothing.

I run Stepthrough, free hands-on AI and software training for white-collar workers, and I built a lesson on a different approach using Yoya, a booking page with a chat on it. (Disclosure: Yoya is a Stepthrough partner. The founder and I were in touch about the lesson, and no payment was involved.)

The example is Priya, who runs a made-up operations consultancy. The guest is Dan, a made-up COO. The screens are real, and the method works for any service business that takes intro calls.

Part 1: build the page, and write the one box that matters

On your Yoya dashboard, click Consultant discovery call, a ready-to-edit starting point.
Under Basics, set the Link to something you'd put in your email signature.
Set the Title guests see at the top. Host name is a Business plan feature. On other plans, Yoya uses your account name, so skip it.
Replace the Description with one line on who the call is for. It screens out people you don't work with.
Check Meeting options. Connecting Google Meet also grants calendar access, which lets Yoya see when you're busy. I skipped that in my practice run, so Yoya couldn't check a calendar there. Do it on your real page.

Then the important box: What should Yoya find out from guests? Here's the text from the lesson:

Ask which part of their operation they want to improve, how many sites or
staff are involved, what they have already tried, and whether there is a
target date. Keep the introduction short. Do not promise a quote, an outcome,
or project availability beyond this consultation.

Note the last sentence. The assistant is talking to your prospects before you are. It should never be in a position to promise them a price.

Click Create booking page, click Start with weekdays, check the Time zone, then Save hours. Without hours, nobody can book.

Part 2: book it as a guest

This is the part to do yourself. Click Preview, and be the client.

There's no form, just a chat. As Dan, I said I was COO at a freight company and that our two warehouses keep missing dock appointments. Yoya followed up on that, guided by Priya's questions. I answered that trucks arrive in clumps, staff book dock times on a shared spreadsheet, color-coding didn't help, about 60 staff are involved, and we want it fixed before peak season in March.

Then a calendar appeared in the chat. Only Priya's hours are offered, and only days inside the 14-day booking window can be clicked. I picked a time, entered an email and a meeting place, and clicked Confirm booking.

Two practical notes from testing:

Use a real inbox you can open, or you won't see the emails. With a reserved test address such as name@example.com, Continue stays greyed out.
The wording of the chat varies from run to run. Yours may go straight to the calendar.

Part 3: read what the host gets

The guest gets a confirmation with the time, the place, a link to change it and a calendar file.

Priya gets New booking with a Meeting brief: Who, Why, Agenda, Background and where to meet. She knows the problem before she says hello. The same brief is on the dashboard under Bookings.

Compare the opening of that call. Before: "Tell me what you're looking for." After: "You've got two warehouses, double bookings from a shared spreadsheet, and March is the deadline. What did the color-coding attempt teach you?"

Check the brief against the chat

A brief is a summary written by AI, so click Transcript and read the full chat. In my run the brief matched: the double bookings, the spreadsheet, the 60 staff, the failed color-coding and March all made it in. That's one run, though. Read the transcript before a call that matters.

What I'd watch before using this for real

Connect your calendar first, so Yoya can see when you're busy.
Keep the intake questions few. Four was plenty. It's a conversation, and nobody wants an interrogation before they've met you.
Say what it isn't. The description line and the "do not promise" rule do more for you than any clever question.
Plans change. Check Yoya's pricing page for what yours includes.

Practice it step by step

The free lesson has 25 steps in 4 parts and takes about 8 minutes. You click through a practice copy of Yoya, it checks your work as you go, and nothing touches your real accounts.

Practice the Yoya booking brief lesson

Transgate: He Said the Desk Spent 11 Hours a Week. He Never Said It Saved 11. Check Your Case Study Claims First

Robert Breen — Sat, 31 Oct 2026 18:34:58 +0000

A customer case study lives on exact words and exact numbers. So somebody on the marketing team replays the whole call to find them, usually the afternoon before it's due.

That's the slow cost. The bigger one is public. A misspelled customer name or a number the customer never said goes out under your logo, and now you're emailing a customer to apologize for the story that was supposed to make them happy.

I run Stepthrough, free hands-on AI and software training for white-collar workers, and I built a lesson on this with Transgate, a transcription tool with an AI chat over your transcript. (Disclosure: Transgate is listed in our free AI tools directory. No payment was involved, and I used my own free trial account.)

The example: you're Dana, in marketing at a made-up scheduling software company. You just interviewed Marcus, a customer at a made-up physical therapy group. The recording is about two minutes. Swap in your own.

The most useful finding: what it got right and wrong

I ran it twice. Here's the scorecard:

Item	Result
11 hours, 9 no-shows, 3 a week, $90	Correct in both runs
The customer's surname, Oyelaran	Wrong both times: "Oialaren"
A "Sure." the customer said	Put on the interviewer's line, both times
Our company name, Lumen Desk	Written as one word
Hours saved	Chat refused the claim. One analysis run stated it anyway

Numbers came through and the name didn't. That makes sense: speech to text writes what a name sounds like. No AI can know the spelling from sound alone, so get names from your own records.

Part 1: transcribe

In Transgate, add your recording to the drop area. The file shows Ready.
Next to Audio language, click Choose the language and pick yours.
Tick Speaker Recognition. A case study needs the customer's words, so you need to know who said what.
Click Start Transcribing. My 2-minute file took about 10 seconds, and each run took 2 minutes off the balance. The pricing page listed a free trial with 20 minutes when I tested it. Check the current plans.

Part 2: fix it before you do anything else

Order matters here. Fix the transcript first, so everything downstream uses the correct text. The analysis I ran afterward did use the corrected spelling.

Click Speaker 1, then Edit speaker..., type the real name and click the blue check mark. One rename changes every line for that speaker.
Read the transcript against the recording. Point at a line and click the pencil to correct it, then click the check mark.
Check names first, then numbers, then company names.

Part 3: ask for what the case study needs

In Ask AI, ask for the quotes and numbers, and limit it to the transcript:

We are writing a customer case study. List the 3 strongest quotes from
Marcus, word for word. Then list every number he gave and what it measures.
Use only what is in the transcript.

"What it measures" is doing real work in that prompt. In my transcript the number 3 appeared twice: once for locations and once for no-shows a week. A list of bare numbers would have invited a mix-up.

I compared all three quotes against the transcript, and they matched word for word. Yours may differ, so compare them. Tables came back as raw text with | marks, so plan to tidy those.

Part 4: test the claim you're tempted to write

This is the step I'd never skip. Marcus said the front desk spent about 11 hours a week before switching. He never said how many hours it takes now.

Marketing copy wants to say "saved 11 hours a week." So ask the chat directly whether the transcript supports it.

In both runs, the chat said the transcript doesn't state the hours saved, only the 11 hours before. It didn't take the bait.

But the automatic AI Analysis in my first run listed "11 hrs/week saved." The second run didn't. Same recording, different wording, and one of them is the sentence that would have gone into your case study.

So your copy can say no-shows fell from about 9 a week to about 3. It can't say how many hours were saved until you ask Marcus for the after figure and do the math yourself.

The analysis can also use the wrong template

Transgate read my customer interview as an interview and added a Candidate Assessment, rating the customer 5 out of 5 stars. That template suits hiring, not customers. Don't forward the analysis as it stands.

The summary paragraph and suggested next steps were still useful. Both runs suggested getting approval to use his name, which is a real step.

The routine I'd keep

Fix names from your own records before running anything else.
Ask for every number with what it measures.
Ask where the transcript supports each claim you plan to publish.
Treat the analysis as a draft. Its wording changes between runs.
Click Download, confirm your fixes are in the file, and get your customer's OK before you publish.

Practice it step by step

The free lesson has 27 steps in 5 parts. You click through a practice copy with a made-up recording, it checks your work as you go, and nothing touches a real account.

Practice the Transgate customer interview lesson

Mixdesk: Your Venue Inquiries Are One Line Long. Have an AI Agent Collect the Five Details Before Sales Follows Up

Robert Breen — Sat, 31 Oct 2026 18:34:20 +0000

Anyone who sells a private event space knows the message. One line, asking about a date. No event type, no guest count, no budget, no phone number.

So the salesperson writes back and asks. The customer answers a day later with half of it. Somebody asks again. By the time your team knows whether this is a 40-person dinner or a 200-person wedding, days have gone by, and the customer may have booked the venue that answered faster.

The expensive part isn't the typing. It's that your best salesperson spends the morning collecting facts instead of selling.

I run Stepthrough, free hands-on AI and software training for white-collar workers, and I built a lesson on this using Mixdesk, a shared inbox for customer messages with an AI agent feature. (Disclosure: Mixdesk contacted me about listing in our free AI tools directory and asked about a lesson. No payment was involved.)

The example is Cedar Hall, a made-up event venue. The same setup works for a caterer, a law firm's intake desk or anyone whose first message never contains enough.

The goal: a first reply, not a salesperson

The agent's job here is small on purpose. It collects five details, repeats them back, and tells the customer a person will follow up. It doesn't sell, quote or promise. That narrow job is what makes it reasonable to put in front of customers in the first place.

Part 1: build the agent

In Mixdesk, open AI, then AI agents, and click Create AI agent. AI agents are on the trial and paid plans. The free inbox plan doesn't include them.
Add an Avatar (it's required, so use your logo) and a Name.
Write the Persona. This is the job description, in plain language.

Here's the one from the lesson:

You are the first reply for Cedar Hall, a private event venue. Most people
ask about a date and leave out what our sales team needs. Before a person
follows up, politely collect these five details, asking one question at a
time: event type, preferred date, guest count, budget range, and the best
email or phone number. Skip anything they already told you. Do not quote
prices, confirm availability, or promise a date. When you have all five,
repeat them back as a short list and say a member of our team will follow
up within one business day.

The line to protect is the one in the middle: do not quote prices, confirm availability, or promise a date. The agent answers in your name. If it tells someone the Saturday is free and it isn't, that's your problem, not the software's.

Part 2: test it the way customers actually behave

Test preview on the right lets you chat with the agent before it talks to anyone. Don't test it politely. Real customers don't answer the question you asked.

Here's what happened in my run:

I sent a one-liner about a date. It didn't confirm the date, which is what the persona told it to do, and it asked one question: the event type. Good.
It then asked which year. I ignored that and gave it the guest count and budget instead, the way a real customer might.
It moved on to contact details and never asked for the year again.

That's the reason to test. On a clean run you'd never see it: a detail can get skipped when someone answers out of order. In my run, once I sent the email and the date together, the agent produced all five details in a list, plus the one-business-day line. So read each summary for gaps, rather than assuming the agent will circle back.

The summary is what your team reads, so read it the way they will. The stars around the labels were raw formatting in Test preview. Check how it looks on your live channel before you trust that it's tidy.

Click Confirm to save. Linked knowledge bases, Tools, Reply delay and Combine replies are there for later. You don't need them for this job.

Part 3: the step people miss

The agent is saved, and it answers nobody. Something has to send conversations to it.

Click Automation. Default Automation decides who gets each new message. Click Edit.
Add an Assign AI agent node to the canvas and click it.
Choose your agent under Select an AI agent.
Tick Show a "Talk to an agent" button, so a customer can always reach a person.
Click Save, connect the trigger to the node, and Save the automation.

To be straight about what I tested: I stopped before saving the automation, so my own inbox stayed as it was. Test it on a live channel with your own phone before you rely on it.

Three rules I'd keep

Always leave a door to a person. The "Talk to an agent" button isn't optional for anyone planning a wedding.
Keep prices out of the agent's job. Its job is collecting details. Pricing is your salesperson's conversation.
Have someone read the first week of conversations. You'll find the skipped-detail cases faster than any test will.

What this saves

Not the sale. The sale still needs your person. It saves the back-and-forth before the first real conversation, and it means your salesperson opens each lead already knowing the type, date, guest count, budget and how to reach them.

Practice it step by step

The free lesson has 15 steps in 3 parts. You click through a practice copy of Mixdesk with a made-up venue, it checks your work as you go, and nothing touches a real account.

Practice the Mixdesk event inquiry lesson

Higgsfield: A Volunteer Posted an AI "Patient" and a Donor Asked If She Was Real. Write This One-Page Policy First

Robert Breen — Fri, 30 Oct 2026 12:51:57 +0000

Picture a community health nonprofit. A volunteer who handles social media makes an AI image of a "patient," writes a recovery story to go with it, and posts it with no label. It looks great. Then a donor comments and asks if she is real.

There is no good answer to that question. The organization is about to lose something that took years to build, and it did not mean any harm. The volunteer was just trying to get a post out.

That story is the made-up case in one of my lessons (Riverbend Community Health), and it is the case I would show any team that has started using AI images and video. AI output is now good enough that viewers cannot tell. That is the business value, and it is also the risk.

I run Stepthrough, free hands-on AI and software training for white-collar workers. I built a course on Higgsfield (affiliate link: I may earn a commission, at no cost to you), and this is its last module. The rules apply to any AI image or video tool. This is not legal advice. Rules differ by country and state and change often, so have your counsel read your policy.

Not a ban: one page and three rules

A ban does not work, because people will use the tools anyway, just without telling you. A policy lets your team use them freely and tells them where the edges are. Three rules cover almost everything.

1. Disclose

If a reasonable viewer could think an AI image, video or voice is real, label it. Do it in two places:

In the caption: "AI-generated" or "Made with AI."
In the platform's own setting:
- Instagram and Facebook: AI info
- TikTok: the AI-generated content toggle
- YouTube: the altered or synthetic content question in Studio

Obvious illustrations and cartoon mascots do not need a label, though they can carry one.

2. Get consent

A person's face, body and voice belong to them. Use them only with written consent for that specific use, and let them approve the final. That includes colleagues. "She works here" is not consent. Keep the form with the asset, and decide up front how fast you will take something down if someone withdraws consent.

3. Never deceive

A label does not fix this category, so it needs its own list:

Fake testimonials, reviews, patients, customers or employees shown as real
Real people, including public figures, shown doing or saying what they did not
Fake events or "before and after" results
Upscaled or edited images used as evidence of what something looked like

Go back to the opening story. Adding "AI-generated" to that post would not have saved it. A recovery story from a person who does not exist is a fake testimonial, and those are treated as deceptive advertising.

Read what your tool makes you responsible for

Most people never open the terms. I read Higgsfield's Terms of Use on October 29, 2026, and they put the responsibility on the user. Check the current version yourself, and do the same for any other tool you use:

Section	What it says, in short
4.2, 5.3	You need the consents for any face, name or voice you upload
5.2	You may not impersonate anyone or pass AI output off as human-made
5.5	You disclose AI output where the law requires, and you keep provenance marks
4.4	Uploads, prompts and outputs may be used to train its models, unless you have a separate enterprise agreement

The last row is the one marketing teams tend to miss. Do not upload client documents, unreleased products or private photos you would not want used that way.

Two more spots matter in practice:

The upload agreement. The first time you upload a photo, Higgsfield asks you to agree that you own the media and that everyone visible in it has consented. Treat that checkbox as a real promise. Only tick it when it is true.
Auto-publish new generations. In Settings, this puts every new image and video in the public Explore feed. Keep it off for client and unreleased work. Check this before your first client job, not after.

Add a check and a record

The three rules cover the big risks. Two small habits cover the rest.

Before publishing, read every word in the image or video. AI output gets prices, pay, dates and names wrong, and warps products and hands. Make sure what is shown is what you actually offer, and that no other company's logo has wandered in.

Keep a record for each AI asset you publish: the tool and model, the prompt, the input files, the consent form if a person appears, who approved it, and the date. When someone asks "where did this come from?" six months later, you can answer in a minute.

Name an owner

A policy nobody owns is a file nobody opens. Put one name on it, set a review date, and share it with everyone who posts for you. That includes the volunteer.

Try it, free

The module takes about 7 minutes and has 9 steps. You see the donor case, read the real terms, and sort four real outputs from the course into OK and not OK. One of them is not OK. There is a one-page policy template to download and fill in, with blanks for your owner, your record location and your review date.

Practice the labels, consent and likeness lesson

Claude and Jinxo: Get the Client Kickoff Agenda Out of One Person's Head and Into the Group Chat

Robert Breen — Fri, 30 Oct 2026 12:51:24 +0000

A small consulting firm signs a new client. Good news. Now somebody has to write the kickoff agenda and the prep list, and the team chat fills up with "who is doing the agenda?" while nobody starts. It gets written from a blank page the night before, by whoever feels guilty first.

The cost is small each time and it never stops. Every new client repeats it. And when one person drafts it alone in a private AI chat, the team still has not seen it until the morning of the call.

I run Stepthrough, free hands-on AI and software training for white-collar workers, and I tested a different setup: put Claude in the group chat with the team. The tool is Jinxo, a group chat for people and their AI bots. (This is my own lesson, not an official Jinxo course, and there is no partner link here. Its Terms say Jinxo is free right now but may not always be.)

The example firm, Quillfield Consulting, and its client, Henderson Dental, are made up. The screens I describe are real, captured in October 2026.

What changes

The usual flow is: open Claude alone, ask for a draft, copy it out, paste it in the chat, and explain what it is.

This flow is: ask in the room, and the answer lands in the room. Everyone reads the same draft, and the team edits one document instead of starting from nothing.

The setup has two halves

This is the part that trips people up, so it helps to know the shape first. One half is in Jinxo. The other is in claude.ai. Until both are done, the Claude row in Jinxo says Not connected yet.

In Jinxo:

Open a room and add Claude to it.
Open Connect Claude. It gives you a name and a connection link.

In claude.ai:

Go to Customize, then Connectors, then + Add, then Add custom connector.
Paste the name and the link.
Choose No sign-in, then Add and Connect.

Claude's Free plan allows only one custom connector, so check that you are not already using it.

Treat that connection link like a password. There is no sign-in step on this connector, so the link is what stands between your room and anyone who has it. Do not paste it into a ticket, a doc or a chat.

Check the permissions before you use it

Once connected, Tool permissions lets you choose when Claude may use each Jinxo tool. Leave approval on while you are learning. You will see exactly what Claude is about to do, such as sending a message, before it does it. You can loosen it later, once you trust the pattern. Starting loose and tightening after a surprise is the wrong order.

Ask in the room

In Jinxo, start your message with @Claude and say what you want back, with limits. Vague asks get vague agendas. The two details that help most are the length of the call and how many prep items you want:

@Claude We have a kickoff with Henderson Dental on Tuesday at 10.
Can you draft a 30-minute agenda and three things we should prep?

Then tell Claude to go read it

This is the step people miss. Your message in Jinxo does not wake Claude up by itself in this setup. In a new claude.ai chat, check that the Jinxo connector is on under + and Connectors, then ask Claude to read the request and post its answer back in the room.

You will see a line like Using jinxo-...: Send a message. That is Claude using the connector to post. Depending on your permissions, it may ask you to approve first.

What came back

Back in Jinxo, the message showed Read by Claude, and the group chat had a new reply. In my run, the agenda had five timed blocks:

Intros and roles
Their goals
Current state
Scope and approach
Next steps

Below it were three prep items: research the practice, draft an access and data checklist, and bring two or three quick-win ideas. Claude ended by offering to tighten it once we knew who was attending from the client's side.

Claude writes a fresh answer each time, so yours will not match this.

Three things to watch

It is a first draft. Read it and fix what does not fit the client. If the agenda lists a detail about the client you never gave it, check that detail and remove it if it is not true. Do not ask the client to confirm something your AI made up.

Every bot in the room can read the room. Keep private client details out, especially while you are learning. Practice on made-up or low-risk details first. A kickoff agenda needs the client's name and a time. It does not need their contract.

Keep it for low-stakes drafts first. An agenda is a good first job because a mistake costs a minute of editing. Get comfortable here before giving a bot anything that goes outside the team.

What this is worth

The time saved on one agenda is modest. The real change is that the blank page and the "who is doing this?" thread both go away. The next kickoff starts from a draft everyone can see.

Practice it step by step

The free lesson walks you through all of it in a practice copy built from real Jinxo and claude.ai screens. It checks your work as you go, and nothing touches your real accounts.

Practice adding Claude to a Jinxo room and drafting a kickoff agenda

OLVA: Stop Half-Listening in Job Interviews. Load the Scorecard First, Then Debrief Against It

Robert Breen — Fri, 30 Oct 2026 12:51:05 +0000

Watch someone interview a candidate while typing notes. Their eyes go to the keyboard right when the answer gets interesting. They write down the story and skip the follow-up question that would have shown how good the person really is.

Then the debrief happens a day later, on memory. Two interviewers score the same answer differently, and nobody can say who is right. For a hiring manager, that is an expensive way to decide: a wrong hire costs far more than the hour you spent in the interview.

I run Stepthrough, free hands-on AI and software training for white-collar workers, and I built a lesson on fixing this with OLVA, an AI meeting assistant. (Disclosure: OLVA is a Stepthrough partner. It links to our lesson from its site. No payment was involved.) OLVA runs on your computer instead of joining the call as a bot, so there is no extra name in the participant list.

The example is a made-up panel interview for a senior bookkeeper at a small firm. The method does not care about the job. Swap in your own role and your own scorecard.

The idea: give the AI the role before the call

Most people start an AI note taker and hope. The better order is to give it the scorecard first, so it can answer "what have I not covered?" and not just "what was said?"

In OLVA, that lives in a context. Make one context per open role.

Part 1: load the scorecard and the resume

Click Manage contexts, the blue button next to the context picker.
Click Add context. It starts as Untitled and saves itself. Name it after the role.
Click Add a note and paste in four things: the role, the must-haves, the questions every candidate gets, and your rules.
Click Save.
Click Add document and pick the candidate's resume (PDF, Word, text or Markdown). OLVA converts it and shows it as text, with Make shorter, Save and Cancel. Read it, then click Save.

The note in the lesson looks like this. Note the last line, because it does the most work:

Role: Senior Bookkeeper. Hybrid. Reports to the Controller.
Scorecard (must-haves): 1) owns month-end close for several clients,
2) bank and credit card reconciliations, 3) QuickBooks Online and Xero,
4) explains numbers clearly to clients, 5) trains junior staff.
Questions to cover: a close that went wrong, clients who send receipts
late, a time they trained someone.
Rules: ask every candidate the same core questions.
Score against the scorecard only.

Part 2: start it, and say so

Click the context picker and choose the role under CHOOSE CONTEXT.

Before you click Start OLVA, tell the candidate you are taking AI notes. Some places require consent to transcribe, so check the rule where you and the candidate are. Saying it out loud also takes about five seconds and sets a straight tone for the conversation.

Part 3: use it to listen, not to type

Turn on Live insight at the top right. The live window opens beside your call.

When a candidate tells a good story, do not move on yet. Click Follow-up under the transcript. Suggestions show in the Chat tab, where only you can see them. Ask the follow-up in your own words, and aim it at a scorecard item. If the candidate found an error in the books, ask how they explained it to the client. That one question tests "explains numbers clearly," which a story about finding the error does not.

Partway through, click Ask about this call and send this:

Which scorecard items haven't we covered yet?

OLVA answers from the transcript and your notes. If two items are still open, you have time to ask about them. Without this, you find out in the debrief, when the candidate has already gone home.

Part 4: what to trust, and what not to

This is where I would slow down. Here is what my test run did and did not show:

A Define card can explain a term you do not know, which helps a recruiter follow a specialist's answer.
OLVA can also add Fact Check cards, but none came up in my test. Do not count on one.
If a candidate says "five years" and the resume says otherwise, note it yourself, as a question for the next round, not a verdict. People round. Resumes leave things out. Ask.

Part 5: a debrief the panel can use

When the interview ends, click End, then Stop Meeting. The meeting opens on Summary. Click the Chat tab, and under Try asking OLVA, click Create a team handoff. Every answer ends with a copy button. Then click What open questions? for your round-two list.

I did not capture those two answers in my test, so the lesson says so instead of pretending. Read them against your scorecard and fix anything misheard before they go to the panel.

Three rules I would keep

Same questions for everyone. The scorecard only helps if every candidate gets measured against it.
The AI keeps the record. You make the call. Do not let any tool score or rank the candidate for you. Software can carry bias, and you need to be able to explain your decision later.
Score before you read the group's opinions. Otherwise the first person to speak sets the score for everyone else.

Practice it step by step

The free lesson has 26 steps in 5 parts and takes about 8 minutes. You click through a practice copy of OLVA with a made-up candidate, it checks your work as you go, and nothing touches your real accounts.

Practice the OLVA interview and scorecard lesson

Gamma: The AI Built Your Leadership Deck in Minutes. Now Check Every Number Before Anyone Quotes It

Robert Breen — Wed, 28 Oct 2026 20:29:50 +0000

Here is the risk nobody mentions when an AI tool builds your slides: the deck looks finished before it is correct.

A clean theme and tidy bullets make every line feel checked. Leadership reads "Overtime down 18% thanks to the shift swap app" and takes it as fact. If your notes only said overtime fell since the app arrived, you just made a claim you can't back up, in front of the people who approve your budget.

I run Stepthrough, free hands-on AI and software training for white-collar workers. In the lesson I built for this, the tool is Gamma (affiliate link: I may earn a commission, at no cost to you). The generating part takes a few clicks. This article is mostly about the part after that: reading the deck the way a skeptical VP will.

The scenario

You are the operations manager at Brightwater Facilities Group. The business, its people, vans and suppliers are made up for the lesson. The Q3 review with leadership is Thursday, and all you have is this:

Q3 ops review - notes for Thursday leadership mtg
on-time delivery 91% this qtr (Q2 was 86%), new route schedule helped a lot
overtime down 18% since the shift swap app
ticket backlog 140 open -> 62 open
2 new dispatchers started in Sept
paper supplier late 4x in Aug, moved 2 sites to the backup vendor
vans 3 + 7 past repair, about $48k to replace both
Q4: replace 2 vans, renegotiate supplier contract, pilot a weekend shift
ask: approve van budget by Oct 31

Those nine lines are your source of truth. Every number on every slide should trace back to one of them.

Why the review pass matters

The meeting has one real goal: get the van budget approved by October 31. If one number on one slide is off, the conversation turns to that number and the ask never gets discussed.

Build it with settings that protect your words

On the Gamma home page, click Use classic generator. If Gamma asks why, click Skip.
Choose Paste in text and paste the notes as they are. Don't tidy them.
Under What do you want to do with this content?, choose Summarize long text or document. Presentation is already selected. Click Continue to prompt editor.
In Settings, under Text content, make sure Condense is picked. Generate adds detail and Preserve keeps your words; Condense tightens.
Under Content, click the first line and replace it with Q3 Operations Review: Brightwater Facilities, so anyone who gets the file later knows what it is.
Pick the Tranquil theme (click View more if it isn't shown). Under Amount of text, choose Concise.
Open Image source and choose Don't add images. AI images use extra credits, and this deck is about numbers.
Click Generate.

Gamma writes its own version every time, so your cards may come out in a different order from mine.

The review pass, card by card

Put your nine lines next to the deck and read each card against them. In the lesson's run, the outline included cards like these:

What drove the gains: New route schedule, Shift-swap app, Two new dispatchers since September.
Fleet: two vans to replace: Vans 3 and 7 are past repair, About $48,000 to replace both.
The ask: Approve the van budget by October 31.
Q3 at a Glance: Delivery: 91% on-time, up 5 points from Q2. Overtime: down 18% thanks to the shift swap app. Backlog: 140 to 62 open tickets cleared.

The numbers survived. 91 against 86 is 5 points. $48k became $48,000. October 31 is right. Condense did its job.

The wording is where I slow down. Three things I'd flag:

"thanks to the shift swap app" turns since into because. Your notes say the drop came after the app, not that the app caused it.
"Two new dispatchers" now sits under What drove the gains. Your notes say two dispatchers started in September. That's a fact, not a cause.
"140 to 62 open tickets cleared" can be read as 78 tickets cleared, or as 140 cleared. The notes say the open count went from 140 to 62.

You decide which of these to keep. The point is that you decide, not the generator.

Fix a line by hand

Click the last card, Q3 at a Glance.
Click the Backlog line to edit it.
Select the line and type Ticket backlog down from 140 to 62. Don't press Enter, that only adds a new line. Click outside the card when you're done.

Word it the way leadership will quote it back to you.

Change a whole card by asking, then verify

Click the Q4 Priorities card, then click Agent at the top right, next to Present.
Ask in plain words and name the slide. The lesson uses: On the Q4 Priorities slide, turn it into a 3-step timeline: October, replace the two vans. November, renegotiate the supplier contract. December, pilot a weekend shift. Keep my wording.
Agent replies that it kept your wording and nothing else changed. Don't take that on trust. Click the Original preview, compare, then click Modified to switch back.

The change is already applied, so there's nothing to accept. If you dislike an edit, press Ctrl+Z.

Export, then review once more

Open the ⋯ menu at the top right and choose Export....
Click Export to PowerPoint, then Export to PDF.
Open the PowerPoint file once. Cards don't always map one to one onto slides, so check the layout.

The PDF keeps the design exactly, so it's the safest copy to email before the meeting. On the free plan, both files carry a small Made by Gamma badge. Click Present to rehearse full screen.

Common mistakes

Trusting the polish. A designed slide is not a checked slide. Read every card against the notes.
Picking Generate for a review deck. Generate adds detail you didn't write. For numbers leadership will quote, use Condense, or Preserve if you want your exact words.
Believing the AI's summary of its own edit. Agent said nothing else changed. Use Original and Modified to see for yourself.

Checklist before it goes to leadership

Every number traces back to a line in your notes.
No "thanks to" or "drove" where your notes only showed timing.
The ask, with its date, has its own card.
The first card has the full title.
The PowerPoint layout was opened and checked.
The PDF is the copy you email.

Practice it step by step

I built a free lesson where you do every step yourself in a practice copy of Gamma, with the same notes, the same Agent request and the same backlog fix. It checks your work as you go, takes about nine minutes, and nothing touches your real accounts: A Presentation from Your Notes in 5 Minutes (Gamma).

Foxit: If You Sign Vendor Renewals, Compare Every New Draft Against the Version You Agreed To

Robert Breen — Wed, 28 Oct 2026 20:29:14 +0000

In a lot of small offices, the person who signs software renewals is not a lawyer. It is the operations manager, the office manager, or whoever owns the budget line. The renewal lands in their inbox with a deadline, and the review is whatever they can fit in between other work.

That is the job I want to make easier here. Not "read contracts better." Just one repeatable check you run every time a revised agreement shows up, before your name goes on it.

I run Stepthrough, free hands-on AI and software training for white-collar workers. The tool in this routine is Foxit PDF Editor (affiliate link: I may earn a commission, at no cost to you). The steps below come from a free lesson I built on real Foxit PDF Editor Cloud screens from my own account, captured October 2026. The company, the vendor and both versions of the agreement are made up.

The problem: you review the new draft, not the difference

When a revised contract arrives, most of us open the new one and read it. That feels like a review, but it is not a comparison. You read looking for what you expect to change, and you fill in the rest from memory.

The expensive edits are the ones nobody mentioned. A number in the payment clause. A notice period. One new sentence in the middle of a section you already approved last year.

Lining up two PDFs by hand is slow. In the lesson's sample it takes about 40 minutes, and it is still easy to slide past one changed figure.

Why it matters for the person who signs

In the sample, the business is Bright Path Marketing, and Dana Reyes is the operations manager. Their software vendor sends version 2 of a subscription agreement and asks for a signature by Friday.

Here is what version 2 quietly changed, compared with version 1:

Section 1: named users went from 25 to 20.
Section 2: the fee went from $1,200.00 to $1,350.00 a month, and invoices are due in 15 days instead of 30.
Section 3: notice to cancel went from 30 to 90 days, and the renewal increase allowed went from 5% to 9%.
Section 5: a new sentence lets the vendor change, add or remove features with 10 days notice.
Section 6: the liability cap went from 12 months of fees to 3 months.

The fee alone is $150 a month more, which is $1,800 a year. The person who signs owns that number whether they spotted it or not.

The routine, step by step in Foxit

1. Start from the version you agreed to. In Foxit PDF Editor Cloud, click Open File and pick the old version. In the lesson that is Sample-Subscription-Agreement-v1.pdf. This matters more than any other step, so I come back to it below.

2. Open Compare. Compare lives on the View tab in the ribbon. Click View, then click Compare.

3. Add the new draft. The Compare PDF Files dialog opens with your open file already set as the Old File on the left. Under New File, click Choose and pick the version the vendor sent (Sample-Subscription-Agreement-v2.pdf in the lesson). Both files show a preview, and the page range is set to all pages, 1 to 2. Click OK.

4. Read the count first. Foxit makes a new PDF called The result of Comparison.pdf. Page 1 is a Summary. In the sample it shows 10 total changes: 9 replacements, 1 insertion and 0 deletions. That count is your reality check against "just a few cleanups."

5. Walk every change in order. On the Compare Results tab, click Next Change. Old version on the left, new on the right. Read the right side for what the contract says now, and the left side for what it used to say. Replaced text has a yellow wavy underline, inserted text a blue underline, and deleted text a red strike. Keep clicking Next Change. It stops on each mark, so you never skip one.

6. Pull the list. On the left edge, click the Comments icon. Every change is saved as a comment, grouped by page. Click a page, such as Page 3, and each card says Replaced Text with the [Old Text] and [New Text], for example "thirty (30" to "ninety (90".

7. Keep a copy. Foxit keeps the result in your Foxit files. Click Download File in the top bar to save it to your computer, next to the contract.

8. Reply instead of signing. Send the vendor the changes you did not agree to and ask which ones they meant. In the sample, that reply lists seven changes. A short, specific list gets a faster answer than "please explain the changes."

Common mistakes

Comparing against the wrong old file. If you compare version 2 against the vendor's last draft instead of the version you agreed to, any change already sitting in that draft looks like "no change." Compare is only as good as the old file you give it. Keep the agreed version of every contract somewhere you can find it.

Stopping at the summary. The count tells you how many things moved, not which ones cost money. Money, dates and notice periods are where a small edit costs the most, so step through each one with Next Change.

Treating a mark as a verdict. Compare shows what changed, not whether it is fair. A lower liability cap is a question to ask, not an automatic no. For anything large, have a lawyer look at it.

A short checklist for every renewal

I opened the version we agreed to, not the vendor's last draft.
I ran Compare with the new draft as the New File and all pages selected.
I read the Summary count.
I clicked Next Change until the end and read the right side each time.
I checked fees, payment terms, notice periods, renewal increases, new sentences and liability.
I copied the old and new text from the Comments panel into a reply.
I downloaded The result of Comparison.pdf and filed it with the contract.
Anything large went to a lawyer before I signed.

That is the whole routine. It does not replace judgment. It just makes sure your judgment is working on the actual changes, not on what you expected to see.

Practice it step by step

I built this as a free, hands-on lesson. You click through the real Foxit screens yourself in a practice copy, with the same two sample agreements, and it checks each step as you go. It takes about 8 minutes and nothing touches your real accounts.

Practice comparing two contract versions in Foxit PDF Editor

Make: Stop Letting Urgent Requests Sit in the Same Pile as Routine Ones. Route Them With a Router and a Filter

Robert Breen — Wed, 28 Oct 2026 20:28:38 +0000

A lot of small offices I talk to have the same setup. A web form collects requests, an automation drops each one into a spreadsheet, and someone checks the sheet when they get a moment.

That works fine until one request is not like the others. A patient with a chipped tooth lands in the same list as a patient who wants a cleaning next month. Nobody sees it until they scroll down. The cost is not a line on an invoice. It is a call-back that happens hours late, to the one person who needed it fast.

I run Stepthrough, free hands-on AI and software training for white-collar workers, and this is one of the lessons I built for it. The example business is Brookside Dental, a small dental practice. It is made up, but the problem is real for office managers, front-desk staff and small business owners who already have one simple Make scenario running.

Why this matters

A scenario that only copies data is a pipe. A scenario that sorts data is a helper. The step from one to the other is small: a router, one filter, and a fallback route.

A router on its own does not choose anything. Without filters, every bundle of data goes down every route, one after another. The filter is what makes the decision. The fallback route is what makes sure nothing gets lost.

The setup

You start from a basic scenario: a Google Forms trigger that adds each new patient request as a row in Google Sheets. The form has a new question, How urgent is this?, with two choices: Urgent and Routine.

The goal: urgent requests email the front desk. Everything else still lands in the sheet, as before.

Part 1: add a router

In Make, Right-click the dotted line between Google Forms and Google Sheets.
Click Add a router. The router now sits between the two modules, and Google Sheets is on the first route.
Click the + on the router to add a second route.

Part 2: build the urgent route

Click the empty module at the end of the new route.
Type gmail, then click Gmail.
Under Actions, click Send an email.
Gmail is a separate app in Make, so it needs its own connection. Click Create a connection, keep the default name, and click Sign in with Google.
Pick the Gmail account the alert should come from, then click Continue.
In To, type the front desk's address. The lesson uses frontdesk@yourbusiness.com.
In Subject, type Urgent patient request.
Leave Body type as it is. Click the Body contents field.
In the mapping panel, under Full name, click Value. Do the same under Phone number and Reason for visit.
Click Save.

In your own scenario, type a few words between the mapped tags, like "Name:" and "Phone:", so the email is easier to read at a glance.

Part 3: filter for urgent only

Right now every response would go down both routes. The filter fixes that.

Click the wrench on the line going to Gmail.
Click Set up a filter.
In Label, type Urgent.
Click the first Condition field. Under How urgent is this?, click Value.
Keep the operator Equal to. In the field below it, type Urgent.
Click Save. The label now shows on the route.

A condition has three parts: a value you map from an earlier module, an operator, and the value to compare it with. If the condition is true, the data continues. If not, it stops there.

Part 4: make the sheet the fallback route

Click the dotted line near Google Sheets. A left-click on the dots opens the filter straight away.
In Label, type Routine.
Set Fallback route to Yes.
Click Save. The arrow on this route changes to mark it as the fallback.

Part 5: set the route order

Routes run one after another, in the order you set, not at the same time. Put the urgent one first.

Right-click the router.
Click Order routes.
Click the up arrow next to the Gmail route.
Tick Auto-align arranges with set order, so the canvas is redrawn in the new order.
Click Apply.

Part 6: test it

Click the Save icon in the toolbar.
Click Run once. In the lesson, two requests are waiting: James chipped a tooth and marked it urgent, and Priya wants a cleaning.
Each route shows a bubble with a 1. One response went each way.
Open the Gmail tab: there is the alert about James, with his phone number, ready for a call back.
Open the Google Sheets tab: Priya's request is in the sheet.

Common mistakes

Typing the filter value loosely. Text comparisons with Equal to must match exactly. "urgent" or "Urgent " with a trailing space will not pass. This is why a multiple-choice question works so well here: the answer is always spelled the same way.

Writing a second filter instead of a fallback. It is tempting to add a "Routine" condition on the sheet route. Then a request that is neither, or a third choice someone adds to the form later, goes nowhere. The fallback route catches everything that did not pass the other filters, so no request gets lost.

Expecting instant alerts. The Google Forms trigger checks on a schedule, every 15 minutes by default. An urgent alert arrives within that window, not the second the form is sent. Tell the front desk that, so they trust the system for what it is.

Checklist

Router added between the trigger and the sheet
Gmail Send an email on the new route, with its own connection
To, Subject and Body contents filled, with name, phone and reason mapped
Filter labelled Urgent, condition Equal to Urgent
Sheet route labelled Routine, Fallback route set to Yes
Gmail route moved first in Order routes
Scenario saved and tested with Run once

Where else this pattern fits

Any time one kind of request needs different handling, the shape is the same: a router, one filtered route for the special case, and a fallback route for everything else. Big orders to a manager. After-hours calls to an on-call phone. VIP customers to a separate sheet. Complaints to a shared inbox. You can combine conditions with Add AND rule and Add OR rule when one check is not enough.

Practice it step by step

I built a free interactive version of this lesson: 40 steps in 7 parts, about 12 minutes. You do every click yourself in a practice copy of Make that checks your work as you go, and nothing touches your real accounts. If you get stuck, "Do it for me" finishes the step.

Start it here: Filters and Routers in Make.com

It builds on the first module, where you connect Google Forms to Google Sheets, so if you have not built that scenario yet, start there.

Lusha: Treat Contact Credits Like a Budget. Filter and Save Freely, Reveal Only the Three People You'll Call This Week

Robert Breen — Tue, 27 Oct 2026 12:40:52 +0000

Most small sales teams I talk to prospect from an old spreadsheet. Some firms on it are too big, some too small, a few have closed, and for most of them nobody knows who actually signs. Every call to a wrong firm is a call not made to one that would buy.

Contact databases fix the list, but they bring a new way to waste money: revealing everyone "just in case." Each reveal uses credits, and each revealed contact is a real person whose details you now hold.

So the habit I teach is simple. Filter and save as much as you like. Reveal on purpose.

The tool here is Lusha (affiliate link: I may earn a commission, at no cost to you). Every click below comes from the free practice lesson I built on real Lusha screens (the free plan, 40 credits, 25 October 2026). The search returns real companies and people, so all of them are blurred in the lesson, and no contact was revealed. The business in the example is made up.

The scenario

You sell outsourced payroll to accounting firms in Pennsylvania. Your best customers have something in common: they're accounting firms, headquartered in the state, big enough to need payroll help and small enough that a partner still makes the call.

That sentence is your ideal customer profile (ICP). Each trait becomes a filter.

Part 1: describe the ideal account

In Lusha, click Company filters. The filters sit in the left panel; the table fills in as you add them.
Click Industry, type Accounting and tick Accounting & Services.
Open Company location, type Pennsylvania and pick Pennsylvania, United States.
Tick HQ location only.
Open Company size and tick 11-50 and 51-200.

Watch the count at the top while you go. In the lesson it starts at about 256,000 companies and ends at a few hundred. That drop is the whole point: a few hundred firms worth a call instead of a database worth of maybes.

Why "HQ location only" matters

Without it, every national firm with a branch office in Pennsylvania shows up, and those aren't the regional firms you're after. One checkbox removes them.

Part 2: save the list where the team can find it

Click Save to table.
Drag the slider to how many companies you want to keep (the lesson keeps 40), then click Create new table.
Lusha names the table after your filters and keeps it Private. Open the menu next to the name, click Rename, type something the team will recognise (PA accounting firms) and click Save.

In the lesson's run, saving companies to a table didn't touch the credit counter. It stayed at 40.

A saved, named table is what replaces everyone's own spreadsheet. Everyone works from the same list, and nobody calls the same firm twice.

Part 3: find who decides, then stop

Tick a company in the table. View employees lights up.
Click the small arrow next to it. Lusha groups people by buying role: Buying group, Decision makers, Potential champions, End users.
Click Decision makers.

You get the people with their role and a fit score. Their emails and phones stay masked. Listing them didn't use a credit in our run either.

Now the button that does cost credits: Show details. It reveals one person's contact details and spends credits from your plan.

This is where the budget habit kicks in. Reveal only the two or three people you'll actually contact this week. The lesson stops right here, with nothing revealed.

Why the budget framing works

Credits are finite. On a free or small plan, revealing a whole table can use up your credits in an afternoon.
Data goes stale. A contact you reveal today and email in six weeks may have changed jobs.
Every reveal is a real person. Once you hold their details, the email and privacy rules that apply to you (such as CAN-SPAM or GDPR) apply to how you use them. Keep the table private and reveal only who you'll contact soon.

Mistakes I'd avoid

Skipping the ICP. Filtering without one just makes a smaller random list. Look at your five best customers first and write down what they share.
Leaving HQ location only unticked for a regional list.
Revealing the whole table. Browse roles freely, reveal two or three.
Leaving the table with its auto-generated name. Nobody else will find it.

The checklist

[ ] Write the ICP: industry, territory, size
[ ] Company filters: Industry, Company location (HQ location only), Company size
[ ] Save to table > Create new table
[ ] Rename it so the team recognises it (it stays Private)
[ ] Tick a firm > arrow next to View employees > Decision makers
[ ] Show details only for the people you'll contact this week

What comes after the reveal

The list is only the start. Once you've called someone, the next job is not letting the conversation go cold: log the call and set a dated follow-up. That's another lesson in the same course, Close CRM: Log a Sales Call and Set a Follow-up.

Practice it step by step

The lesson, Lusha: Build a Target Account List and Find the Decision Makers, walks through every click above on real Lusha screens, with every company and person blurred, and it doesn't touch any account. It's part of my free AI for Sales Teams course. To do it for real, Lusha has a free plan with a small number of credits; check lusha.com for current plans.

Practice it step by step, free: https://learn.ynteractive.com/content/lusha-target-account-list

Close CRM: Write the Call Notes for Whoever Picks Up the Account Next. Log a Call, a Dated Task and the Deal in Two Minutes

Robert Breen — Tue, 27 Oct 2026 12:37:43 +0000

The test I use for any sales process is simple: if the rep who took the call were out tomorrow, could someone else pick up the account and keep the promise?

In most small teams the answer is no. The notes are in a notebook or on a phone. "Send the proposal by Friday" is a sentence somebody said out loud. The budget the prospect mentioned is in one person's memory. The account doesn't go cold because anyone did a bad job. It goes cold because the next step was stored in the one place nobody else can read.

The fix is a two-minute routine at the end of every call, done before you leave the lead's page. I'm using Close (affiliate link: I may earn a commission, at no cost to you), and every click below comes from the free practice lesson I built on real Close screens (a trial account, 25 October 2026). The company, the contact, the phone number, the notes and the deal in it are made up.

How Close thinks about an account

Four objects, all on one lead page:

A lead is a company.
Contacts are the people at it, each with a phone and email.
An opportunity is a deal with a value, a status (pipeline stage) and a confidence.
A task is a dated to-do that lands in your Inbox.

Because all four sit on one page, whoever opens the lead sees what was said, what was promised and what happens next. That's the whole hand-off.

The scenario

You just finished a discovery call with Dana Whitfield at Lakeview Accounting. The call went well. Dana wants a bookkeeping proposal by Friday, and the work is worth about $1,200 a month.

1. Create the lead (company and contact)

Click Leads in the left bar, then Create Lead. If you already have leads in Close, use the + next to Leads.
Type the Company name, the Contact name, the Phone and the Email. Close checks for duplicates as you type.
Click Create Lead.

The lead page opens: details on the left, the activity timeline on the right.

2. Log the call, don't dial it

This is the step that trips people up. The Call button starts a real phone call through Close. Your call already happened, on your mobile or in a meeting, so you don't want that.

Click the small arrow next to Call (not Call itself).
Choose Log a call.
A call entry opens on the timeline. Type your notes.
Click the clock (it starts at 0s) to set the duration (15 minutes in the lesson).
Click Done. Changes save automatically.

What goes in the notes

Write them for the person who picks up the account next, not for yourself. Four things:

The need: what problem they're trying to solve.
The timing: when they want to decide or start.
The budget: what they said, in their words.
What you promised: the proposal, by Friday.

If your raw notes are a mess, I keep a call notes to CRM update prompt that turns rough notes into a clean update before you paste them in. Check what it gives you before you save it; it can tidy your notes, but only you know what was actually said.

3. Set the follow-up, dated the day before the promise

Click the + next to TASKS.
Type the task: Send bookkeeping proposal to Dana.
Click the date and pick the day before it's due. Dana wants it Friday, so the task is Thursday.
The task is assigned to you. Click Create Task.

Why Thursday? Because a task dated Friday reminds you on the day you're already late. A day of slack is what lets you send something good instead of something rushed.

4. Put the deal in the pipeline

A deal that isn't in the pipeline isn't in anyone's forecast, so nobody notices when it quietly cools.

Click the + next to OPPORTUNITIES.
Type the value (1200) and switch One-time to Monthly.
Open the status list. These are the stages of Close's default pipeline. Keep the first one, Demo Completed.
Close date and confidence are optional (50% is the default). Click Save.

The lead now shows $1,200 monthly.

5. Check that it will come back to you

Don't trust it until you've seen it.

Click Opportunities in the left bar. The Sales Pipeline board shows the deal in Demo Completed, with its annualized value.
Click Inbox, then Future. The task is waiting there with its date. On Thursday it moves into your Inbox.

Mistakes worth avoiding

Clicking Call instead of the arrow. Call dials. Log a call records a call that already happened.
Notes only you can read. "Good call, send prop" helps nobody on Monday. Need, timing, budget, promise.
A task with no date, or dated the deadline itself. The date is the reminder. Put it a day early.
No opportunity. Without it the deal isn't on the board or in the forecast, so a cold deal looks exactly like a deal that never existed.
Leaving the page first. If you plan to "do it after lunch," you won't. Two minutes, before you close the tab.

The checklist

[ ] Lead with company and contact
[ ] Arrow next to Call > Log a call, notes (need, timing, budget, promise), duration, Done
[ ] + TASKS, the next step, dated the day before the promise, Create Task
[ ] + OPPORTUNITIES, value, Monthly or One-time, first stage, Save
[ ] Opportunities board and Inbox > Future checked

Every call ends with a logged call and a dated next step. That's the habit, and it's also the hand-off: if you're out on Thursday, a teammate who opens the lead sees the call notes, the dated task and the deal in one place, and can keep the promise.

Practice it step by step

The lesson, Close CRM: Log a Sales Call and Set a Follow-up So No Deal Goes Cold, walks through every click above on real Close screens, and it doesn't touch any account or make a call. It's part of my free AI for Sales Teams course. To do it for real, Close offers a 14-day free trial; check close.com for current plans.

Practice it step by step, free: https://learn.ynteractive.com/content/close-crm-log-call-follow-up

Google Analytics 4: Three Channels, One Hour Each, No Idea Which Works. Sort Traffic Acquisition by Engagement Rate

Robert Breen — Tue, 27 Oct 2026 12:32:34 +0000

Here's a week I see in a lot of small offices. Someone posts on LinkedIn twice. Someone keeps the Google Business Profile up to date. Someone writes the monthly newsletter. Each of those takes hours, and if you ask which one actually brings people to the website, the honest answer is "no idea."

So every channel gets the same time, and none of them gets better.

The answer is usually already sitting in Google Analytics 4. Nobody opens the right report, or they open it and read the wrong column. This post is the ten-minute version: two reports, one dropdown, and one decision at the end.

All the clicks below come from the free lesson I built on real GA4 screens. The property in the lesson is a real website, so every number on those screens is blurred, and I'm not quoting any figures here either. You'll read your own.

What you need

Access to a GA4 property. The Viewer role is enough. Reading reports changes no settings.
No property yet? Google offers a free demo account (the Google Merchandise Store) you can practice on.

First, three words GA4 uses all the time

Get these straight and the reports stop being confusing:

A user is one person (strictly, one browser).
A session is one visit.
An event is anything they do: a page view, a scroll, a click, a form sent.

And one more that matters for the decision later: an engaged session is a visit that lasted more than 10 seconds, viewed two or more pages, or included a key event. Engagement rate is engaged sessions divided by all sessions.

Step 1: read the snapshot (the "how many")

Open analytics.google.com and pick your property. Home shows a quick summary.
Click the Reports icon (the bar chart) in the left bar.
Reports snapshot opens with four cards: Active users, New users, Average engagement time and Event count.
Set the date range at the top right. It's the last 28 days by default; Last 90 days smooths out a quiet week.

This tells you how many people came. It doesn't tell you where from, which is the question you actually have.

Step 2: Traffic acquisition (the "from where")

In the left menu, under Life cycle, click Acquisition.
Click Traffic acquisition. It groups every session by where it came from.
Google may show AI insights at the top, then a chart of sessions over time per channel. Scroll past both to the table.

The table has one row per channel, using the dimension Session primary channel group:

Organic Search (Google and other search engines)
Direct (typed in or bookmarked)
Organic Social
Referral (links from other sites)
Email
Paid
Unassigned (GA couldn't tell)

Each row shows Sessions, Engaged sessions and Engagement rate. Click a column header to sort.

Step 3: sort twice, not once

This is the part people skip.

Sort by Sessions first and note the top two channels. Then sort by Engagement rate and note the top two again.

If they're the same channels, easy: that's where your visitors come from and they're interested. If they're different, look harder. A channel with fewer sessions but a much higher engagement rate is bringing people who actually read. Those visitors may be worth more than a bigger channel full of ten-second bounces.

That's the whole point of this exercise. "Most visits" is not the same as "best visits," and the hour you spend on a channel should follow the second one.

Step 4: drop down to the exact source

A channel name is still a bucket. "Referral" could be one industry directory or forty random sites.

Click the Session primary channel group dropdown above the table.
Pick Session source / medium.

Now each row is a site plus a medium, such as google / organic, linkedin.com / referral or newsletter / email. The top rows are where your visitors really come from.

Changing the date range or the table dimension only changes your view. It doesn't touch any setting for anyone else; settings live under Admin.

Why your newsletter might look like it does nothing

If your newsletter or social links aren't tagged, those visits often land in Direct or Unassigned instead of Email or Organic Social. The channel looks dead when it isn't.

The fix is UTM tags on the links you control (newsletter buttons, links in your social posts). Once they're tagged, GA4 can put those visits in the right row, and the comparison in Step 3 becomes fair.

Mistakes I see most

Judging by Sessions alone. A big number of short visits can hide a small channel that brings your best people.
Reading a 7-day window. One slow week or one viral post skews everything. Use 28 or 90 days for a decision like this.
Writing off a channel that was never tagged. Check Direct and Unassigned before you decide the newsletter doesn't work.
Stopping at the channel. Open Session source / medium before you change anything. The exact site usually tells you what to do more of.

The checklist

[ ] Reports > Reports snapshot, last 28 or 90 days, read the four cards
[ ] Acquisition > Traffic acquisition, scroll to the table
[ ] Sort by Sessions, note the top two
[ ] Sort by Engagement rate, note the top two
[ ] Switch the dimension to Session source / medium
[ ] Write down where your hours go today
[ ] Move one hour a week to the channel with the most engaged sessions, and check again in a month

That last line is the business part. The report is free and takes ten minutes. The value is in actually moving the hour.

Practice it step by step

The lesson, Google Analytics 4: See Which Channel Brings Your Visitors, walks through every click above on real GA4 screens, with the numbers blurred and no account touched. It's module 1 of my free course Google Analytics 4 Basics: Where Visitors Come From. Module 2, Find Your Top Pages and Form Submissions, picks up where this one stops: once you know where visitors come from, which pages they read and whether they fill in the form. There's also a one-page GA4 cheat sheet if you want the clicks on one screen.

Practice it step by step, free: https://learn.ynteractive.com/content/ga4-traffic-sources