So What? How to make your own 2025 wrap-up

Katie Robbert – 00:35

Well, hey everyone. Happy Thursday. Welcome to So What?, the Marketing Analytics and Insights live show. This is our last show for 2025. Fellas, how you feeling about it?

John Wall – 00:46

The holidays are dry.

Christopher Penn – 00:48

It’s been a good year.

Katie Robbert – 00:55

Well, good. I’m glad to hear it. Also, fun fact, apparently StreamYard now has timers on their banners and that’s really annoying. So happy New Year, everyone.

We thought we would end the year doing something a little bit more fun. We always try to do something practical and useful to teach you how to use all of the tools that have been thrust upon you, and this week is no different. Since it is the end of the year, we want to show you how to make your own 2025 wrap-up.

You can use what we’re going to show you to say, “Here’s everything we did this year.” I’m assuming there’s a lot of different use cases where you can say this is our end-of-quarter wrap-up or this is my own personal career wrap-up. Basically, you’re seeing Spotify, Apple, and LinkedIn—everybody has their “Year in X” and what you did. Wouldn’t it be great if you could build your own? So Chris, where should we start today?

Christopher Penn – 01:57

The hardest part of this entire thing is the data—getting the data and getting it into some format that a machine is able to process. Once you have that, the tools to do a wrap-up are actually pretty straightforward, and we’ll show probably the easiest one for folks to get started with.

The hard part is where’s the data and what data do you want to use? In rehearsing for this episode, I did a version where I used some of Trust Insights internal data to do an internal wrap-up and I realized I can’t show any of this on the air because it has tons of confidential information. I’m like, “Oh, you won this client for this amount of money.” That’s true, and we celebrated that, it was a good thing—but I can’t show that on the air.

So I decided we’ll do the Analytics for Marketers Slack group as a way to do that as an example. That’s safer data than proprietary confidential information. First, you have to get the data. If you are an admin of an environment like that, you can literally just go into the admin portal, hit “export the year,” and it will give you a big file.

If you don’t have that, then you need to figure out how you’re going to get a hold of the data and what you want to use. Say if you wanted to do your own version of LinkedIn Wrapped with stuff that you actually cared about, you could go to LinkedIn, go to your settings and privacy, and say, “export my data.” Then, obviously, go through the spreadsheets that are generated, cut all the stuff that isn’t 2025, and use that. You could even go into your email and put all of your emails for 2025 into a folder, export that, and put that into a system to see the “greatest hits” of your inbox.

Katie Robbert – 03:56

You’re poking fun at it, but there are certain people—think about sales, for example—where your KPI is how much outreach you did. If they’re not tracking things in a system like HubSpot or Salesforce, that is a really good way to get your hands on that data. Think about the KPIs that you’re being held to as an individual and where that data might live. Email is actually not a terrible use case; it’s a matter of the context of what it is that you’re looking for.

For John, for example, your whole role is making those connections and getting the contracts in the door. Looking at the emails that John has sent is not a terrible idea. We’re not actually going to do this, John, don’t worry. But if we were a larger organization and weren’t intimately aware of what we’re all doing at all times, it would be a great way for John to present, “Hey, here’s everything I did. Look at all of the cold outreach I did or look at all of the connections that I made.”

Moving into the next year, it is a great way to say, “Here’s what I want to make for money and here are the goals that I want to set.” So I just shot down your idea that it was a bad use case.

Christopher Penn – 05:20

It’s not a great use case for something I would do on the air.

Katie Robbert – 05:25

And that’s okay. But I think in general, when we talk about your own wrap-up, yes. On air, I’m interested to see what you pulled for Trust Insights privately, but publicly I agree.

Christopher Penn – 05:40

You’re absolutely right. A big thing that we do that for with emails is for client review. At the end of the month or end of the quarter, we’ll export the client mailbox within our system and say, “What did we do with this client? What were the greatest hits of the quarter?” Particularly when you have clients who say on a regular basis, “What have you done for me lately?”, we can say, “Here’s what’s happened.” So it’s a great way to do that.

Let’s walk through the example of how to do this. I will say it again: the data manipulation part is the hardest part. Everything else after that is easy. We’re going to use the Analytics for Marketers Slack group. When you go into Slack as an administrator, you get a zip file. When you open up that zip file into a folder, you get all of your Slack channels and then you get a JSON file for every single day of every single channel.

You get 1,731 files all put together in one massive folder. This is a pain. If you’re going to use a tool like Google Gemini or ChatGPT, 1,731 files is a lot. However, if you are skilled at the command line, you have the ability to process this data to glue it all together and put it in more manageable chunks.

If you’re not sure what your computer has, this is a great opportunity to talk to ChatGPT or Gemini and say, “I’ve got 1,700 files in nested folders. Help me write a command line command that will glue them all together into one text file.” I’m going to go ahead and do that. Now I have one great big file. This thing is 27 megabytes of text. You all were real chatty this year.

Katie Robbert – 08:08

If you want to join the conversation, you can go to our free Slack group at TrustInsights.ai/analyticsformarketers. It is a very chatty and engaged group, and that is something that we really take pride in.

Christopher Penn – 08:23

Exactly. Now here’s the problem: if you try and load that file into just about anything, it is going to blow up. Let’s go into Google AI Studio because you can see just how big that file is. AI Studio is Google’s developer AI portal. It allows you to have the same access to all the models in the Gemini ecosystem.

This is 8.7 million tokens. That exceeds the memory of Gemini; it can’t process that. There’s no AI model on the market that is generally accessible that can process that much information. As ridiculous as it sounds, we now need to take that big file and split it into smaller files—but fewer smaller files.

We started with 1,700 files. Again, using the commands built into your computer, we’re going to split this into files of 20,000 lines each, which for most plain text is going to be about 1 megabyte.

Katie Robbert – 09:39

While Chris is doing that, I think it’s a good opportunity to remind people that this is why you always want to do some semblance of requirements before you start exporting things and putting them into other systems.

Chris, one of the things I noticed that when you exported the data from Analytics for Marketers, it exported every channel. I would argue that we wouldn’t necessarily need to look at the data from every single channel. There are a few channels in there for specific uses that we don’t necessarily need to bring in.

That’s why you want to have a user story. As the admins of Analytics for Marketers, we want to know what the top conversations were for 2025 so that we know how to engage with our community in 2026. Having requirements is going to help focus down the data that you’re looking at, because we all have a lot of data and it can get really unwieldy. John probably has the most data of all of us, so we want to make sure that he’s focusing down his conversations so that we can do justice for his 2025 wrap-up.

Christopher Penn – 11:00

All right, so we end up with 35 files. Each file is about 800 kilobytes. That’s a lot more manageable. 35 files is something that Gemini still can’t handle because it’s the same amount of text, even if it is split up into pieces.

So we would use a tool like NotebookLM. NotebookLM can handle that much data very easily. I’ve gone ahead and put all those files into NotebookLM. You can see them all listed here as sources. We start off by saying, “This corpus represents the Analytics for Marketers Slack group, a private community operated by Trust Insights. 4,500 marketers have regular conversations about analytics, data science, and AI. We want to provide a summary of the top 10 most impactful conversation topics in the corpus in descending order.”

Impactful means that many members participate in the conversation or conversation threads were unusually long. It goes through, processes all 27 megabytes of data, and says your three main topics are AI practicality and tooling, operational and strategic challenges, and non-work and fun conversations.

For those who missed previous episodes of our live stream, you can create media on the right-hand side, which are great for things like slide decks. A prompt for the slide deck might be, “Make a top 10 topics Analytics for Marketers slide deck using the Trust Insights brand standard guidelines. Count down the topics in reverse order from 10 to 1, with each slide covering one topic.” I actually pasted the output in so that you get exactly what you want.

Are you ready to see the wrapped for Analytics for Marketers, Katie?

Katie Robbert – 13:45

I am.

Christopher Penn – 13:48

All right, let’s go into what analytics marketers really talk about. The top 10 things include AI practicality, operational challenges, and community connection.

Number 10: AI tools for content creation and adoption. Users evaluated the feasibility of local AI setups and shared which models are best for tasks.

John Wall – 13:46

Believable.

Christopher Penn – 13:48

Number nine: Outdated sales and marketing tactics. Interruptive tactics like cold calling need to be retired. I remember that thread. John, this was something you put in the Rants channel.

John Wall – 13:58

Yeah, this is like my weekly complaint post. So I’m not surprised.

Christopher Penn – 14:06

Number eight: Non-AI marketing tactics. The irreplaceable value of direct, authentic conversations and listening first. I think this was one of your threads, Katie.

Katie Robbert – 14:20

To be fair, they’re all my threads; I ask every question of the day. But this is a topic that I know more about, so it makes sense that I likely weighed in on it more.

Christopher Penn – 14:36

Number seven: Marketing analytics terminology and confusion. Frequent confusion between different terms, differences between metrics and KPIs, and frustration with Google’s inability to name any product sensibly.

John Wall – 14:52

That sounds correct. It used to be bad, but this year is just times 100 as far as horribly named stuff.

Christopher Penn – 15:07

Number six: Frustrations and critiques of AI. Concerns about hallucinations, the complexity of model naming conventions, and the societal and environmental impact of widespread AI adoption.

Katie Robbert – 15:23

We’ve heard a lot of our members chime in about that, which is a big topic in and of itself.

Christopher Penn – 15:31

We could spend entire shows just on that. Number five: Technical SEO and large data corpus processing. Leveraging tools like NotebookLM, technical tips for managing context windows, and advanced techniques like “vibe coding” for exploratory research.

Katie Robbert – 15:48

It’s been a minute since we’ve talked about vibe coding, but technical SEO and AEO are very top of mind.

Christopher Penn – 16:04

Number four: AI automation, tooling, and implementation challenges. Workflows, automation, agent-to-agent tools like Opal, and Claude’s agents.

Katie Robbert – 16:16

I’m seeing a theme, and I’m sure it has a lot to do with the types of questions that we ask because we want people to respond with their shared experience. A lot of times the shared experience is, “I’m struggling with this thing, can somebody help me?”

Christopher Penn – 16:39

Exactly. Number three: The great Thanksgiving side dish debate.

Katie Robbert – 16:45

On Fridays, we always ask a non-work related question. It’s as professionally unhinged as we can get while keeping it on the rails. People in this community love to debate things about food specifically, which is hilarious. This community is very passionate, for good or ill, about their food.

Christopher Penn – 17:29

Exactly. It’s not a real holiday meal if you can’t argue passionately about the merits of green bean casserole.

Number two: Operational bottlenecks and constraints. The biggest hindrances to workflow efficiency, internal political hurdles, and feeling overwhelmed with the constant news cycle.

John Wall – 17:55

I love the quote about Agile being way too rigid. That’s awesome.

Katie Robbert – 17:59

I think I said that. Anyone who wants to say we have to do true Agile methodology should buckle up. It is not easy. It can work, but you have to have everyone in lockstep.

Christopher Penn – 18:37

Does anyone want to take a guess at what the number one topic of the year was?

Katie Robbert – 18:46

If it’s Taco Bell, Kelsey is fired.

Christopher Penn – 18:51

It is fast food fries.

John Wall – 18:56

That was the Taco Bell thread.

Katie Robbert – 19:35

She has turned every thread into the Taco Bell thread. It is very funny to me how polarizing certain foods are. I appreciate that everybody in the community comes to the table with an opinion and feels comfortable sharing it, knowing that it’s a safe place to do that.

Christopher Penn – 19:35

Exactly. That’s probably the most important value of the community, particularly this year, which was very stressful for a lot of people. It’s having a safe place where you can let your hair down and be among peers. These patterns—practicality, operational challenges, and community—mirror who Trust Insights is.

Katie Robbert – 20:16

I’m the one in charge of asking the question every day to start the conversation, so obviously I’m going to ask questions that relate to the work that we do. I try to give a well-rounded set of questions, but even when we say, “Tell us the good about AI,” it naturally turns into frustrations and critiques because that’s the human experience with these tools right now.

As someone trying to figure out how we can best support people, you can’t beat this kind of free market research. This is directly from the mouths of people who are in it. It further confirms the direction of where we want to take our services and our ICPs. This is priceless data.

Christopher Penn – 21:22

Even though the group has been called Analytics for Marketers for years, we had very little actual analytics discussion because of the 800-pound gorilla in the room—AI. We want to meet people where they are, and where they were this year was around all things artificial intelligence.

Katie Robbert – 21:56

I do ask non-AI questions, but someone in this room always takes over the thread with something AI. I could say, “What’s your favorite color?” and someone in this virtual room would give me an AI response which would derail the thread.

John Wall – 22:29

She’s onto me.

Christopher Penn – 22:34

We took the data, put it into NotebookLM, and then created the output. We have the slide deck, an infographic version, and you can also do video explainers. If you want to do your own 2025 year-end summary, NotebookLM is the tool I recommend you start with.

You can use ChatGPT or Gemini’s image generation models to roll your own if you have the data. The hard part really is: how do you get the data out of the system? If you have no data but want to make something anyway, you could get credible third-party data. An example would be the Reddit API. You could grab a subreddit’s conversations for the last year and summarize that. We do that every quarter for ourselves and our clients to see where the broader community’s heads are at.

Katie Robbert – 24:38

I’ve used one of those notebooks and found it very helpful to understand people’s opinions of the C-suite in terms of AI adoption. Nothing in the conversations surprised me—there’s a lot of people feeling like the C-suite is detached from the day-to-day—but it was nice to have the data to support that.

Coming back to Analytics for Marketers, I can see a lot of other ways you could be using this data. If you run a community and want to promote engagement, you could ask your notebook, “Who are the top 25 most engaged community members?” or “What are the non-AI topics that were bubbling to the top?”

We can use that infographic as a social post to remind people that this is a great place to have conversations. I guarantee our community members are going to be thrilled to see the kinds of things they’ve been talking about all year.

Christopher Penn – 26:12

You could take the top places people want to have a conference or whose general fast food is better if you want to start fights between Kelsey and everyone else.

Katie Robbert – 26:33

We’ve also asked questions like, “Who are the influencers or thought leaders you recommend following?” When you look at a community, that data feels hard to pull out, but when you put it in a format like this, you can see who the community thinks are influential voices.

Christopher Penn – 27:06

The person I was most connected to this year is on this call with me—John.

Katie Robbert – 27:20

LinkedIn has sent out their version of 2025 Wrapped. It shows the emoji you use most, how many new connections you have, and the person you connected with the most. Unsurprisingly, Chris and I connected with each other the most on LinkedIn.

Christopher Penn – 27:46

NotebookLM is powered by the brand-new Gemini 3 Flash now, which means it is very good at reading JavaScript Object Notation, or JSON. What we provided in here was not plain text; we provided JSON files.

A tool like NotebookLM can read this data with a good deal more accuracy because it’s like a spreadsheet in a version it is capable of reading. JSON is easy for Gemini to understand and it does a better job of doing basic operations like tabulating. If you’re going to do these kinds of wrap-up things, remember to provide data in that format if you can.

Katie Robbert – 29:26

Could you export data and then bring it into a Google Colab and ask it to transform it into a JSON file?

Christopher Penn – 29:35

Yes, and you can even do that in regular Gemini. As long as the file doesn’t exceed the size of what it can work with, it speaks that language well.

Katie Robbert – 29:46

John, I’m adding to your to-do list: when are we going to see the top Marketing Over Coffee conversations of 2025? I would imagine that the conversation tends to skew towards AI. Do you feel that’s true of the interviews you’ve done this year?

John Wall – 30:21

No. The problem we have with Marketing Over Coffee is that close to 70% of the content is so topical that it’s stale within three months. The amount of reuse is astonishingly small. I did three years of material once and less than 20% of it was even relevant anymore; so much of it was just complaining about MySpace.

A Marketing Over Coffee wrap-up would pretty much just be stats—which interviews and episodes hit. Even that is weird because once in a while an episode will get picked up in some random channel and get a ton of action that has nothing to do with the content. Coming up with a “best five things of the year” is definitely a challenge.

I did decide one year to do an outtakes and funny clips episode. It took so long that I am never doing it again. If I had hours over the holiday, I could load up 50 episodes and have AI pick the five funniest clips, but then I’d be relying on an AI’s opinion of what’s funny. It’s a rough problem to crack.

Katie Robbert – 32:25

That goes back to Chris’s point: you can make a wrap-up of anything provided that high-quality data exists and has the correct context.

Christopher Penn – 32:49

NotebookLM supports MP3s, so you don’t even need to transcribe them.

John Wall – 33:05

It’s not about the MP3 as much as it is the audience feedback. There’s no way we can track people laughing in their cars. The other thing is that sponsors don’t want their data out there, like how many leads or clicks they got.

Katie Robbert – 33:35

But at a high level, you do have the data for the top five topics discussed. You can subscribe at marketingovercoffee.com. The podcast has been going on for what feels like forever—you’re entering your 19th year?

Christopher Penn – 34:12

Yes.

John Wall – 34:12

2,308 years, actually.

Katie Robbert – 34:17

If you were to pull the top five topics, my guess would be equipment, AI in the news, and some books.

John Wall – 34:44

That would be worth doing—to see the percentages of topics and the breakdown. I’d also be interested in the demos on the authors, like how many women versus men were guests, how many books, and how many vendors.

Christopher Penn – 35:04

Let’s talk through this. John has a really sensible file naming system where the name of the show every week is the same except for the show number. I could say to Gemini, “Create a list of URLs that decrement by episode number starting from 898 and show me the last 52.”

There is a terminal-based utility called Wget, which is basically a browserless browser that can fetch files. I gave it 50 some odd episodes of Marketing Over Coffee to download, and I have all of these episodes. NotebookLM supports MP3s, so I just drag them in and ask, “What were the top five topics that Christopher Penn and John Wall talked about in 2025?”

In the time you and Katie were talking, I was able to execute this. We have:

1. Generative AI, LLMs, and AI strategy

2. Gear, hardware, and tech

3. Events, conferences, and travel

4. Economic, geopolitical, and market issues

5. Social media platforms and algorithm changes

Katie Robbert – 37:33

That’s a really helpful pull of information because John can use that to promote the show. That can become an infographic.

Christopher Penn – 37:53

I can actually cue that up. I can also ask for the top topics John discussed with guests, excluding shows where I appear. Gemini is able to digest this stuff.

As for the gendered stuff, I’d want to experiment with that offline because every model has biases and we don’t explicitly say, “My woman guest today is…”

Katie Robbert – 38:49

There are some ungendered names that someone of any gender could use, so it’s very hard to say for sure.

Christopher Penn – 39:13

The top five guest topics were:

1. Generative AI implementation strategy

2. Marketing strategy, branding, and consumer behavior

3. Social media management platforms

4. Operational technology and software development

5. Social and cultural integrity

John Wall – 39:41

It definitely reflects that nothing is too nerdy or too tactical for this show. It is all deep-dive stuff.

Christopher Penn – 39:52

Except for that doofus this morning that sent us a pitch saying, “As a fellow spa owner…”

Let’s take a look at the infographic. There you go—top five topics.

Katie Robbert – 40:16

In terms of infographics, John, that’s not bad. I hope to see that on social media.

John Wall – 40:20

I love number four because we used to laugh at AI-generated images of gear with wires sticking out of it. This clearly shows a drone, an Apple Watch, a mic, and headphones. We’ve nailed it.

Christopher Penn – 40:39

If you wanted to, you could provide lists of individual YouTube videos to NotebookLM and see the topics we covered on this live stream for the year. There’s no shortage of ways to do this as long as you can get the data.

Do this whenever you are up for a review. Take your inbox, your Slack, Asana, or Jira—export your data and say, “Here’s what I did in the last year. Give me a raise.”

John Wall – 41:43

Prove your worth. That’s a huge time saver for any annual review process.

Christopher Penn – 41:49

If you’re a manager, consider doing that across your team. One of the things I suffer most from is recency bias—remembering the last few things somebody did and not the whole record from the entire year. It can help you overcome that bias to see what a person did for the whole year, not just the last two weeks.

Katie Robbert – 42:34

Think about the data you have and the points you want to make. Make your own 2025 wrap-up and join the conversation in our free Slack group. You too can weigh in on the best cut of French fry or the best Thanksgiving side dish. More to come in 2026.

Christopher Penn – 43:30

That is going to do it for this episode and this year of the So What? live stream. To everyone who has watched, thank you so much for being here with us every single week.

We are on Thursdays at 1:00 PM Eastern time most weeks. We hope you participate in the Slack group and the email newsletter that comes out on Wednesdays. If you have questions for our podcast, which we record on Mondays and air on Wednesdays, we’re more than happy to answer them. Thank you for being a part of our community. Take care and we’ll talk to you all in 2026.

Be sure to subscribe to our show wherever you’re watching it. For more resources, check out the Trust Insights podcast at TrustInsights.ai/tipodcast and our weekly email newsletter at TrustInsights.ai/newsletter. See you next time.