From SurveyMonkey to Straight Answers in One Conversation

All right, welcome to the Camp Tech Club podcast. This is our first kickoff, where we get the chance to demo real live technology that camps are running that leverage tech, all sorts of AI and other tools, so that they can focus on what really matters, which is amazing experiences on camp. In case we've never met before, my name's Eric Wittenberg from Camper Machine. And this is? Carl. You got to tell him your last name, dude. All right, this is Carl Ngan from Camper Machine. So Carl's been working steadily for a couple of years on a special, arguably our most important client, the Berkshire Soccer Academy.

It's the camp that my wife and I and Dan Zincal started back in 2010. And they're now absolutely crushing it. They have all sorts of marketing technology, but they also run a great camp. And one of the things that has led them to run a great camp is Dan Zincal's focus on survey and survey analysis, specifically for us in a one week context window to loop back and play, learn from one week to the next. So Carl has been automating and helping, or the last two years, Carl?

Yeah, I think it's the third year. This is the third year. This is the third year we've been automating analysis to help save a high value, high brain powered human from having to do analysis in the crunch time of Friday afternoon, Friday night before the Saturday reset begins for the next week. So I'm gonna kind of step back. I'll interject with some key questions or slow downs or yeah butts, what ifs. But Carl's gonna demo what Berky has been running this year leveraging the latest technology. All right, yeah, super excited to be sharing this.

I believe this is a very, all right, let me know if you guys see my screen. Cool, okay, so yeah, I'm very excited to share this with the world. I believe it, I mean, possibilities are endless, but at the end of the day, I'm starting from the results, but I'll kind of go back and go through what's going on, what's the problem, but allowing camp directors to be able to run their operations, look at data, ask questions, do everything inside, for example, Claude or ChatGPT or an AI tool that they already use, because there's no need to log into 10,000 different systems. There's no need to go through, go through different platforms just to export import data, do that clickly clack.

You can manage your camp, have all the context right inside your Claude. So without further ado, let me just- Hold on a second, hold on a second. You dropped the punchline first. You're managing survey data in Claude? Yes, now it's possible. All right, you gotta take this from the beginning. Okay. So I'll go over again. Okay, all right. Let's go. Let's start from the beginning. So normally this is what happens, right? Every week, parents will fill out a survey, and then the survey goes into SurveyMonkey.

I won't, I mean, paper and pen, whether it's paper and pen or it's digital, that's a different story. But what happens is surveys, results, several results, they end up inside SurveyMonkey. So SurveyMonkey has results week over week. Let me just unapplause this rule. There we go. So every single week we have surveys being filled out. And then without going into all the questions on the survey, what happens is someone will manually take the results of the survey, they'll export it.

And then if you look at dashboards, SurveyMonkey does give you some, I'm sorry. I mean, the summaries. So SurveyMonkey does give you some useful insights, right? They will analyze your results. They'll tell you different, they'll tell you numbers, they'll aggregate things for you. They'll put them in nice graphs. But what happens is most of the data here, as we know, like it's not enough for a camp to make, not insightful enough for a camp to make important decisions. So what happens is someone will manually take this, dump it into spreadsheet, do manual calculations, and then a few moments later, you end up with something like this.

So this is based on the results of the survey, and this is a summary page, right? So every single page has something where it's specifically, let's just say meaningful for a camp. I'm gonna go through, I'm just gonna flip through these, but Eric, if you have anything you want to... Well, hold on a second. I just wanna be crystal clear. This is an individual session survey poll. So if you went in to pull just week nine, and look at all the counselor's names, these are raw scores, right? Yes.

Because even, I just wanna, because I used to have to do this as the director earlier, I used to have to pull this and look at it. The, aren't we already using a baseline from last year was to be able to pull this and drop this data in here from week to week so that Dan could look at it the way he wanted? Yes, this is a... So it actually, the raw data is even more raw. Yeah, so before this, I actually skipped this step, right? There was this raw data that just comes directly from SurveyMonkey. It's pieces of data area where we have to... Well, we wrote custom code for it, but being able to connect those pieces of data, match the response with the right question, and then find out which survey it came from.

So that was a little bit of work, but I think those days are over. So yeah, thanks for reminding me of those tedious days. Yes. Okay, so I'm gonna flip through those little bit more tabs as like, the point is just to show how many... If you were to fill this data in manually, doing a lot of copy and paste, this could easily eat up a couple hours, right? Hold on, Carl, there's even more you're not bragging about. You're forgetting that we added... So last year, so this is... And this is one of the principles that we've kind of taken from scoping and doing a lot of these AI projects.

If you go back to one of these other tabs, like comments from one of those, one of the things that we added last year for the Berkshire team was to rank and sort and organize some of these comments. Go to one that has a category. Yeah, maybe this. So we're reading, this is the first AI step was to read the words that were written out in an elective experience feedback, and just simply categorize them. And most camps probably already know the sort of standard type of feedback. This is feedback around flexibility.

I didn't get to do the zip line nine times like I was promised or things like that. And depending on your personality as a camp director or as an operator, you may wanna hear about the negative ones first and so that you have a chance to react before another kid has the same feedback. And so this is really, again, this is kind of custom. built to Dan's brain and the Berkshire way, but they want to know about the negative things that they can react to, even though this might happen in week eight, they're going to fix it for week nine. And so this was an AI categorization based off of these comments. And then ranking the negative categorizations to show up first. So in a matter of seconds after people see this, I'm sure Dan and the director sees it, they know what happened and they know the age and session of the kid that attended and said that.

Yeah, this is so this is like, yeah, behind the scenes, what we're doing is we're using AI to to automatically categorize depending on right, depending on Dan's feedback, depending on the, you know, we see, I think what we see, we improve this, I think this is like the probably the third year, fourth year that we're 30, right? That we're doing this. So year over year, we're making it slightly, just slightly better. So next year we have more, more insights. And the goal is to be able to help Berky make better decisions, right? And know what's and be transparent in terms of like what the feedback was, what are we going to do with it? Yeah. So moving on, like just to show how many, all right, there we go. Moving on to show how many tabs there are. So if someone were to manually put these tabs together, right, it's actually, it's quite a bit of work, I would say. And it could, it could easily eat up a couple hours, right? And then also finding a normal thing and mentioned is finding anomalies, right? So sometimes in a survey, you get a comment where it, let's just say it deserves someone's attention, right? It deserves a director's attention.

But maybe that comment was missed if, or we don't react to it in time. But with a, with something that's automated like this, anything that deserves our attention will surface immediately. Right? So going back, this, this is the output that we're, that we're getting. And this is based on the SurveyMonkey input. So in previous years, this was done, this was put together somewhat manually and some, some pieces of it was done in code. But now what's going on is whenever surveys get filled out every single week, I can come in here and I can, I'm going to kind of use the same session that I was using and I'll be like, week nine's survey results are in. Can you help me grab the Excel sheet or generate the Excel sheet?

Oops. I was, let me do that again. So week nine campers survey results are in. Can you help me generate the Excel sheet? All right. So you do something like that. And because we are already connected to the, the Camper Machine, MCP server or Camp Stack MCP server, we can automatically, we can automatically pull data directly in Claude. And then we can ask Claude. So I'm going to kind of come back to this later, right? When we have the data, this is a bit, this was a big step. This kind of blew my brain when you first added this. This is your, the thing that's making this possible is that the Camp Stack built this connector that now we can add to either Claude or ChatGPT. Yeah, Claude ChatGPT or any AI tool that you use that allows you to connect to outside tools. Yeah. So this, this, that, that connector allows us to bring in specific AI

projects that we've built and give the interface directly to anyone that's using Claude, Gemini or chat. And we can customize the tool to the camp. So like in this case, all of those really nitty gritty ranking and sorting and presentation based formats that all is custom to it's a straight out of Dan's brain. And so theoretically, if there was a week 10 for them next week, and Dan or anybody, [names removed] and the team had access to this, to this connector in this chat, they could just say run the week 10 Excel sheet.

Yes. So yeah, definitely. So if there's another week or if we let's just say we wanted to update the SOP midweek or we wanted to update it mid season, we can definitely do that. And then the MCP, we don't need to do anything fancy. The MCP or to the camp director, everything responds automatically. Got it. So meaning like if somebody gets fired or hired and there's a new staff member that needs to get included in the SurveyMonkey list or taken out, that would just get dealt with inside the SOP. Yeah, that can get dealt with like instantaneous.

So I mean, if someone on boards and there's someone new they want to do inside of their Claude, all they do is it's a one line thing connected with their MCP. Do it once and then you kind of never have to worry about it again. If someone leaves, obviously, I think you can revoke access to that email to that Claude and then no longer have access. Yeah. Okay. So this is now is this are the limitations of this project just on what you've trained it to output? The limitations of good question and what is what scope are you are you asking this in?

Because if does the does this chat now have the context of the spreadsheet as well? Yes. So instead of getting a spreadsheet out of this project, this chat, and then going and uploading and asking Gemini that knows nothing about how this was created or the questions that I always ask, could you build like a stock? Like what are the things I need to know about? So now you get to save the camp directors from having to read the 35 spreadsheets for the top two things that they need to look at. I think like so in theory, yes. And in theory, you can I mean, so we can include any context we want. So I think we kind of think about it this way.

As next year, I'm hearing level up. Yeah, this next year, like for add bells and Whistles to make this product better, we can add the correct wording for it is adding it's a balance, adding the right amount of context to the conversation. So we don't blow up the context and blow up your usage. But at the same time, know enough to answer the question that you have. Know enough to give you enough insights. Yeah. So here's I'm thinking of the example. So this is something we've dreamed up together with Dan for Berky this year.

But we ended up cutting it because it just seemed like a little bit too much for this moment. But if there was a negative comment around cabin counselors, to say that there would be that would trigger a text or SMS to the camp director saying, hey, this negative feedback thing happened, you might want to look at it right away. So I'm thinking of these now upstream. Like, well, now that people can think, you know, if you're a camp director, you're looking at this and you're saying, I never I don't even ask for surveys during the summer from kids because I don't have time to look at them.

So what's the point? I mean, that's that's a realistic sort of outlook that some folks might have. And then there's other folks that are dying for the data and they want it, but they can't take the time to analyze it. The actual comment text. Yeah, I mean, there is interesting. I mean, yeah, yeah. So it can run through thousands of comments very, very quickly. But yeah, like what Eric just said, now your conversation here has full context of the report and of like if it needs to fetch additional context, it can go through our MCP server without having you just dump a whole, you know, all thousands of comments or thousands of survey results in a folder.

Yeah, because this is where I think a lot of AI has helped so many people be more productive with their time. And there are but many folks are kind of treating it like Google. And so they end up doing asking the same kind of questions. Yeah, so like we probably will blur. these out. So these are going to be intentionally blurred out in the video. But as an example, there was some flagged some flagged events maybe that took place like a slightly risky activity perhaps took place. These are things that rather than not knowing these, how many words is this, Carl, this is probably like five paragraphs, you know, maybe that's like 100 words, maybe a couple of more, maybe 150 200 for 200 words of some of like, these are the these were the bad things that happened. Just from this one week session. And like how many how many suck I don't have to ask this building. In each of those

sheets, you know, it's gonna be 200 campers deep and many, you know, five, at least five columns across and there's 30 columns, like, that level of data, there's no way you're going to be able to read these individual isolated events that fast. North Sea patterns, like that's the other thing now that so Berkshire runs nine sessions. There are different types of sessions, but you know, the girls are always 10 to 16, nine to 16, nine to 15. And so the patterns that might emerge in age groups and cohorts and cabins and, you know, you could theoretically append this data to, you know, say that does the weather temperature in influence this rally, you can go really bananas with what you ask. And it kind of comes down to what every camp director cares about.

Yeah, I think the possibilities are endless. And I'll kind of drop this in for like the for our next next podcast, right? I think the beauty of this is you can we can even connect it with other tools like your you know, your camp mind or your camp brain, right? Your CMO, where it can read multiple systems, read our read data from multiple system, be able to aggregate that data, and then kind of stitch that data together without having to build, build or use custom tools on top of it, which normally, like we can do previously without, you know, without without, you know, without AI, but it takes time, right? It takes a lot of time. It takes, you know, it takes a, maybe sometimes a developer, it takes code to do.

But now, when we, when we package this together inside the MCP server, now are inside this connector, everything is done behind the scenes, or we you know, we do the work so that you can enjoy using this directly inside of Claude and not have to worry about the technology underneath. Yeah. So this is, you know, synthesis, you know, across all these things. Have you tested it for synthesis across multiple weeks and sessions? I've been asking it random, random questions. And a lot of times, like it hasn't, so it hasn't hallucinated on me yet. And I've, and if I tell it to give me the sources, it'll actually tell me like where this thing where this came from, which tab, which role?

Oh, really? So it's not, it's not just like, okay, I think this is bad, right? It's like, okay, this is, this is a negative comment. This is bad. This is where it came from. This needs your attention. Hmm. And I think the, you know, I kind of want to want to say this, like, I feel like AI is most useful when it's invisible. Hmm. Yeah, this certainly makes you know, you're just surfacing the very last the last 10%, maybe 5% here inside of Claude. Um, can you do if you can you share one of this, the food, comment or surveys because I know as a camp director, you know, when kids don't eat everything else is worse. You know, everything else goes bad. And so this some of these questions for most favorite breakfast food, most least favorite food.

And looking at some of these age groups of 15 plus versus 12, you see this negative ratings. As we can imagine the 12 year olds or older kids might get a little bit pickier. But so could you theoretically ask it at this point through Claude to read the food comments and say, what are the eight year olds like or not like across the nine sessions. You know, I would do these on these separate spreadsheets. Could you even reference other questions from the respondents like first year nine year olds?

Yeah, I think you can do I mean, let's do nine year olds because I think this is a little bit more interesting. Yeah. Oh my gosh, that was a foreign object in the pasta. Well, the Berkshire Soccer Academy is a world class institution. These are all freak incidents. Probably nine year olds hallucinating, not Claude. Can you help me go through the food comments in the in the responses and tell me what the nine year olds like or don't like? I didn't have my dad. I did not activate my speaker. Okay. Oh, you're voice chatting into this. Okay.

Oops. Oh, can you help me go through the food comments in the survey and tell me what the nine year olds like or don't like with their food? Sounds good. That's quite clever. So yeah, while this is happening, I think the point where we want to get across is like it's helping someone get from A to B in the fastest possible way, right, in the most reliable way without friction. Yeah.

Now these in the same way. Often, I feel like many of our projects seem to be disconnected, but then they end up meeting in the middle. How do how do you see camps being able to use like a survey project with another internal custom project that they use for a backdrop or labels or things like that? Okay. Well, actually, okay, let's let's go. Welcome. Welcome back to let's read the answer first here. What did the nine year old say about the food? So what did the nine year old say about their food?

So it definitely these are independent reports. One first hand. So I love this is amazing. The second hand accounts of traumatic food experience. Yeah. So, you know, I definitely picked up the foreign object in in food. Right. That picked up that comment. Yeah. So, I mean, this is we're we're looking at thousands of camper weeks and looking at the four kids that had something, you know, a strand of hair in their grapes. So now we ask our nine year old hallucinating or hallucinating.

So there were a couple of instances. What is essay that session eight? Yeah. SESSION eight. So so it's abbreviated to know the weeks. So basically once or twice a week, there's something random that a kid doesn't like in their food. Now you could could you cite this? Could you could you pry in? So like that week eight, they somebody thought they had the milk, their milk was bad. Who who actually said it? What cabin were they in? And I could you could you could you bring us to the raw data?

Yeah. So I will. Interesting comments on session eight. Can you help me dig into this a little bit deeper? And let me know where the comments came from. Can you find out what cabin? Etc. All right. So while this thinks about that, we're going to have to flip this and do some bragging and hear and analyze some of the greatest comments about, you know, why Berkshire was so amazing because. Yeah, there's a I mean, there's definitely.

The actual food. This is interesting. The actual food preferences. They love it. Nearly all. Of his unqualified praise, it was delicious. So yummy. Yeah. Yeah. So for the most part, for the most part, it's good. Right. It's scrub a little bit. What did it say at the very bottom there? So that the age group. So. You want the same age? I know. Kevin. It's just look at it's offering another age group analysis. I mean, yeah, the next question, you know, if we were going to be doing real post camp lessons learned, whether we change for next year, which is the season, you know, as soon as camp directors come out of the woods, and they're not in a sort of silent on the beach week. The first thing they're doing is looking through what do we have to change for next year? What can we add or do or do differently? And I would be super interested looking at all these things across age and session differences. Okay, so

this is telling us that we're so for the food comments, we're not able to find the cabin, but for but it's telling you, okay, for the cabin comments, final comments and like mental skills, they do carry a cop cabin name. So for those were able to backtrack and reference the the cabinet. Kind of but but the the actual comment for that food one. Oh, here's here's just more stuff. So yeah, cool. The actual comment, there was like a hair in the grapes or something like that. It could just because I'm sure there are some I would say AI skeptics out there that are thinking, is this even real? Does it, you know, is it pulling the spreadsheet? Does it change things and embellish? And so if some one of the takeaways was somebody found a hair in their grape during session nine, there was a nine year old, they, you know, therefore were drawing conclusions, nine year olds might need special attention.

But but I will. Yeah, I will say I will just add another, you know, sentence like, I will say this like, so if we really do want to find like, okay, where did this come from? But I think that's like engineering on like, it's not Claude, right? But that's us. We haven't we haven't said site the cell. Yeah. Okay. And kids are so funny. An eyelash on the pizza. An eyelash on the pizza. How do you know it wasn't yours? Okay, cool. So this is this is these are red flags, some orange flags, some yellow flags. And the idea is save time. And there's nothing worse as a camp director to find out about a problem from the parent when they call in. And so every second you can get to things earlier, you can get full details, you can, you know, maybe even kind of address the root problem for us bigger.

So this is probably not the most exciting thing for some folks. But for some people, this is where they live. Can we just ask it? I feel compelled to ask it about some of the great like, I would just say maybe across the ages and sessions, like, could it make a graph for like the thing that was the best for each age and session? Like I usually I ended up asking Claude to there you go. Then you got me to speech that and that's that's that's real engineering laziness is a good highlight. So I'm constantly asking for little matrix to get built. Yeah, so you know, I'll say this, like the possibilities are really endless, like, you know, piggybacking what you said earlier, like, you know, we separate it into positive, negative because of Dan's preference, but we can easily separate it into red bucket or red flag, a bucket for red flags, orange flags, etc. Yeah. Yeah. And even, you know, again, as a camp director,

especially in a weekly session, even a two weeks, two week sessions, if you're getting camper feedback, I want to be able to give some sort of score and praise like this is as an owner operator, I want to fix the worst things first. But as somebody that's dealing with frontline staff that are living in cabins, you want to hype them up and get them also excited. Be like, Oh, my gosh, these people love you. The girls came back, you know, praising you so much. They just loved the experience. They thought x, y and z about you. You could even spin this into create a bullet one hour meeting agenda to review policy. And this is where the sort of like the the machine starts really helping you. You could have a specific format for a half hour after dinner meeting where you address what went not so well, what went really well and a goofy game to fix it in. If you had your mission vision values aligned, staff training to recite, like you could really go crazy and have a staff training skill augmented here. So this is a cross grid.

Yeah, well, Claude works on that. So normally, it depends on, you know, also say this, like I put an Opus 5 high right now, but depending on the models you use, like it could be a lot faster or if you wanted to think more in depth and you can use like, you know, a Fable 5 or something like that. So yeah, it's really, there's a lot of control. There's a lot of flexibility in terms of like, how hard you want the model to think on like, I mean, even if you put it in Haiku, like the lowest model, you'll still get the spreadsheet, you'll still get right. I mean, because our tools will still I mean, the tools will work the way they're supposed to. But I think it's everything that comes afterwards, right? It's delivering insights rather than just a character, the spreadsheet.

So I want to do a little bit of foreshadowing for some of our subsequent conversations that we might as we revisit this. Right now, we have Oh, this is fancy. Yeah, as a fast little graph. I mean, obviously, I think it needs a little bit of Yeah, it needs some UI, UI work. We haven't really asked for any of this stuff. But this is kind of what I was looking at. So okay, look at the age. Right? Yeah, so these are the big takeaways that certain kids, certain age groups have big differences. So the 15 year olds love cabin cup more than the Berky tournament. Yeah, okay, so that's okay.

The eight year olds think that yeah, this makes sense. Seeing the, you know, your your new idol be goofy on stage, and then you know, be the one that's tucking you in, men taking care of you in your cabin is amazing. That doesn't move the needle for a 15 year old. So this is an example of like an additional output, though. Right? Like, are you thinking about this in in terms of inputs and outputs? You're saying am I thinking about this in terms of inputs and like, like, like the analysis, the analysis of because really, this is this is really a second window. Now that we're mostly done with the summer, and that there's and that there's there's in summer feedback responses, which is really what that survey tool was primarily built for. Then there's this additional fall wave of, okay, what are what were some bigger trends that you might want to respond to? Maybe the 14 and 15 year olds, like, opening campfire wasn't even listed as a top thing for 15 year olds. So like, maybe they have a different campfire that ends earlier, or just something like that. So these

big programming changes that might be informed by this data. I'm just realizing that the output that we've mostly highlighted has been the in summer response to in summer surveys. And this is there's probably a desired output that is not really, you know, you and I, every every human might have a different version of what they want to see or what they might ask about. But there's a probably a new output that we could use this data for. If only Berkshire had, you know, this MCP connector in their accounts, which they will soon. Yes, and I'll in date, like, so data is data, right, we have access to all the data, right?

Because of the MCP server, we made data accessible through Claude and how we want to look at it or the perspective we want to look at this data and the insight we want to get. I mean, you can, there's so many ways to skin the cat, right? You can look at it from the perspective of, okay, in session, or when we're in session, what are things we have to react to versus like what you said, Eric, outside of session, what are some of the big improvements, big changes that we can potentially or consider potentially consider improving on? Yeah, yeah. Super cool. I mean, these things, the needs and outputs are to keep kind of like emerging for what we could do in different times, different audiences.

One thing that I'm thinking about, because I know I've been looking at the Berkshire Soccer Academy's Camper Machine and enrollment data as it comes through. I would be so curious if we could match up and see which of these responses are re-enrolled already. Ah. But then we're getting into enrollment, CampMinder sort of API integration with the MCP server to say, hey, how many of the kids that were enrolled in 26 are back at this point and which sessions and bunks were they in? Yeah, and that would be a very, and that's like the conversation for next time, right?

Where we have multiple tools. So now we're just talking about, like for this conversation specifically, we're focusing on, we're honing in on a SurveyMonkey. Yeah, summer camper survey analysis, really what it is. Yeah, summer camper survey analysis, but we can also stitch that together and aggregate it with enrollment information. We can aggregate it with other pieces of information, right? Enrollment, cabin information, et cetera. And then be able to make useful insights out of that data. Yeah.

And so if you're listening to this and don't have access to the video right now, we'll share the show notes that are available on the campermachine.com. If you click insights, you can find the podcast. You'll be able to see a couple of screenshots and scroll through and watch selected clips if it was interesting to you. This is the beginning of all of the tools sort of molding and melding into one. And so these MCP servers that allow us to connect, Carl and I for years have been moving data in and out of tools in itself.

And now the data continues to flow, but we're building tools and outputs and results based on top of that. And so if you're curious if this conversation has sort of opened your mind of what's possible, what you would wanna build for your camp, if you have a specific request, there's some things that we do stock now, survey analysis like this. And then there's kind of anything that you wanna dream up. So we've got an option to just kind of talk to Carl and I and we'll figure out what is possible and what makes sense to build for you. And the only thing I'll add to that is if you're thinking about it, if you're starting, if you're vibe coding yourself or you want someone to do it for you, these things have in our experience are very iterative and take time to lay out and develop and progress.

And so even if you need a thought partner, Carl is definitely the guy to help you there. So you can reach out and learn more about that. We'll share a link on campermachine.com. You can find a way to connect and book down a chat with Carl and or I about your AI dreams. All right, gang, we'll talk soon. Thanks, Carl. All right, thanks, Eric, good stuff. All right, see ya.