Luxury Bag Restorer Finally Analyzes Her Sales Channels with AI

Lynn runs Dr. Bags, a luxury bag and shoes restoration business in Singapore - 11 years, 3 stores, about 30 staff. Her team is already stretched thin building a new ERP system, training staff, and handling daily operations. Analysing marketing data? That would take two days nobody has.

We built her a marketing dashboard in under an hour using Claude Code with a month of her deal pipeline data. It wasn't perfect - and that turned out to be the most useful part.

What you'll hear in this episode:
- Why SME owners know they should analyse their data but never do
- Building a marketing dashboard from raw ERP data, live in one session
- Discovering that WhatsApp was getting credit for wins that started on social media ads
- A pile of deals stuck in limbo that nobody had noticed
- Why business acumen matters more than any AI tool - and how Lynn proved it

Hosts: Eric Tan (non-technical builder) & Yaohong Ch'ng (engineer, Superuser HQ, ex-Stashaway head of Data)
Guest: Lynn Kee - founder of DrBags, luxury bag and shoes restoration

Got a problem you want us to solve live? Fill out the form:
https://forms.gle/DSyLzPAoR6x2M4Np9

Listen/watch:
- YouTube: https://www.youtube.com/watch?v=w5u9ShYoodc&t=56s
- Spotify: https://open.spotify.com/episode/43mFsd6uOHqcYuE2c6nmkp?si=2MxLxpcFQyu1fvPYIiO6BA
- Apple Podcasts: https://podcasts.apple.com/us/podcast/luxury-bag-restorer-finally-analyzes-her-sales-channels/id1877603539?i=1000759069037
- Snipd: https://share.snipd.com/episode/829084df-c22c-4833-96e6-9fb7714b7275

Connect with us:
- Eric on LinkedIn: https://www.linkedin.com/in/erictisme/
- Yaohong on LinkedIn: https://www.linkedin.com/in/yaohongchng/
- WhatsApp community: https://chat.whatsapp.com/Dmp5eEEsAZhJTB6LjcIG3c?mode=gi_t
- Substack: https://substack.com/@coderiff
- Email: code.riffs.ai@gmail.com

Tools used in this episode:
- Claude Code: https://claude.ai/code

Tools we use:
- Buzzsprout (podcast hosting): https://www.buzzsprout.com/?referrer_id=2371679
- Snipd (AI podcast highlights): https://get.snipd.com/pAbF/36jzrvki

LEARN ALONG - Glossary:
- ERP (Enterprise Resource Planning): Software that tracks everything in your business - inquiries, jobs, inventory, delivery. Lynn's ERP follows a bag from first WhatsApp message through cleaning, QC, and return.
- Dashboard: Charts and numbers on one screen so you don't dig through spreadsheets. What Lynn wanted instead of two days of manual Excel work.
- Attribution: Figuring out which ad channel actually brought a customer. Lynn's data said 89% WhatsApp - but many customers found DrBags through Instagram or Facebook first.
- Funnel: The journey from first contact to closed deal. Lynn's funnel: new inquiry to engaged to won or lost. 61% of deals were stuck in "Engaged."
- Data cleaning: Getting your raw data accurate before doing anything with it. Yaohong called it "plumbing" - if your data isn't clean, your dashboard lies to you.
- POC (Proof of Concept): A quick prototype to test if an idea works before investing serious time or money.
- Plan mode: A Claude Code feature where it reads your data, proposes an approach, and waits for your OK before building anything.

ABOUT:
Code Riff - messy real-world problems, solved with AI, so you can too.

One of us can't code. The other's been coding since he was 13. Every episode, someone brings us a real problem and we try to solve it with AI in one hour live.

Transcript

Trailer

SPEAKER_05 0:00

We need to understand that if we want AI to be helping us, we need to put in our effort and our our time to correct Yeah, to build the system first.

SPEAKER_02 0:15

Yeah, let's do it.

SPEAKER_08 0:17

Yeah, you see. More appropriate for the brand today. Okay.

SPEAKER_05 0:30

See now I'm excited. I think what's really good about today's exercise is that I'm gonna question all the numbers, right? Yeah, what's the lead source? What's this, what's that?

SPEAKER_00 0:40

Yeah, you'll try to understand your data a bit better.

Meet Lynn - 11 years restoring luxury bags and shoes

SPEAKER_08 0:43

So actually, yeah that's the underpin of the whole thing of automation, right? Your data, data is the most important thing.

SPEAKER_00 0:57

Hi everyone, welcome to the Code Rift Podcast. I'm Eric, the vibe coder, and with me I have Yao Hong, my AI orchestrator, and Lin, who is the founder of Dr. Bax. So welcome Lin. Uh, do you mind just sharing a bit more about yourself and Dr. Bags as well?

SPEAKER_05 1:17

Right, thank you so much, Eric and Yao Hong, for doing this with me. Um, I am the owner of a luxury restoration company. We specialize in cleaning, coloring, and repairing bags and shoes. Um we are in the luxury brand segment. So we on a daily basis there are ladies, men who bring their bags and shoes um for us to upkeep and hopefully do some magic for them as well. Yeah. So that's the business that I have been in for the past 11 plus years now.

unknown 1:51

Yeah.

SPEAKER_00 1:52

Wow. All right, cool. And and before this, we had a nice chat as well. Could you just share a bit about your motivation for you know getting on this call with us and sharing with the rest of the world as well?

SPEAKER_05 2:06

I got reconnected with Yao Kung some months ago, and he kind of like urged me and pushed me to like automate some of my stuff. S T U F F. So um, and I had a deep thought about it, and it got me really interested actually. So at the same time, I'm also doing this like post system um to link up my customer service like from the front end all the way to the back end, um, back of house, and then back to our um customers again. Yeah, so um everything is in progress now, and I'm seeing a good results. Um, we're trying to automate some of our replies, we're trying to gather some of the data that we have had for the past 11-ish years and putting it into this central system. Now, but at the same time, I all my staff are mixed up. S T A F F, yeah. Learning how to use this new system, um, really pushing themselves to connect the different dots together. So I'm thinking like how as a founder and as a steward of the company, I can help value add um for my people.

SPEAKER_00 3:24

Yeah, and I know Lynn, you mentioned that you you would eventually hope for a fully automated dashboard where you don't have to ask anyone, you don't have to press any buttons. Yes. Um, but I think just for the purpose of this one hour, uh, we also uh asked you to export some of your data for us. Uh we've cleaned it up already, so that's uh so no one's uh details are going to be leaked. Um but we're just going to briefly see what we can do in an hour. Our goal for this is really just to show everyone and hopefully more business owners as well that you might not need to build a full production app to make your life easier, but at least knowing the full gamut of like what's possible, I think we'll will greatly improve your working life in some sense. Yeah.

SPEAKER_05 4:08

Yes, I look forward to that.

SPEAKER_00 4:11

Yeah.

SPEAKER_08 4:11

Alright, alright, let's go.

SPEAKER_00 4:12

Alright, so I'm gonna just share my screen. I have this thing called warp. Warp is basically another form of terminal, but Yao Hong asked me to download this the other time. So I'm just going to uh so you can have taps, yeah.

SPEAKER_08 4:26

Yeah, so and and don't have to worry about finding your window. Leno is like, why is the screen so black?

SPEAKER_07 4:37

Don't read my mind.

SPEAKER_08 4:40

I'm sure many people are thinking that. You could change color lah, you could change it to white if you really want. Or pink, pink, pink also can yeah.

SPEAKER_02 4:46

Yeah, let's do it.

SPEAKER_08 4:48

Pink on. Yeah, you see. Wow. Okay, very nice. More appropriate for the brand today uh point.

SPEAKER_00 5:00

Okay.

SPEAKER_04 5:00

I mean see now I'm excited.

New feature in Claude: Auto

The problem: marketing spend across Google, Meta, Instagram - where's the ROI?

SPEAKER_00 5:02

Yes, all right. So put to the black. So Claude basically released this new thing called uh Claude Auto. Which is which, if you look at this little chart here, hopefully it's simple enough to understand. In the previous episode, we used this thing called bypass permissions, which is at the bottom right here, which is as you can see, very low on safety, um, but very high on task autonomy. So we built Brian and oil um analysis app in a very short amount of time, but that was actually not the safest thing to do, especially with company data. Right? So now with this thing called auto mode, as you can see, my mouse, uh, it basically goes up in terms of safety, yet still retaining some of the task autonomy uh that you can do. So before this session, uh Lynn mentioned I think one of the things that she wanted to do was to create a dashboard, right? Do you mind just explaining a bit more on this dashboard, especially for those who are just like listening in and might not see our screen?

SPEAKER_05 5:59

So we've been around for 11 years, and obviously we have um we've been very grateful for our journey. We have a lot of clients who are repeated clients who come to us regularly. At the same time, um there are clients who reach out to us through different platforms as well. So on a monthly basis, Dr. Bex actually spends some ad dollars on Google, on Meta, and also on IG itself. Moving forward, there are plans for us to um understand our data better and then spend money on ad dollars or advertising budget on different other different platforms. But until then, we want to very be very cautious about how we spend our money as well, right? We want to reach the right audience, we want to reach new audience. Yeah. So I do really want to understand is our marketing dollars really working for us now? For example, like if I were to spend X amount of money on um Google search, right? The the words, the keywords, which keyword actually um ultimately not only give me the leads, but also give me a confirmed um service closed. Yeah, a case one, for example. Um I would also like to um understand in one at one glance, like instead of pulling data from different sources, I hope to see on a dashboard which, for example, is it Google or Meta that's given giving me um better results? Yeah. Is it a real that's giving me better results as compared to uh word search? Yeah. So if I if you can if you guys can work this out for me, then you know I I don't need my staff to spend so many hours pulling in the data, like how we traditionally do it. And also I'm able to write it, I mean like write down the formulas in so that I know for sure, because like different platforms um you you calculate the ROI differently, right? So if we can already write this all in, then that would be perfect.

SPEAKER_00 8:25

And how much time does your staff spend on this these days?

SPEAKER_05 8:28

Um to be honest, it is very um it is not a regular thing that we do.

SPEAKER_08 8:37

Okay.

SPEAKER_05 8:37

Yeah.

SPEAKER_08 8:38

So how often it should actually be regular.

The pain of pulling data

SPEAKER_05 8:41

Correct, yeah. So I'm actually trying to build it, but as I said mentioned just now, right, to be really, really, really honest. Everyone is working so hard um at this point in time. Yeah, we're trying to build the ERP system. Um, we're trying to educate our our colleagues to adhere to certain like new standards that we are trying to build and learn and keep strictly to the different processes. This itself is already taking up a lot of time. Yeah, on top of all the daily tasks that um my managers are doing. Yeah, so I do hope to have a regular meeting. Yeah, so if they were to do this, then they would it would be quite painful for me as a boss, lah. Maybe they will have to take like two days, I would think, yeah, to churn data like that for me. Because it's not on a regular basis, right? So they have to download all the data and then put it into different matrix, yeah, and then come back to me. Yeah.

Feeding voice notes into Claude plan mode

SPEAKER_00 9:41

Yeah, for sure, for sure. And I think this is a classic case of how AI works. Oh, yeah, yes, how SME works, I guess. Yeah, very, very uh see that every day. Always fighting fires. Um yeah, and and also I guess how AI also would be able to uh free you up not just for the work you're doing now, but for the work that you would not have been able to do otherwise.

SPEAKER_06 10:12

Yeah.

Scoping decision: live dashboard or demo prototype?

SPEAKER_00 10:12

What I'm gonna do is I'm going to um just kind of copy whatever we just talked about here. Okay. There are applications out there that let you talk to your computer. So honestly, whatever you just said would have been processed quite well by an LLM. So what I'm gonna do is I'm going to switch to this thing called plan mode, pressing shift and tap, um, shift and tap to switch into plan mode, and then I'll just put it in. And so you can see it's quite messy, right? Like it like um not that your thoughts are oh yeah, my my thoughts are messy. Okay. Okay, I'm just going to add in that the the main task is to make a new dashboard for this, right? Based on what you told us, and then based on uh a bit of what you typed to us as well. But uh mostly based on what you told us, uh, which is yeah, what we talked about over WhatsApp before this call. Okay, so so we're just gonna go into a plan mode, and what are we building in this session? A live working dashboard connected to real data or a polished demo prototype. Um so I'm going to Okay, so um Okay, we don't have any APIs, which is basically how your data would connect in a live way, right? Without you having to export data. So I think we'll just do a working prototype with sample data. And then for data sources, uh we already have a clean uh data source. So I'm just gonna copy the file name and paste it in here.

SPEAKER_03 11:53

Okay.

SPEAKER_00 11:54

Yeah.

SPEAKER_08 11:55

But right, if we have access to the API, then it will be live. So you would have to check with your team tomorrow after you installing clock code and then you can uh access everything via API.

SPEAKER_05 12:09

Okay.

SPEAKER_08 12:09

Yeah.

SPEAKER_00 12:10

But I also think what you're doing.

SPEAKER_05 12:11

Basically they're not scared of me. Like every day I ask them a different thing.

SPEAKER_08 12:15

That's okay, that's normal. The boss is 11.

SPEAKER_06 12:18

Okay.

SPEAKER_05 12:19

They're quite receptive, eh? Cool, huh? Yeah, I think so.

SPEAKER_00 12:24

No, as in I think I think a very important trait of being a company that is you know future proof is one that is also thinking about how it can help its employees to grow. And I think this is yeah.

SPEAKER_05 12:38

And my team, they are really like, I think I'm really thankful for them, like, because they are they really look for problems. Like they really they really feed back to the developer and say, you know, there's a problem here, there's a problem there, like how can we do it better? And yeah, vice versa. So happy to see. Yeah, and I do want um definitely my brand to be future proof. Yes.

SPEAKER_00 13:05

Yeah, and and if you're there, there are some organizations out there that are that are fully using plot code. So these are very, very future-oriented organizations. Um and and even for the non-technical folks they are that are using it, they they find their productivity kind of skyrocketing, and they also find themselves growing because they are working in new ways, basically. Yes. So yeah, that's just something to think about.

SPEAKER_08 13:36

So the idea is that you are doing things, not just automating the old ways, but maybe you start questioning does it make sense to do even do it this way? Right? Is there something better? Yeah, and yeah, when you start trying to solve problems, that we actually can be quite interesting what people do.

SPEAKER_05 13:55

Yeah.

SPEAKER_00 13:56

So here's what the here's what Clock Code is proposing. Okay. Uh, there are 659 deals with source attribution, funnel stage, deal values, response times. Uh, it already has found that WhatsApp is 89% of one deals. Uh, Instagram and Facebook bring leads that barely convert, so that is a funnel. Um so you don't I mean I I wouldn't trust this immediately. So feel free to feel free to just jump in as well, Lynn. Um then there were four 404 deals that were stuck in engaged. So that's a big opportunity there, with response time varying wildly, which likely correlates with conversion. So, what it's gonna try to build is a single HTML dashboard file. How I see it, it's just something you can open and then it opens your your browser.

SPEAKER_08 14:46

I think it's it's quite cool because it already kind of realized that I mean we are doing a live session now, so it's not gonna over overbuild it for us. Yeah. But it's a good uh kind of like a proof of concept, right? That you can bring it to the team to idea upon it and without spending too much time.

SPEAKER_02 15:02

Yeah.

SPEAKER_00 15:02

Yeah. And ideally, we would have done a design phase before this. I would kind of draw it out as well, like how I want it to look like. But we'll also just trust the system and see what it comes up with based on its own.

SPEAKER_08 15:16

Let's scroll down and see what's the build plan.

SPEAKER_00 15:18

Yeah, so um this HTML dashboard file is going to answer six of your questions with charts and KPI cards. Uh so what it will do is parse the Excel to JSON and build a section by section, which okay, we don't have to go through that. Um so this is a plan that uh we would typically read um as well. So you run Dr. Bags, uh, but has no clean way to see which channel brings revenue versus enquiries. Um managers need two days to do this, uh, or even more, right? You have a sanitized Excel report. Yes, the goal is to build a visual marketing dashboard within one hour, and that you can glance at and understand where the money where your money is working. Okay. Um and then there's some of these key stories here, which we talked about. They are telling you the kind of um software packages or software that they'll use. Um so step one, they will they will read.

SPEAKER_05 16:15

Is that what you do all the time, Yahoo?

SPEAKER_08 16:17

Yes.

SPEAKER_05 16:18

Wow.

SPEAKER_08 16:19

In fact, this step is very important because this step you you need to kind of read the plan and figure out whether it's going the right direction and then you start changing.

SPEAKER_06 16:29

Yes, okay.

SPEAKER_08 16:30

Because a lot of a lot of times I see um folks, they actually just trade to solutioning, right? And you end up spending a lot of time trying to course correct along the way. But it's actually easier when you do a proper plan, you start from there and then and then you start making changes. Like, for example, some of these things, if you don't like it, you should just highlight. For example, now let's say we go down and say that okay, we want to step two, build the dashboard HTML, row one, you have KPI cards, right? So this part here, you you start thinking whether it makes sense to have total leads, right? Uh one deals and win rate percentage, total revenue, right, average response time.

SPEAKER_00 17:13

And please jump in anytime, Lynn, uh, as you're reading as well.

SPEAKER_08 17:16

Like if anything jumps out, just just say, oh, actually, I don't really want to see it this way, I want to see it the other way, right? And then you yeah.

SPEAKER_05 17:24

Now understand why why my developer always asks us to give constant feedback whenever like something is launched. Yeah, so yeah, like how to better communicate with him would be actually to draw a plan or to to explain to him. So as we build our entire ERP, nobody in the world, I I don't think there's a restoration company in the world that has an ERP system as tight as ours. Because um it is it is a traditional business in 2026, right? So for a restoration business to run in Singapore, fully in Singapore, um, and not like what our competitors do to run overseas, we need to really watch the KPIs. Yeah. So like that is the business, that is a big part of what I do on a daily basis, yeah, to make sure that the team runs effectively and efficiently. Yeah. So all these little things that I'm seeing here would be connected to the back of the house. Yeah. In the end. So yeah, I mean, like, thank you for doing this. I will continue reading, and if I have continue to tell you.

unknown 18:39

Yeah.

SPEAKER_08 18:39

In in real life, uh, what would happen is that this is just the first layer.

SPEAKER_06 18:43

Yeah.

SPEAKER_08 18:44

You can then add the data from other sources uh from ERP, and then you make it even more robust. Yeah, yeah. Like obviously, that one you have to sit down and think quite deeply now how you how that can be done. Uh, what needs to go in?

SPEAKER_06 18:58

Yeah.

Plan walkthrough: 6 rows of the dashboard explained

SPEAKER_00 18:59

Alright, so row one, KPI cards, row two, channel performance. Uh, row three would be the funnel, which is the funnel visualization, from a new inquiry to being engaged and being won or lost. Row four is the revenue and timing, which is your revenue by channel, and then uh timing-wise would be the deals by the day of the week to find best closing days. Interesting. Uh row five would be response time impact. So uh scatter or grouped bar, response time versus conversion outcome. Row six, campaign performance. Uh so there'll be a table with your campaign name, leads, wins, conversion rates, revenue, and answers which ads or keywords work. Okay, then there'll be step three, which is the polish and brand, which will be the brand colours. Do you have any design materials? Uh website website thing.

SPEAKER_04 19:53

We do, we have a website.

SPEAKER_00 19:54

Okay, drbs.com. Perfect. Um, okay. I'm going to ask Claude to oh, okay, never mind. Okay, I'll just take a screenshot first for later. Just take a screenshot and put it in a pick up as you pick out the colors. Yeah, but it's I'll just let it go first because now I can't add a photo. Okay, so you'll use the brand colors and all mobile friendly. Oh, we don't really need that, but sure. Okay, why not? Um so the files is gonna create some files, it's gonna refer to some files, which is the data files. Um there's a verification where you run a Python script uh on my computer, and then it will open this thing in the browser and you check the dashboard and verify that no customer names are leaked, uh, which we have done already. And then you can test on phone. Okay, no, I don't know if you can test on your phone, but sure. Um let's go.

SPEAKER_05 20:58

Okay, so we're just going to Can we also give like advice on what we should do?

The build starts

SPEAKER_08 21:04

Yes, what you don't don't don't do that now. Just build it first. Okay, sure. Remember that thought, remember, hold that thought, yeah. Clear context and go. Yeah, because what happens is that the context is actually uh uh very critical in this stage. Yeah, you don't want to give it too much context until it goes uh it gets uh bit confused.

SPEAKER_00 21:27

Okay, additional thoughts, um advice and actionable uh insights insights on what to do based on the data. Um just wait for it to do it. I'll just approve it's yeah, you didn't you didn't do the uh accept auto. Actually, this auto mode thing is actually not very auto. Oh, what is it? I haven't tried yet. Yeah. It is uh getting me to approve a lot of things. I think one thing to note is that uh Clockco is not just For coding, right? You can use it to like move files around, to rename files, to open files, even to do research for you.

SPEAKER_08 22:11

Market research, uh, we did that last episode, you know, look at what competitors are doing, grab prices off. If you think about marketing, you could actually look at what other people are selling or talking about. Yeah. Or the number of uh like uh if you monitor their polls and see how's the uh results from what they are doing, you know. So actually quite quite a lot of things we could do.

SPEAKER_05 22:37

Yeah, yeah, it's so fun.

SPEAKER_00 22:39

Yes, it is quite fun. Um maybe as as we usually do in each episode, we have uh another task that we just try to do, try to do alongside this. Is that anything that pops into your mind?

SPEAKER_05 22:55

Yeah, yeah, yeah. I want to do the before-after picture thing that you suggested.

SPEAKER_00 22:59

Okay, okay, before after picture.

SPEAKER_05 23:01

Because we actually take before all the things.

SPEAKER_00 23:03

Okay, then so where can I get this fast?

SPEAKER_05 23:07

Um, Instagram.

SPEAKER_00 23:08

Okay, I'll go to Dr. Bags to get your style. Yeah, and then I do need some before and after photos. Is it here as well?

SPEAKER_05 23:15

Um yeah, one of the there are some posts that has okay.

SPEAKER_00 23:18

Which one is should I open?

SPEAKER_05 23:21

At the bottom, uh I mean you scroll down.

SPEAKER_00 23:23

Yeah.

SPEAKER_05 23:23

Uh that shoes the shoes before after.

SPEAKER_00 23:26

Yeah. Okay. You change the colour.

SPEAKER_05 23:30

Yeah.

SPEAKER_00 23:31

Alright. Well, from yellow to pink. Very cute. Okay.

SPEAKER_05 23:35

It was it was pink.

SPEAKER_00 23:37

Oh, sorry.

SPEAKER_05 23:38

It became yellow.

SPEAKER_08 23:39

Huh? No, before it's yeah. Oh, but your original colour was pink.

SPEAKER_05 23:45

Yeah, I I think the original colour was lightly pink.

SPEAKER_00 23:48

Wait, doesn't it write before as yellow and after is pink?

SPEAKER_05 23:52

I don't know if you're just it became yellow.

SPEAKER_00 23:54

It is colored and it became yellow. It became yellow.

SPEAKER_05 23:57

Yeah, you notice the I get it.

SPEAKER_08 24:03

Hue is quite dirty on the left side. Okay, you can see how how much I know about skin eyes.

SPEAKER_00 24:10

Yeah. Yeah. OCD people is like that. Okay, I'm gonna open a new one. Clot permission auto. Um and I'm going to say, I'm just gonna prompt hi Claude. Uh, we want to build something that takes uh before and afterback photos to write social media posts. Um I have a few photos that I'm going to kind of uh download and drag into a folder. Can you just open that folder for me as I drag some photos in? And then I'm also gonna give you some examples of uh previous writings uh from their Instagram page.

SPEAKER_05 24:48

It's like talking to your best friend.

SPEAKER_00 24:50

Yeah, yeah. It's just like talking to someone like it's it's not. Your soulmate. Yeah, yes.

SPEAKER_08 24:55

Um I'm not gonna go there, but uh I would think that in a couple of months a lot of us will have to go and see psychiatrists.

SPEAKER_00 25:04

Yeah, um, let's not get too dependent on this, but yeah, I will paste in what I just said um and then I'll let it open a folder for me. Okay, I'm going to just go back and approve more stuff. Oh. Okay, that's what always happens. We always get distracted because clawed code is very fast.

SPEAKER_08 25:21

Oh, I seriously think my attention span is much shorter these days.

Dashboard revealed - 659 leads, 219 deals, 33% win rate

SPEAKER_00 25:25

So sorry, Lynn. Uh we are also breaking your attention span, but we're going to go back uh towards the marketing dashboard, and we're gonna get you to comment on it. So for those that don't see the screen, this is a marketing dashboard for Dr. Bax. It shows that we have 659 leads, one 219 deals, which is a 33% win rate, and there's an average response time of 1 hour 44 minutes.

SPEAKER_05 25:52

That one you can keep.

SPEAKER_00 25:53

Yeah, leads by channel.

SPEAKER_08 25:55

That's very slow there. 1 hour 44 response time. Or is it uh two ways though?

SPEAKER_00 26:00

Is it two ways or okay first reply time? Okay, there's leads by channel. WhatsApp is the most uh at 405, uh 166. Lead source. What is lead source? Anyone knows?

unknown 26:17

I don't know.

SPEAKER_00 26:18

Okay, there's a source which says lead source. That's why. So let's leave it there. Uh Instagram at 58, Facebook at 30. Uh I'm just gonna uh oh sorry, on the right side.

SPEAKER_05 26:29

I think what's really good about today's exercise is that I'm gonna question all the numbers, right? Yeah, what's the lead source? What's this, what's that?

SPEAKER_00 26:38

Yeah, you'll try to understand your data a bit better. So you went into it.

SPEAKER_08 26:42

Actually, yeah, that's the underpin of the whole thing of automation, right? Your data, data is the most important thing.

SPEAKER_05 26:50

Yeah.

SPEAKER_08 26:51

If if your data is not right, then your output or your processes won't be right.

SPEAKER_06 26:55

Correct.

SPEAKER_00 26:56

And Yao Hong, here's where I have a question for you, also, right? Because you know, this is just one month of data, right? So even for one month of data, we are like questioning whether it's true. We are we are getting confused. Like, how would you recommend a company do this for like, I don't know, like a year's worth of data, right? It will get so confusing. And like, how do you like verify?

SPEAKER_08 27:19

I think we try to get it right the first time around. Yeah. Your initial one, you you must make sure that the data is clean, plug all the holes. I mean, it's it's literally data engineering, right? Yeah. We actually go down, it's plumbing, right? You have to clean the data sources, you've got to clean any transformations you have to get to what you want.

SPEAKER_00 27:40

No, which I find quite difficult as someone that's non-technical, right? And trying to even judge whether the data is clean and judge whether data is correct and and know what to look out for.

SPEAKER_08 27:50

But sometimes uh for businesses, because you you know the numbers intuitively, you can kind of catch quite fast whether it's something is wrong. Yeah.

SPEAKER_05 27:59

I mean, this transition period is important, right? Because that's what the government wants us to do as well. Correct. So so managers need to understand that this data needs to be really clean. And we need to understand that if we want AI to be helping us, we need to put in our effort and our time to correct to build the system first. Yeah. You know, I I think as business owners, um, it's very important that we are the engineer of this entire thing that we're looking at now.

SPEAKER_08 28:30

In fact, this is the part where most people say, Oh, AI is not good, then therefore I'm not gonna do anymore.

SPEAKER_05 28:37

Yeah, that's that's the part that most people but it's a it's a fallacy, right?

SPEAKER_08 28:41

Because yeah. Yeah.

SPEAKER_05 28:45

The difficult part is always I think now. Yeah, but if we were to really press on, um have good friends like Yao Hong and friends like Eric, then we would all win today.

SPEAKER_00 28:57

No, or maybe you could just try this with yourself, I mean, on your own.

SPEAKER_05 29:00

Yeah, but that would take take us 10 years. Then the world would be different.

WhatsApp attribution problem

SPEAKER_00 29:06

That's true. Yeah, that is true. Very good point there. Yao Hong, we still have a different thing.

SPEAKER_08 29:10

Actually, I quite uh quite a finding interesting that it says Facebook is only has 13.8% conversion rate. That's quite low.

SPEAKER_05 29:19

Yeah, so it's it's pretty interesting because most of our most of our call to action would bring us to WhatsApp.

SPEAKER_08 29:27

Ah okay, so actually the attribution is not correct already. That means in terms of the way you design the flow, you must attribute WhatsApp, must have attribution back to what uh Facebook. So I mean you can't tag it to WhatsApp, yeah.

SPEAKER_05 29:41

Yeah, so that's why I think this is weird. And also, I don't know, like because of the way we design our ERP system, we need to start. Remember, just like at the front, um, when Eric asked about the entire flow, right? My sales team have to start an order. And this order that we start might not have come from WhatsApp, like this specific WhatsApp number, it might have come from other sources, like other WhatsApp numbers that we have. But ultimately, maybe the system only captured like it came from resource.

SPEAKER_08 30:22

The system needs to have a bit more granularity that you may need to go back and redesign a bit. Yeah, good. I see we are we're uncovering some potential issues, yeah. Yeah.

SPEAKER_00 30:32

No, I I think it's very important that that Lynn you are like open to even thinking about all these changes that you have to make. But um let me just carry on first. Okay, because we we can talk about this forever. Uh so moving on to deal funnel, all leads is um enhancing okay, and 61% of deals engaged, 33% won, two percent lost, uh deal by channel, which uh revenue by channel, which is a bit off as well, yeah. Um, due to the attribution to WhatsApp, uh even though some came from Facebook. And then, of course, if you look at the deals, it starts very high at month in on Monday, it goes down uh all the way to Friday, and then it starts going up again on Saturday and Sunday.

SPEAKER_05 31:17

Um deals are the one in blue. Or leads are the one in blue, yeah.

SPEAKER_08 31:22

Yeah. So you see that actually quite interesting. Right? Sunday has uh quite like quite a lot of leads on Sunday, but the conversions from Sunday leads. I'm not sure. Does it trace the conversion?

SPEAKER_05 31:34

I mean, this looks normal to me because on Saturdays and Sundays, most Singaporeans will be at home, so they'll be texting us. Now, but my my office will be closed, right? So the sales team or or whoever is uh knows how to reply and has been in like my team is very robust, so we will be replying. But on Mondays, from Mondays to Fridays, we would be the customer service team would be back at work. So they'll close the deals, yeah. That's when like the bags come in as well. So because we do things fast by Friday, yeah, yeah, it will be cleared. Yes, and then we'll be doing other things like following up of orders, yeah, inquiries and things like that.

SPEAKER_00 32:24

Okay, cool, makes sense. Yeah. Um, and then next there's a response time versus conversion. I'll zoom out a little bit. So it shows that um for zero to five minutes. That well, how do you even read this? Zero to five minutes.

SPEAKER_05 32:43

So you need to make it pink, Eric. I'm sorry. No, no, you don't have to.

SPEAKER_00 32:49

Okay.

SPEAKER_03 32:49

You don't have to, I was just thinking.

SPEAKER_00 32:51

Okay, okay. Yeah, I still don't I'm still trying to. Can you can you explain this, Lin? I mean, you are the boss here.

SPEAKER_03 32:57

Like I got Lao Hua. You need to make it bigger.

SPEAKER_00 33:02

Two of us can Okay, one hour bucket shows 40% conversion, higher than faster buckets.

SPEAKER_03 33:10

So I still can't see young men.

SPEAKER_00 33:13

Okay. I'm going in.

SPEAKER_08 33:15

Okay. Zooming in, zooming in. I guess the intent is higher when people talk longer on WhatsApp review.

SPEAKER_05 33:24

Yeah.

SPEAKER_08 33:25

So that's what the metric is trying to say. Yeah.

SPEAKER_00 33:27

But their insight is quite wrong because they're like, you don't need fast replies. But actually it's not.

SPEAKER_05 33:32

The one pass out 38%. Oh, I understand. Higher than faster baskets.

SPEAKER_08 33:38

It's the kind of customer that asks you a lot of questions, but then they will still give you the business.

SPEAKER_05 33:42

Correct. And they're saying it only makes out 38.8% of the entire conversion. Is that correct? Yeah, but in the end, they will still give you business. And that's that's the customers that we want to talk to because we believe that educating them about services is extremely important.

Deals stuck in "Engaged" - a big untapped opportunity

SPEAKER_00 34:00

Yeah. Okay, okay. 404 engaged problem.

SPEAKER_05 34:05

Before I went to Hong Kong. So basically, we were looking at this entire no, actually, we've been looking at it for like two months now. Like this, we realize also, although we don't have data, but like what Yao Hong said, we are in the business every day, so we look at where we can improve on. Now that um we are trying to build systems, right? We realize that there is a large like customers do not only say yes um to us on the day they inquire with us. Yeah, most of us aren't so decisive, right? Especially when we mention that these bags are important items to them. So it might take more than 24 hours, it might take more than 48 hours, and therefore, like at Dr. Bags, we are trying to program our communications with clients um in a more sensible manner. Yeah, like maybe we re-engage with them, we re-talk, we talk to them more. Yeah, so so this is this is why we also realize that there is potential in this engaged column. Yeah, and I feel like through re-engagement, through more engagement, we will be able to um have customers understand us better and trust us with their items more.

SPEAKER_07 35:25

Yeah, yeah.

SPEAKER_05 35:25

And even if we don't they don't say yes to us today, they might remember us for the quality of our replies, yeah, at a later stage.

SPEAKER_08 35:36

Yeah. But what I think I will do here just to throw it out there, right, is to look at uh past conversations, measure the time between each messages, right? So we have that kind of granularity, and see what's is there a pattern between uh the last message and when they get stuck in engaged, right? And then to warn. And then from there we can try to then try to uh optimize for the time where there's an automation back to the customer. Hey, are you still thinking about this?

unknown 36:11

Right?

SPEAKER_08 36:12

Yeah. Because I'm not sure if your team do follow up when people are in engaged or like how many times do you message back? Yeah, that kind of thing. So there's a lot of strategies you can play. So from there you can do like testing, like, oh, we send one message at this time, and you can test. It's like basically an experiment, right? You you send uh after one day, you send a message, or then no reply another day. Maybe after three times you say, okay, drop. So it doesn't perpetually stay engaged, lah, right?

SPEAKER_05 36:40

Yeah, that that's what I'm doing.

SPEAKER_08 36:42

Okay.

SPEAKER_05 36:42

Yeah, but you your your your your suggestion sounds even more complex, which I like. So I wrote it down.

SPEAKER_00 36:55

Because you have WhatsApp uh business as well, right? Um there are always ways to kind of build your WhatsApp bots, uh, depending on the complexity, and then you know, maybe just having a nudge, like, hey, are you still interested? Correct. Might even increase, uh, might decrease some of these stock deals as well. Um but you could start. Imagine 10% of this converter is 40 more wins, huh?

SPEAKER_08 37:17

Yeah, yeah.

SPEAKER_00 37:17

So that's yeah, but you could start small. I mean, you don't have to go like full, fully like uh WhatsApp agent that can do anything, right? But I guess this would be one small problem that if you were to work on, you know, just a kind of notification message, you could actually add some deals as well. Um I would caveat against jumping straight to using an agentic AI solution here because you don't understand the problem yet, right?

SPEAKER_08 37:45

You try AI there, it's gonna get confused. You've got more confused than humans.

SPEAKER_05 37:49

Yeah.

SPEAKER_08 37:50

Yeah. Yeah. Yeah.

unknown 37:53

Yeah.

SPEAKER_05 37:53

So what I'm doing now is I'm putting them into different baskets. Yeah, at this current moment, lah, I'm trying to put them into different baskets.

Sidetrack: Instagram copywriting task

SPEAKER_00 38:02

Yeah. Which is good. I mean, it's a good step to classify and and understand your business better. So, yeah. Um, I understand we are almost hitting an hour uh and we still have that one Instagram task. Let's finish this one.

SPEAKER_08 38:19

I want to see this one.

SPEAKER_00 38:21

Let me just let's just, you know, let's just wing it. Um I'm just gonna paste in the photo and say, hey clock code, I have no time. Can you please make uh you know a very good Instagram post uh that uh basically is very short, very readable, um hopefully just you know, uh two or three short lines and then with pretty good um uh hashtags as well. Uh these are the photos for you to base it off on. Uh Lynn, is there any other things you add to the how you write your social media posts?

SPEAKER_08 39:03

This shows how Eric doesn't not a very design sensitive person. We should be talking about the the mood and feel of the images for the the brand brand guide, the brand style. Yeah. Yeah.

SPEAKER_00 39:18

The new life, making it feel refreshing.

SPEAKER_06 39:21

Yeah, yeah.

SPEAKER_00 39:23

No examples, I'm in a rush. The episode is ending. Okay.

SPEAKER_03 39:30

He's your best friend.

SPEAKER_08 39:32

This is gun tripping the AI, you know.

SPEAKER_03 39:34

Yeah.

SPEAKER_00 39:37

Okay, this is bad. Yellow in, yellow out, thank you. Oh, wait, is there a uh picture that I gave it a picture just now? Which picture did you give it? This one. I'm just gonna give it okay. Okay, here's one example. Yeah, okay. Um I'm going to give you one last picture to try on.

SPEAKER_08 40:00

Um, so this shows the the limit of uh creativity in in the LOMs. Yeah. I struggle with this a lot, right? To ask them to write something that's creative. Yeah. It sometimes gives you very like kind of yeah.

unknown 40:16

Okay.

SPEAKER_05 40:17

But likely it's because it's the first time that we're talking talking about.

SPEAKER_08 40:21

Correct, correct. Yeah. So usually how you make this better is you put in a build a relationship. No, no, I mean kind of. You actually send it a lot of examples in the past that you have so that it it learns uh your your brand, your your brand uh your tone of voice, yeah, and things like that. Identity, correct. So it you'll try to write the same style.

SPEAKER_06 40:43

Okay.

"This Adidas is being written in a Ferragamo copy"

SPEAKER_00 40:45

Yeah, okay. This is weird as well. Now the Adidas is being written in a Ferragamo copy. So from dull and dingy to fresh and bright with a star emoji. Wow. These Adidas Continental 80s were brought back. Okay. Yeah. Lynn, any thoughts on this?

SPEAKER_08 41:05

Very bad. You see, the no one can say no.

SPEAKER_03 41:12

No, no, I think maybe maybe this is how Claude Code knows of Eric. I don't know, I don't know.

SPEAKER_00 41:21

Uh no, no. This is not from memory. It's a fresh uh it's a fresh instance.

SPEAKER_08 41:25

Yeah, a fresh one. So uh so we should actually feed it more slowly.

SPEAKER_05 41:30

But I understand where you're where you're heading towards. Yeah, yeah, yeah. And yeah, I think it's important. I think if we feed it the correct information, um, yeah, it might likely give us better results.

SPEAKER_00 41:45

So I I think not just information, but um, after you know, feeding in a few or maybe like 50 captions, um, I would also ask it to kind of extract out the principles uh behind some of the marketing copy so that it's not just copying word for word or or line by line, right? It's like actually trying to understand what you're thinking. And then of course your your marketing team can go in and say, no, that's not what I was thinking of. Uh here's the principle instead. And and of course, so it's a mixture of like principle and examples so that um you're not just Does it work for videos as well? Uh in in a cap you mean in text for videos, is it? Or or like or you mean you upload a video and then you write a caption?

SPEAKER_02 42:30

Yeah.

SPEAKER_00 42:31

Um no, I I think for now, quick answer, I don't think the videos would be as good as as yours uh for now. Um definitely yeah, I mean TikTok and and reels are getting there. But um yeah, I think you still need quite a bit of human intervention for videos, um, just because yeah, there are a lot of limits for for videos, even though there are uh some software packages out there that can edit videos, but it's super basic. It's mostly just cut and put a picture on and maybe trim the video. Very simple stuff. So we are not there for video yet. So yeah.

SPEAKER_05 43:11

Okay.

SPEAKER_00 43:11

Okay, I'm just gonna stop sharing my screen so that uh you can get back to your your work and rest. Um but I just wanted to uh rest. I need to go back to work. Yeah, yeah, yeah. So but yeah, maybe just any any quick reflections, Lynn, on on the hour that we had today. Anything that you would uh start doing differently from today?

SPEAKER_05 43:40

Definitely to understand the the data that we were using just now. And to yeah, like what are the different um to clean that data up to so that we can really utilize it to do what I want to do. I already know that um the the possibilities that I thought could be materialized can actually be materialized, but my data needs to be correct first.

SPEAKER_00 44:11

Yeah. Yeah. And actually, maybe just to caveat, the the things that we did today could have actually been done in uh clot.ai session. Yeah, what do you have? I mean, I'm shooting myself in the foot.

SPEAKER_08 44:28

Right. You didn't really need clot cooler. I think this was you can have used the desktop app.

SPEAKER_07 44:34

Yeah.

SPEAKER_00 44:35

Actually, this would have worked by you just uploading your data into uh especially since it's just one month of data.

SPEAKER_05 44:41

Yeah.

SPEAKER_00 44:42

Um if you just uploaded it. Yeah.

SPEAKER_05 44:44

But in um but when I finally get to do this, it wouldn't be a one-month thing, right? So it would be uh you need to look at trends and everything. Yeah, correct. So I'll pull out all my data and understand it. Yeah. And I think that would be uh yeah. How long do you think it would take?

SPEAKER_08 45:03

Not very long, I would think.

SPEAKER_00 45:06

You mean just to understand? You mean just to understand your data or how how long?

SPEAKER_05 45:10

No, to really do up with the dashboard. Yeah, to build it. Given my intellect.

SPEAKER_04 45:17

Sorry, sorry, sorry.

SPEAKER_03 45:20

I like pink. I wish your PPC now. Not very long ago. Yeah. I cannot.

SPEAKER_00 45:34

But there's still a process of like a normal human being. No, I mean I would as a normal person, like for me, I would actually ask like what are the steps for data, like proper data analysis to make it clear, right?

SPEAKER_08 45:48

Uh I am to train you or tell you what and check with somebody, maybe. Yeah. And it's good to chat with your tech team. Basically, you can send it to me.

SPEAKER_01 45:58

Really?

SPEAKER_08 45:59

Help you, help you vet, help you vet, yeah. Well, I mean it's interesting. To me, it's always interesting problems to solve. Yeah. Yeah, yeah.

SPEAKER_05 46:07

So Nightwise for me.

SPEAKER_08 46:10

It's it's problem solving. Yeah.

SPEAKER_00 46:11

Yeah, yeah, yeah. And then I mean it would it will give you some steps, like you have to clean the data first and all that, right? So I mean it really depends how how comprehensive you want it to be.

SPEAKER_08 46:22

Um it can be very complex, yeah, if you want to do it properly. Yeah, because you actually need uh proper data warehousing, your data must pipe there from all the sources, clean up.

SPEAKER_05 46:34

Yeah, but I'd rather be complex and then later on. It's always like that, right?

SPEAKER_08 46:38

Like, correct, correct.

SPEAKER_05 46:40

Then the simplification comes after the complexity.

unknown 46:44

Yeah.

SPEAKER_08 46:45

Also, what we are doing now is a very quick um POT just to look at the numbers, and straight away you see a bit of a discrepancies already.

SPEAKER_05 46:52

Yeah.

SPEAKER_08 46:53

Right. And from there, it's it's it's helpful actually. Nevertheless, it's too helpful.

SPEAKER_05 46:57

Yeah.

SPEAKER_08 46:58

You you otherwise wouldn't have been able to spot this on your own. Yeah. Or you spend quite a lot of time making it.

SPEAKER_05 47:05

Definitely, like imagine I make use of all the data and then I spend two hours keying all the things and sorting it out, and then realizing bam, it's wrong.

SPEAKER_08 47:13

Yeah.

SPEAKER_00 47:15

Now we just talk about just talking.

SPEAKER_05 47:17

Yeah.

SPEAKER_00 47:18

Yeah. Which there is a lot of trial and error. There's a lot of trial and error for this kind of thing. So if you do have fires to fight, then I do imagine that you do need to measure that trade-off, right? Of like how much time and complexity you want to spend on this. Um But actually, actually, I do take back a bit of my words. Like if you were to do this in a normal clock chat, I don't think you will get so far, but you you might get uh a dashboard, but you might not get as accurate results and uh comprehensive kind of data analysis, especially within like two or three prompts that we just did. Um I think that's just a caveat. Uh you might have to prompt a lot more times on a normal chat interface.

SPEAKER_08 47:57

Yeah, because uh what Cloco did was that it created a program in the back, right?

unknown 48:01

Yeah.

SPEAKER_08 48:02

To do summation or aggregation of your data.

SPEAKER_05 48:04

Yeah. Yeah.

SPEAKER_08 48:08

So yeah.

SPEAKER_05 48:10

And I also learned that we need to communicate, continue to communicate to Claude.

SPEAKER_00 48:18

Yeah, and I I think communicating with an AI would teach you how to communicate with tech people. Like that's what I learned as well as a non-technical person. Because a lot of times I have ideas in my head.

SPEAKER_05 48:29

Oh, you sound extremely technical to me.

SPEAKER_00 48:31

Oh yeah.

SPEAKER_05 48:32

Um what's me? What am I?

SPEAKER_00 48:37

Normal people. Yeah, we're all just normal. Actually, I mean, yeah, I don't know what this what it means to be technical if you don't have to code with code anymore. But but yeah, it's good for uh training our communication skills and just being very clear about that. Yeah, I think that's true. Okay, alright then. Thanks so much.

SPEAKER_05 48:58

Thank you so much, boys. Thank you.

SPEAKER_00 49:00

Thanks for coming on. Have a great night. All right, thank you so much.

SPEAKER_05 49:05

Thank you.

SPEAKER_04 49:06

Yeah, bye. Okay, bye.

SPEAKER_00 49:10

Thanks so much for joining us for this episode. If you found this helpful, feel free to leave us a rating on your favorite podcasting apps or like, share, subscribe on YouTube. And if you want to get into the next episode, feel free to fill up the form as well. So, yeah, thanks so much for joining us. Uh, really appreciate that you even made it to the end. And see you on our next episode. Bye.