---
title: 'Research Interview: Most Time-Consuming Tasks — Oriol'
source: notion
interviewee: Oriol Altisent
date: '2025-12-17'
type: raw-transcript
status: active
tags: [research, interviews, raw-transcript, oriol]
---

> **Note:** This is a raw interview transcript. It has not yet been processed into JTBD opportunity cards.

---

# Understanding your most time consuming tasks - Oriol


**Transcript**

17 December 2025, 03:03pm

**Desmond Leo** started transcription

---

**Desmond Leo** | 0:03

I'm. I think to start normally I would, so I've got a series of questions that I kind of want to start to basically ask, but before we begin. I'm not exactly sure what your role is here, and in general, what do you what do you do?

**Oriol Altisent** | 0:28

But there you go.

**Desmond Leo** | 0:29

Yeah. Are you a researcher?

**Oriol Altisent** | 0:33

I'm technically, I'm. Marketing scientist and and what basically do is all the advanced analytics, but that's in in theory regarding the studies, in practicality, I I do a lot of of different things that shouldn't be. My my role here, but you you theory is more abundant.

Analytics means running segmentation driver analysis the category points or Max devs. All these all these kind of analysis in in a project so helping. The teams to set them up and run the analytics.

**Desmond Leo** | 1:21

OK. When it comes to. Quantitative projects, where are you involved the most that you're you're involved in the analysis side or are you involved earlier on on the briefing side as well?

**Oriol Altisent** | 1:38

Everywhere in the briefing side to to decide which kind of of technique we need to apply in in the study. For example, what can we offer to the client when when we have the the briefing, they're in the setup of the questionnaire, but everything is that included if.

**Desmond Leo** | 1:49

OK.

**Oriol Altisent** | 1:56

Including correctly also then the the analysis part I get involved to to run analysis and then sometimes in the in the deliverables to to the client.

**Desmond Leo** | 1:58

OK. OK. OK. Then I think these questions might be relevant or we might be able to to continue with the same line of questioning. Can you, can you think about a recent quantitative project? And walk me through how it actually played out, starting from when you were first involved, maybe at the briefing stage all the way to the final delivery.

I'm I'm particularly curious about what felt straightforward. And what felt like it took a bit more effort than it should have.

**Oriol Altisent** | 2:48

Yeah, but before we start 'cause, now we're doing this kind of presentation. Can you also define what is empathy IQ for the role and what we're doing?

**Desmond Leo** | 2:51

Mm-hmm. OK.

So empathy IQ is the temporary name for the the the software platform that we have and the software platform that we have right now is designed around. It's like a survey platform. It also does focus groups and or or online discussions and stuff like that.

In terms of what my what I'm trying to understand is rather than build just another platform and just another like technology tool out there. Where there's a lot of point solutions, there's there's solutions out there that like automate different quantitative research methodologies. There's point solutions that do specialised in focus groups and and stuff.

And what I am specifically exploring is. How we can help researchers across the entire research process to make it easier and better for them to do.

And so I'm looking for those long, those points of of of that take a lot of time. We could potentially automate through workflows automate with AI or build a software system that makes that specific thing a little bit easier to accomplish.

And I'm talking to as many researchers as we can, at least internally at first, to get an understanding of what problems we have as a team.

And hopefully I can speak to external researchers as well to understand what the market is. Is is saying like these are kind of like the biggest sticking points or the biggest problems and and and so forth that we have because the last thing I want to do is it's like if we there is this plan to. Launch or go go into public beta. With our platform and I need to make sure that it's got enough delighters in there that actually make it stand out from the the rest of the other solutions that are just other survey tools, right? I don't want to launch another survey tool, so I'm trying to.

I'm trying to figure out where what can we, what problems can we solve for researchers? That actually make a big enough impact where it becomes a must have technology for people to use.

**Oriol Altisent** | 5:25

And the survey platform is to use internally or is also for commercialised to other client and they can build their own surveys and do their own research.

**Desmond Leo** | 5:30

So currently.

Vian. I mean, I think the dream is to get it to public so that other people can can use it. Buzzgrave currently access currently have a tenant or a version of the the software already. I have a friend in the UK who runs UK Omnis and he's using it as well.

And so it is. It is. It is. The intent is to try and build a solution that solves problems for market researchers effectively built by market researchers as well. Market researchers that understand like core problems in the market research process. Because then and then we build software to solve those problems.

So.

We are trying to get. We are trying to get the software in a position to have all of our internal researchers use it instead and kick out decipher. And we will be leveraging and using our internal researchers basically as as as feedback points of feedback content and what we need to improve in order to just skip full adoption internally and then we'll go external.

Does that is that? Is that answer? Yeah. Do you have do you have? Do you have other questions?

**Oriol Altisent** | 6:47

Yeah. No, no.

But I, I I knew the part about the survey platform that we are creating, but then I also see all these messages that you're sending lately about dischargeability problems that are trying to help us with solution. I wasn't sure if there is also. This part that is coming more, yeah.

**Desmond Leo** | 7:09

So the the, the the chat, so the the the prompt improvement thing actually spawned from just like an internal.

I noticed that some prompts that people were using were particularly inefficient and and weren't really good at at like fine at. Getting the answers that that people would need and so then that's I was just like very quickly. I was like I'm going to put together a little thing to to help improve how people write prompts and so that's that's why I created that.

It's it's not necessarily part of of empathy IQ at all.

The questionnaire reviewer is a proof of concept that one's a little bit more complex, involving quite a few different documents and work flows and standardizations and stuff like that.

Eventually that will be. I'm hoping that will be something that we bake into the platform as well.

So someone then says, oh, you know what, I'm I'm actually just at the research design phase and I need to get a starting point. They can talk to the questionnaire reviewer in the platform. It designs a questionnaire. All of the the back and forth and the the edits happen inside there and then and then it automatically transfers into a programme survey that they can. Then launch.

That's kind of there. There's ambitions for that there, but I need to see how much interest there is in the market for for that type of solution in the first place.

When we break down the the research process, though, we go briefing design, field work reporting and so the the. I have an idea that. There will be a way that we can help across every step of that process and so for the design process, that's where the questionnaire GPT comes in play, where actually maybe we've just got a quick way that we can already help people come up with very quick tablets. For questionnaire designs that they can then that they can then use.

**Oriol Altisent** | 9:12

Mm-hmm.

**Desmond Leo** | 9:13

It also it's also it also helps resolve some of the sometimes reformatting a questionnaire really sucks and is boring and is time consuming and you don't want, you know, no one wants to be spending their time doing that.

So like if you have so with with that specific. Like GPT. Like if you have a questionnaire already designed and you just need to make it better formatted for programming, you throw it in. It formats everything for you, but if you also just have a brief and you don't understand what questions you should ask in the 1st place or you don't want to spend the time coming up with those basic questions. You can just be like, give me a starting point. It gives you the starting point and then you can take that and and and refine it.

**Oriol Altisent** | 9:57

Mm-hmm. Nice and useful 'cause. You know, we so.

**Desmond Leo** | 10:03

The intent I'm hoping is is to make live seed make the life of market researchers easier.

**Oriol Altisent** | 10:09

Hmm.

**Desmond Leo** | 10:10

You are.

While you are different because you are a marketing data analysts, specialists, advanced analytics person, theoretically you're still in the body of of market researchers. I think if if you don't think so let me know.

**Oriol Altisent** | 10:29

Oh yeah.

**Desmond Leo** | 10:30

OK. And so then my question to you is is I do want to kind of get your opinion and in terms of. How you work through things, how projects come to your plate, you might focus on different areas, but I still want to know, you know how those areas are are done and and you know if there's any efficiencies that we can we can use.

So if we think back to a recent quantitative project, can you walk me through how? It actually played out.

**Oriol Altisent** | 10:58

Yeah. Yeah, I I going going through a standard project 'cause there's a lot of a doc things that that I'm doing that probably I don't want and never repeat again. So this is probably less interesting here.

But yeah, let's talk about a traditional project and and as you know that we have empathy, we have DBA, and we have salience plus, so salience plus DBA is always quite the structure is always the same, but the type of project that we have for empathy is a kind of the traditional one that. That we came with gonna start with with a brief.

We will receive the brief of the client or we'll receive some information from the client. And from there we have to understand the request. What? What they wanna do. And how we're gonna solve it. We're gonna propose to them.

And this was. I'm not gonna say pain point, but when I was thinking about this meeting, it was one of the things that I was thinking related with the problems that you were created. Creating that's something that can be used.

**Desmond Leo** | 12:11

When, when? When it comes to? Sorry if we if we just stop at that specific you know we're if I understand you correctly we're at the briefing stage you get together I mean even at the briefing stage you're trying to figure out what solutions to put in place.

**Oriol Altisent** | 12:17

Yeah.

**Desmond Leo** | 12:27

Does that require a lot of work on your side?

**Oriol Altisent** | 12:27

Yeah. Yeah.

Well, for you, you received the request from proposal. They work in the request from proposal.

It depends on the client because sometimes it's they go to a point and clearly they they know what they want and they they are talking about the solution or or it's quite clear from the proposal.

Sometimes our proposals is is 2030 page. Just. That they sent you have to read them so this can be kind of a playful point when when the client send you these RFPs that are more more than 20 pages.

So in other utterances, what we're doing is first up using using a prompt, summarise it and you have. Affect summary of what the client is is giving you.

You never as you know you never can trust in those kind of summaries because sometimes you forgot important information, so you always have to read it and go to it. But it helps a lot this kind of. Summary from and and it's something that I'm doing in any proposal that I receive going to to to summarise that as a first and then using it as a brainstorm to generate ideas.

What kind of solution they can they we can provide, how how we can. Like those issues, how to say that when you see the proposal, sometimes you're like unrealistic or something that is not, you know, gonna help.

But it's like if you have another person in the room proposing ideas, giving you ideas, and sometimes you can over a farm, what would you? You were already thinking or or propose sometimes different different ideas.

That probably you're gonna. You're gonna wait.

So as as a first step is something that that used to used to help in when we received them the the briefing from the client.

**Desmond Leo** | 14:21

None.

So I I I didn't intend to to. This is just a bit of a sidebar. I didn't intend to discuss what I'm about to discuss now.

Would you like a custom GPT specifically trained?

**Oriol Altisent** | 14:27

Hmm.

**Desmond Leo** | 14:38

On market research methodologies or books or like it is, how is it? How is the current? How is your current approach with with ChatGPT actually working and do you want improvements to it?

**Oriol Altisent** | 14:49

No, I I I don't think I I don't think I I need it or I want it. I is not gonna make. I don't think my my life easier.

But thinking what you were mentioning about a platform for market research and when you start a project you say OK I want to obtain that information.

If, if the artificial intelligence can tell you, OK, there is this three solution that can work here. That's something that that can help.

For example, I want to know which are drivers that that explain consideration in my category so. If if this is an intelligent directly, tell me oh, so you have these different options. You can do a driver analysis so you can include those kind of questions on the survey.

You can include a Max diff that is this kind of questions. Are you gonna obtain this kind of results so giving you these different solutions can help you to build a better a better survey. Or and to and to answer your question in a better way.

So you come initially with one problem. I want a price study.

So suddenly the tool is going to tell me if you want a price study, you have these different options. They've got Western Dome. You can do a conjoined, you can do Gable Granger.

So there is different solutions that it can give you to tackle that problem. And sometimes people are unaware of those solutions.

If we're creating a typing tool for a typing tool, a tool for market research, it will be great that they they they can see this different options from the beginning and then pick the one the the is one better answer the the questions.

**Desmond Leo** | 16:38

Oh, is that?

**Oriol Altisent** | 16:39

That, that, that, that was my my approach here and my idea.

I don't think it's gonna help me. Well, I I don't think we need to spend time for me to create this kind of thing because with what I'm doing right now is is more than enough.

I don't. I don't think wasting time. I'm creating a tool for me is gonna help but it was more thinking for researchers in general that that can also help them.

**Desmond Leo** | 17:07

How often are you? Do you do you at any point run into a? Bottleneck where you're where you're you have too many of these analysis that you need to to do and.

**Oriol Altisent** | 17:20

No, we we can have if if we if we have a maximum I'm gonna have is 2 proposals in a week and that's in a in a very busy week.

So it's not that these kind of RFPs come come very often. So for me again normally now we do not have any and some. And you can have to win a week maximum, so it's not, it's not something that that you need.

**Desmond Leo** | 17:41

Right.

So so the process is an RFP comes in, some of them are very long and detailed and some of them are very some, some of them are are straight to the point and you they know exactly what they want and it's very easy to understand like what kind of research to give it and what you know what to what to propose.

Other times it's it's longer and more detailed and and confusing to understand, but you have to spend the time to read through all 20-30 pages in order to understand to get a grip on what they're trying to accomplish, and then your role is to then. Basically help decide what type of research should be used, what what methodologies should be used to achieve whatever the goal is for for that RFP.

**Oriol Altisent** | 18:25

Yeah. Umm.

**Desmond Leo** | 18:26

OK. What happens after after that?

**Oriol Altisent** | 18:33

We need to include this in the proposal and I mean it's still not something that I think is related with what we are doing here but but yeah, it's you need to pick those ideas that that you you have from that. RFP and you, you have to include in a proposal to convince the client that this is the best solution and how to how to tackle it.

And that's probably maybe because my limitation is worth taking more time in my job and is is creating the proper the PowerPoints not for not not related to the content is more related about about the statics and the format and.

**Desmond Leo** | 18:53

Mm-hmm.

**Oriol Altisent** | 19:08

Kind of things and I think this is nothing that that we're gonna solve here at least probably not for the proposal state. Maybe yes, in the in the reporting state because I don't know if the tool is gonna have a kind of way of of display the results as well in a. In a nice way.

So that's something that, but that's in the last stage is gonna help in in the reporting. But yeah, not only for my skills. PowerPoint is something like always take took along but it's not something that.

**Desmond Leo** | 19:35

OK.

**Oriol Altisent** | 19:43

That came getting.

**Desmond Leo** | 19:43

Well, I mean, never seen ever. I mean I think what what I hear is that one of the things that take a long time for you is is actually just responding with a pretty PowerPoint because it's that formatting, NSD aesthetics that makes it difficult.

OK.

But let's say we get, let's say we win the project then. How does that? How does that process work with you?

**Oriol Altisent** | 20:02

Then next strategy is well is getting involved in a questionnaire.

You you will have to get involved in, in, in, in interviews or in the in the discussion with the stakeholders about about what we're gonna do, what we're not cheap, but basically it's going. Understand the problem more and create the questionnaire and that's that's the next step.

The questioner normally is is created by any of the of the researchers, but you have to to be sure that for example, if we have proposed a conjoin that the conjoin is is is well implemented in the. In the question, so is the proper question and the design is is the right one.

For example, if we are proposing a segmentation, we need the proper questions for a segmentation that not any question work for a segmentation.

So checking that all these things are are included in the in the survey in in the in the in the proper way for for them doing analysis.

**Desmond Leo** | 21:04

And that is. That responsibility is squarely on you then, to make sure that the the right questions are in place.

I mean, the researcher can design the questionnaire, but once it gets to like the advanced stuff, you have to make sure that you're you're the person that's making sure, like all the segmentation questions are correct and in the right place in the right. Yeah.

**Oriol Altisent** | 21:25

Yeah, yeah.

For an example is for segmentation. We use choice questions when when we use a best practises for including, for example, choice questions, you don't want to include questions about if I want to do a study about holidays and I ask you, do you like to go to the moon time? You wanna say yes and do you want to go to the beach? You're gonna say yes.

So with that information, I cannot do anything because everyone wants everything.

So the type of question I'm gonna ask is, what do you prefer going to the mountain or to the OR to the beach? And then you have to choose between one or the other.

So. Ensuring that that's a silly example, but we have these kind of questions instead of the previous ones at economical question. It didn't help me. A question like this is is going to help me to obtain a good segmentation.

So I I need to ensure that the questions that we include is for example a ranking. Question instead of a rating question or.

**Desmond Leo** | 22:22

Do we have? Do we have documentation at all on terms of what the best practises are for the for different things that we look out for? Or is it just in your head?

**Oriol Altisent** | 22:27

Yeah.

We have some kind of documentation about the kind of analysis that we can do and some kind of best practises in kind of analysis. The type of question is to include or something like this.

Yes, we have in not not detailed but we have some PowerPoints to help.

**Desmond Leo** | 22:47

Right. And and is this something that like there are there must be books about this then I mean how do you how did it, how did how have you come to learn about all of this stuff?

**Oriol Altisent** | 22:48

Yeah. What is?

**Desmond Leo** | 23:00

Just from training in school and and stuff like that or is?

**Oriol Altisent** | 23:02

No, no basically.

For for example segmentation is because I've been doing segmentation for 20 years.

So in in different companies you learn different things and and based on experience and for the the silly example about choice and having.

**Desmond Leo** | 23:07

Hmm.

**Oriol Altisent** | 23:17

Is something that is. It's quite common in some of the in some companies, but in other companies they never do it. And when you see their cementation you can see that end up being a disaster and it's a best practise to avoid.

So sometimes you saw you saw some, some. Some analyst or some people explaining how to the segmentation and they explain the rules like if they was kind of something an innovation of but a new one.

And sometimes you think well, this is the basic. So how you how you gonna apply it.

So we think that for. For a lot of companies, is is the normal the the basics, but there is a lot of companies that are not not using it.

**Desmond Leo** | 24:03

MMM.

**Oriol Altisent** | 24:05

So if if you go if you online, probably you're gonna find a different kind of of ways to do it. Options.

But in my experience I would say that that needs that's the option and that that's that should be the the basics.

**Desmond Leo** | 24:19

Are you ever involved in so? So you questionnaire review comes in.

You have reviewed it from the kind of like the advanced analytics perspective of it. Feed that back to the the researcher.

Do you find that to be a particularly time consuming process? Yeah.

Or is it just like, hey, by the way I've read it, you need to make these changes and then they go ahead and make it.

**Oriol Altisent** | 24:42

Well, it's it's time consuming because sometimes there is discussion.

The the question had to be approved for the client. So the client to look at it, for example the the statements that you include, you are driving analysis or and you need to include twenty statements, I might Steve.

**Desmond Leo** | 24:49

Yeah.

**Oriol Altisent** | 24:58

So there is always a discussion about how to reduce it, how to include more or less.

Again, I I use a tool in using an I using semantic embedding to try to find which which statements are very similar to each other and the ones that are very similar you.

**Desmond Leo** | 25:02

Yeah.

**Oriol Altisent** | 25:17

Recommend the client to remove them because you don't want to to have two statements that are too too similar.

And because always the client wants to include more statements than than the necessary ones. So these kind of tools help me a lot to reduce the number of statements in a questionnaire.

**Desmond Leo** | 25:28

And.

**Oriol Altisent** | 25:36

Before, before going to field work and I think this this can be time consuming, not very time consuming but more the discussion discussion with the client to the side.

If we have to remove it. If not, how we change it, how we frame it, how we ask those statements?

Which finding the proper statements for for a survey can be also time consuming or because because you can you can pick the statement from our previous survey or you can do a qualitative study to find the surveys and the statements.

So there is multiple ways. That you can find the.

**Desmond Leo** | 26:19

The the.

**Oriol Altisent** | 26:19

The proper statements, including in the survey, so again, it's something that maybe AI can help.

I use it some time to time to see which statements to include, which statements are are. Are following the.

**Desmond Leo** | 26:36

None. The best practises?

**Oriol Altisent** | 26:38

The Yeah, the proper structure.

Not again. We have a a couple of slides explaining how the statements needs to be. So I upload these slides or this information in charity BT and always ask, OK, look at these statements, review them where propose alternative is necessary or give.

**Desmond Leo** | 26:41

Yeah. Mm-hmm.

**Oriol Altisent** | 26:58

But for them, so it's always a step that we do. I wouldn't say that is time consuming, but it's a step that you you have to do is always try to reduce them, try to analyse them and try to find the best statement for the for the category.

**Desmond Leo** | 27:12

The You mentioned earlier that you you use a tool for semantic analysis to see how similar things are is that that's like a like an actual thing that you pay for that you that that's specifically designed for it or.

**Oriol Altisent** | 27:25

No, it's.

Is in I I use it in R in R it's like Python, yeah? So I.

**Desmond Leo** | 27:32

Oh, and art, yeah. Oh.

Oh, it's just. It's just a piece of code in R that you use to see.

**Oriol Altisent** | 27:37

Is a piece of colour card. Yeah.

And then you you have to, you have to call open AI and. Yeah, but then the amount of money that you cost is is is almost nothing.

So. So yeah, that's.

**Desmond Leo** | 27:47

Right.

**Oriol Altisent** | 27:50

That's how we use it.

**Desmond Leo** | 27:50

OK. So it's it's it's, it's still like an LLM. Call to just look for like what's similar and OK OK, cool.

Umm.

So you've gone through the review process, you've moved on to the, I mean, are you involved in? Ever in the the building the testing, the field work or anything. Or do you just you just it comes back to you and you've got data to then?

**Oriol Altisent** | 28:13

No.

OK, in theory I'm not involved in interviewing the questionnaire. When when he's in in, in, in field work or before going field work, I only. Get involved in creating the the survey, but here, here. Sometimes yes.

For example latest project that we have for for Moose grave. Yes I had to because we had some complex conjoined. I had to go to to a survey to see that the conjoin was just plaguing. Properly checking the versions, taking a look at it, but is not the is it not the normal thing?

Maybe because it was the first time we were doing a congen by ourself. I had to get more involved, but it's not something that I'm planning to to to spend time on on reviewing.

**Desmond Leo** | 29:01

OK.

**Oriol Altisent** | 29:06

Yeah.

**Desmond Leo** | 29:06

Right.

So the reason the reason why you were involved in checking for that one particular study was simply because the risk and uncertainty of having it being their first, the first one that we're doing by ourselves, OK.

**Oriol Altisent** | 29:17

Mm-hmm. Exactly.

And then for for new questions, we like to sometimes to to check if these kind of questions is. Are displaying properly, so we're we're trying a slide that for example that we haven't used before.

So we're trying to see how it was working, if if it's at, is it all in those cases that I get, I get involved in in reviewing the the salary.

**Desmond Leo** | 29:29

Hmm. Right.

**Oriol Altisent** | 29:42

Yeah.

**Desmond Leo** | 29:44

OK so.

Field work finishes at some point. You just get data and then that's when you come in to do more of the advanced analysis.

**Oriol Altisent** | 29:48

Mm-hmm. Yeah.

Well, right now I also get involved in exporting the data 'cause we are having we we are having some issues and problems, empathy to getting the data export properly.

Something that now with Liam. Probably. Yeah. I'm gonna get less less involved.

**Desmond Leo** | 29:59

OK.

**Oriol Altisent** | 30:14

But I also getting a bit involved on this.

So in in the part of exporting data is something that I'm. And.

**Desmond Leo** | 30:21

Exporting from empathy, IQ or exporting from Donata.

**Oriol Altisent** | 30:23

No. Right now if I use that forced platform, yeah.

**Desmond Leo** | 30:29

Forster. Yeah, yeah, yeah. The Forster platform. Yeah.

**Oriol Altisent** | 30:32

Using from there and and the main issue is because we cannot philtre out as far as I know, we don't know how to philtre there by way.

If we were duplicating respondents in some projects, so we were having a a lot of troubles, a lot of problems there. That I need to to get involved, clean the data, merge it properly.

That's one of them, of the of the parts that I I'm getting involved in theory in the future. I'm gonna get less involved.

Some kind of data transformations that I have to have. For example when we need to stack data, so it's something that I also I'm getting involved with.

Well in theory is something that. Hmm.

**Desmond Leo** | 31:21

So you're you're wearing many hats.

You're you're basically there's, I mean, we're a small company. So you're being like, OK, fine.

If someone needs me to export data and then transform the data so that it's actually usable because it's difficult to be able to like, I I'm guessing an example of what you're talking about is.

**Oriol Altisent** | 31:26

Yeah.

**Desmond Leo** | 31:38

Maybe you have multiple waves in one study and you expert that data, but then you still only want to show the latest.

You still only want to report on the latest wave, but the researchers themselves don't actually have a way to actually isolate those records to to then create a a unique file for that.

So you're the one that goes and and does that.

**Oriol Altisent** | 31:56

Yes.

**Desmond Leo** | 31:58

Yeah.

**Oriol Altisent** | 31:59

Yes, and the the the plan.

Well, yeah, there is different problems for different projects.

But what we were having in the past is the the the data files were becoming too big. So they started splitting the projects into and then they needed to merge them.

So they they stop.

For example, I think it was, for example, Musgrave. They stop the old survey in until 2424 and they continue storing data from 2025.

But they needed to merge it because this is a tracking study, so they want to to see the trends they want to see all the data.

So despite and they couldn't continue the having the, the the full study together because it was it was too big.

So then exporting in and managing it was it was too big. So what they decided. The rule is stop the project, start again, but we need to merge the data each time, so that was one of the things that they needed help to how to merge the data properly.

**Desmond Leo** | 33:10

Right.

When we say too big, it was too big for dinana to allow us to download because it takes too long or something like that.

**Oriol Altisent** | 33:15

Yeah.

When we were in too long, when we were in, in the four-star platform, every time you wanted to review something, when you have 1-2 million respondents, you become a massive.

**Desmond Leo** | 33:28

Right.

Massive, but there's it's not a limitation of SPSS like SPSS can have a million plus so so someone just needs to put the OR some or some or something like that. I'm I'm I'm I'm.

**Oriol Altisent** | 33:36

The.

Yeah, I think that there is also some limitation in the species like I never find an issue for that.

But yeah, we use Q in Q.

In theory we don't have as far as I know, any limitation. We didn't find it, but you can see that everything goes very, very slow when you start having lots and lots of data and lots and lots of variables as we were having.

So you become a bit messy, and then we were having problems merging data files because every time we were doing exporting we were exporting the whole data set one year data when we were merging it, we were duplicating waves so.

So then I I I need to get involved to only merge the current wave each time, but it's something that hopefully somebody IQ is not gonna not gonna happen when we export the data and we can export every time.

**Desmond Leo** | 34:25

Hmm.

**Oriol Altisent** | 34:34

Then the respondents upon it.

**Desmond Leo** | 34:40

OK. That's interesting.

At a future point, I might want to dig into how these long tracking studies operate and go into more of the specifics, but not not right now for this conversation.

**Oriol Altisent** | 34:56

And then there is a another thing in part of data processing, there is all the data cleaning all all this kind of stuff that some of them have been done during field work, but some had to be done then, then a doc depending on on the project.

**Desmond Leo** | 35:10

And. What are you looking for when it comes to a data cleaning or is it ad hoc each time?

**Oriol Altisent** | 35:15

There's a for example flat liners.

In some. In some respondents there's needs to look if if there's a question about how much you spend you, you're looking for for Old liars.

You want the answers make make sense.

All this kind, all this kind of of things you like. Take a look.

**Desmond Leo** | 35:18

Mm-hmm.

Right.

I mean, Leanne mentioned that she, this is stuff that she looks for as well and she does an R and stuff like that. Do you do another layer on top of that?

**Oriol Altisent** | 35:37

Hmm.

It depends on the study, The Flatliners. All these things sometimes is already on in field work, so you don't need to do it for for for questions about how much do you spent, sometimes you you need to you need to check it.

If Leon is doing it now, probably we're not gonna have that issue anymore. But in the past we we used to have it, so we needed to check there is a all different kind of analysis that depend on on the, on the speed of the response in in some kind of questions like an implicit response test for example this more related with. With DPA that that we also need to to do some data cleanings there.

If if some response take too much to to answer that question, it's not that he's doing anything wrong. But for that kind of analysis you cannot use that respondent. So you have to do this kind of of data cleaning for. For the respondents.

So that's.

**Desmond Leo** | 36:35

Oh, interesting.

Do you do you have a series of of defined rules that you look for when you're when you're cleaning things like you, you've just mentioned kind of one rule there in one specific situation.

I mean, is there like a defined?

**Oriol Altisent** | 36:47

In in that case, yeah, I have my my code in in R and when I run the code these kind of things are clean. I I need to I need to check it but but is I I I don't wanna know where the number line but but it's like if you are two times bigger than the median or or or three times bigger.

So I'm I'm looking looking. Something like this?

**Desmond Leo** | 37:08

OK, so you've already built it and you've already got a solution.

What what might be interesting is in a future conversation, maybe understanding what parts of the stuff that you've built, you might want to establish. Within our product as a standard that we then kind of roll out to everyone and we can be like empathy IQ, we put the empathy standard.

Here's all the different data cleaning criteria and and so forth. So just as just as an idea.

**Oriol Altisent** | 37:27

Hmm.

Yeah, but this kind of solution are more for DBA or salience plus because they have always the same kind of a study. The server is always the same. The questions that they ask, are they always the same? So then this this kind of tool makes sense for them when we are working with empathy normally.

**Desmond Leo** | 37:46

OK.

**Oriol Altisent** | 37:57

The the surveys are more a doc and you need to find a a solution survey by survey instead of an overall solution. Not gonna work for for everything if it's a tracking a study for sure you're gonna have the solution for for always, but if not.

**Desmond Leo** | 38:00

Hmm. Right.

**Oriol Altisent** | 38:14

Can be more more difficult to to replicate.

**Desmond Leo** | 38:18

Right, OK. And so?

**Oriol Altisent** | 38:21

For example, flatliners in in DBA we also look for flatliners, but sometimes it includes forest for statements.

You have 4 statements and you write the same in for the four statements. That's a normal thing to happen.

But when you have 20 statements or or 10 statements rating always the same you can say yeah. You are a flat liner.

Then you apply it with four statements. You cannot really remove respondents for this.

Yeah.

So so you you need to look at at this kind of of a small things per per study.

But overall the the rules are are the same for.

**Desmond Leo** | 38:43

Yeah, yeah.

But for statements is not that. Yeah.

Yeah, yeah, yeah, it's not. Right.

**Oriol Altisent** | 38:57

For all the stories that we have done, yeah.

**Desmond Leo** | 38:59

OK.

And so we you go through data cleaning and then you you probably run some sort of scripts to help do the segmentations and whatever the advanced analytics are.

How does what what? What is the output of that?

Like what do you do like? What do you? Does it go into a report?

Do you build a report to go to the researcher who then builds another report? Or?

**Oriol Altisent** | 39:20

Yep.

But. But.

I I sometimes I have to help with the waiting. Not always, but from time to time. I also help them with with the waiting.

And I I suppose that's something that we also count in there in the platform that if we want to to report results, we need to to include the waiting and with with Q again we also had few.

**Desmond Leo** | 39:30

Oh yeah.

**Oriol Altisent** | 39:44

Problems in the past up at some point when we create the waiting, we can we can look at it to try to to avoid it, especially when we are doing tracking tracking data.

**Desmond Leo** | 39:56

Do you understand the math behind waiting?

**Oriol Altisent** | 39:58

Yeah.

**Desmond Leo** | 39:59

So one day, if we ever need to get like, ask, ask for someone to be like I don't.

I understand waiting. I don't understand the math behind waiting. I don't understand how it applies to each individual record.

And one day if we in the future if we want to put that into the platform to have some sort of preliminary I like we're going to need to.

**Oriol Altisent** | 40:01

OK.

Mm-hmm. Yeah.

**Desmond Leo** | 40:23

So that that can be you, you, you you would know that.

OK, cool.

**Oriol Altisent** | 40:23

Yeah. Yes, perfect.

**Desmond Leo** | 40:32

Beyond waiting, then, you do the advanced analytics.

**Oriol Altisent** | 40:35

Yeah.

Then we're going to to analytics and again we have a a huge portfolio of different analytics that that we can run and we in theory have this IDs in the brief and proposal stage.

So when we include it on on the questionnaire, so we have decide which kind of analysis we are gonna do and. In the analysis, spaces of time to to run, to run it, and and to run it, and depending on the analysis, it takes more or or less time.

For example, a segmentation is something that used to take a lot of time. If it's arrival analysis quite fast. If it's a market, if it's very fast.

So depending of the. Of the analysis that we're doing takes longer or or shorter, we almost have more or less we have tools, tools of code, script to run any, any kind of analysis always need to be tweaked depending on the project.

But but for segmentation, yeah. Now we have our workflow that we run.

We need to of course work with the variables, prepare the variables to be ready to use, decide which variables to include, which ones not.

So that's what takes more of my time. Is this kind of decisions, which variables are the best one to include which not?

Preparing those variables, levelling them correctly, that's the thing that preparing this script or adapting the script or the workflow is what more time tames, and then when we run the analysis, looking at the results, because if it's a segmentation, you're gonna obtain different solutions and you could get to find and interpret which one is the best solution.

And then show it to the client. Discuss it with the team.

**Desmond Leo** | 42:15

But theoretically speaking, you're, I mean, deciding which variables to to use to run a specific analysis that's happening at the design phase.

**Oriol Altisent** | 42:24

If if it's a drive analysis or a Max. If yes.

If it's a segmentation now you decide what to include in the survey, but then you need to to look which variables are the ones that are more discriminating. 'cause if I were able, you could. I were able. It's not discriminating. Then it's better not including into the analysis, so that is. A particular case on segmentation that you need to understand which variables are are more estimating or more splitting the data better.

But same gonna happen if we do a forecasting model.

You're gonna have all information in the survey, but then you need to to find which variable. Really predict.

So then you. It's not that you're gonna use everything. And maybe in the first step you used everything, but then you need to decide which ones go out, which ones, which one the state is because if you are not explaining, you don't need to to stay on analysis.

If it's a then you don't need to do anything. It's like URL. Include those statements. Those statements are the ones who want the analysis. And that's all.

So depending on the type of the analysis you you need more or less.

**Desmond Leo** | 43:24

Yeah.

OK. Yeah, that's and that that makes sense.

Do you are you involved at all in in then kind of again the output is just like a something for the researchers to reuse and then put into a a report or do you are you involved in building the report as well?

**Oriol Altisent** | 43:49

Sometimes yeah, sometime now.

It depends on on, on, on the needs of of the team.

If the team had a very busy then you you've been more involved in in building the the reporter not not the report, but at least the the slides regarding my analysis.

So this can happen with distinctive bad for example, but again this is quite a structure and it's always the same, but maybe it's the only one that it happens and in a specific occasion for for some teams that they never have done analysis before, they don't know how to. How to plot it? How to present it?

Then you help them more.

But in theory it's not. It's not my my, my work is providing the results there or create a simulator or this this kind of analysis, but they they be in the report normally.

**Desmond Leo** | 44:25

Right.

OK.

**Oriol Altisent** | 44:41

But I mean, when I have to do it, I mean for me that always a painful a pain point because it takes a lot of time to to create a slight.

I'm not. I will go that PowerPoint, but that's more more my problem and another thing.

But in in this in this kind in this kind of platform I think for example Leanne was.

**Desmond Leo** | 44:52

None.

No, I think that you'd be surprised.

**Oriol Altisent** | 45:01

I think all these dashboard that directly you you have the result for the client.

So sometimes it's something that that can be solved there.

**Desmond Leo** | 45:10

So.

So then across all of the projects that you typical quant projects, so let's not do the distinctive that and the CPS and stuff like that across those the the ad hoc quantitative projects when you're dealing with researchers.

**Oriol Altisent** | 45:17

Mm-hmm.

**Desmond Leo** | 45:27

Umm.

Where do you feel? Where do you personally feel the most friction or or drag where things just kind of take a little bit too long and do you?

Are there any things that you you you would think?

I shouldn't be doing this not because not because I shouldn't be doing this, because it's not my job and you know.

But. But I shouldn't be doing this because it should be done. All automatically anyways.

**Oriol Altisent** | 45:56

Mm-hmm.

Yeah, there is a thing that need to be for me some sometime when when we need. Sometimes for some some slides, for example, in a tracking.

I think this should be done directly. You know you push a button because we are doing the same slides every day and the slides shouldn't be polated directly.

And I suppose that's something that now for example in muscle, if we're going to create them.

**Desmond Leo** | 46:17

Hmm.

**Oriol Altisent** | 46:27

The new survey for next year probably is something that we can we can include that directly from Q in at least for the for the monthly reports.

That's that's something that we need to work on it and and is gonna happen.

In the in the I try to automate as maximum as I can.

For example, when we run segmentation we obtain 100 solutions if times, so we need to find which of these 100 solution is the best one and I don't. I'm gonna waste my time looking solution by solution to find which one is the best one, but we already create some kind of a scorecard that is telling us providing some statistics that tell us which one is the most promising one and you end up looking at 3/4 of them and deciding which one is a better solution.

**Desmond Leo** | 47:16

MMM.

**Oriol Altisent** | 47:22

So we already tried to automate all all this kind of inefficiencies.

If I get involved in open hands for example, I'm gonna say that is not my job or not, but is is something that they are sometimes can be quite time consuming.

But again that's more related in in my case.

**Desmond Leo** | 47:41

For for is that that that's from a coding perspective you mean?

**Oriol Altisent** | 47:41

We've.

Yeah, in, in DBA, sometimes we do some kind of text analytics.

So you you need to to take a look to the results, do some data cleaning that's that's something that that sometimes is kind of time consuming.

**Desmond Leo** | 48:00

Mm-hmm.

**Oriol Altisent** | 48:01

Hmm.

But the the rest is is is more.

Yeah, you spend a lot of time in the in the analysis in the in the coding, in the data preparing sometimes.

**Desmond Leo** | 48:15

OK, OK.

**Oriol Altisent** | 48:15

This this can be kind of a a thing, but it's something that needs to be done and and and we also have tools to try to to automate it and do it as fast as possible.

So.

**Desmond Leo** | 48:27

Hmm, but it would be better if it was actually just came out coded in the first place.

**Oriol Altisent** | 48:32

Yeah.

**Desmond Leo** | 48:34

OK, that's interesting.

I think those are all the questions that I had for you.

I really wanted to understand your process and flow from a research perspective and I'm trying to get like again everybody's kind of, I just want to see if everybody's head is in the same place in terms of what. What? Creates. What friction points are the best friction points to resolve that are most common for everybody?

Because it then that becomes our proposition with with the software you know.

**Oriol Altisent** | 49:14

Yeah, but for for me, sometimes it's things that is more annoying or takes more time is in the data processing.

If the data don't come as you expected and then you need to work and reward the data because it doesn't come in.

**Desmond Leo** | 49:25

Hmm.

**Oriol Altisent** | 49:30

In the proper way.

**Desmond Leo** | 49:30

What? What does it? What does it mean to not come in the proper way?

**Oriol Altisent** | 49:34

What does it mean?

Can can come in multiple different things, but if if for for example that you have two variables with with the brands two different variables and the number of the brands, the brands are ordered in a different way.

So if if in. In variable one that is, which is your favourite brand, you have a list of brand leader list in number one in that in that in that list and if then there's a question about last time you went bugging. Little hard code 20, so these kind of things make up my work quite quite difficult.

So then you have to create a an A doc code. Take take a look at the numbers and everything match.

So these data mapping can be. Can be very, very time time consuming.

So for me that's what.

**Desmond Leo** | 50:30

Which which can be resolved so, so and but. But theoretically that's something that should be resolved at the questionnaire design stage.

**Oriol Altisent** | 50:36

In exactly.

**Desmond Leo** | 50:38

And then in the cases where it is not caught in the questionnaire design stage, then.

It takes time to then try and identify that it's not that that it's an issue and then and then basically write some code to like recode whatever answers so that it does match.

**Oriol Altisent** | 50:57

Yeah.

OK.

If it's a multi country survey, I mean in one country that question about which is your favourite brand, one is legal in another country. One the Tesco, when you manage all the assets, that's a mess.

**Desmond Leo** | 51:10

Yeah.

**Oriol Altisent** | 51:13

And that's what normally takes more time for me to to solve, to solve those issues.

But yeah, but it's it's basically these, these kind of things that can make my work from very easy. You receive the data, generalise it and you have results to, it's gonna take a whole.

**Desmond Leo** | 51:14

Yeah.

I can see.

Yeah, I can. I can see that being annoying.

**Oriol Altisent** | 51:33

Day to to make sense of the data and then you can start analysis.

**Desmond Leo** | 51:40

That's interesting.

That's interesting.

That's an interesting challenge, especially to even catch it, but I will keep that in mind as we kind of interview everyone else to see what other challenges there are across the studies.

I mean, like I. Part of me thinks we can probably build pretty quick. And simple internal tools and prototypes that might make things easier that that stuff doesn't go to clients.

It doesn't go customer facing. It doesn't need to go to production.

It's just like one off scripts. And it'll be just things that we test to say like if it works and we get it working and we think it's worth putting into a product, you know into the product maybe maybe I need to think through like how we handle unique challenges like that, but OK.

Excellent.

Thank you so much.

This was this was super informative I this covers this covers the the the entire spectrum of where I was, what I was hoping to cover, frictions, workarounds, how you work from from start to finish understanding the most time. Consuming components so it was super helpful.

Thank you.

And if it's OK, I'm gonna. I'm probably gonna reach out again to talk about some of the other things in a little bit more detail as we as we progress forward.

Umm.

Especially with, it sounds like if if you're a wealth of knowledge for waiting and stuff like that, once we once we do kind of explore that in the platform, then we'll we'll definitely need to lean on you pretty heavily to figure out what the the math is behind everything.

**Oriol Altisent** | 53:23

Perfect.

**Desmond Leo** | 53:24

Cool.

OK.

Thank you so much.

**Desmond Leo** stopped transcription