Meeting Title: Client Engagement Ownership Interview Date: 2025-12-15 Meeting participants: Jasmin Multani, Amber Lin
WEBVTT
1 00:02:33.860 ⇒ 00:02:34.760 Amber Lin: Hi there!
2 00:02:35.500 ⇒ 00:02:37.090 Jasmin Multani: Hi, Amber, how’s it going?
3 00:02:37.090 ⇒ 00:02:40.060 Amber Lin: Pretty good! How do I pronounce your name correctly?
4 00:02:40.450 ⇒ 00:02:41.740 Jasmin Multani: Jasmine?
5 00:02:41.740 ⇒ 00:02:46.899 Amber Lin: Jasmine, okay, I wasn’t sure if it was a Jasmine or Jasmine, I wasn’t sure.
6 00:02:46.900 ⇒ 00:02:53.510 Jasmin Multani: No worries, I get that. I’ve gotten just mean before, too, so I’m like, no Just basic jasmine.
7 00:02:53.700 ⇒ 00:02:56.130 Jasmin Multani: Am I pronouncing your name correctly?
8 00:02:56.130 ⇒ 00:02:57.500 Amber Lin: Yeah, just Amber.
9 00:02:57.500 ⇒ 00:03:01.160 Jasmin Multani: Okay, nice, nice, nice. Where are you based out of, Amber?
10 00:03:01.190 ⇒ 00:03:03.549 Amber Lin: I’m in LA. What about you?
11 00:03:03.550 ⇒ 00:03:05.990 Jasmin Multani: Oh, I’m in LA too! I’m in Santa Monica!
12 00:03:05.990 ⇒ 00:03:09.210 Amber Lin: Hey! Oh, wow, we’re so close, I’m in Culver.
13 00:03:09.500 ⇒ 00:03:12.280 Jasmin Multani: Oh, sweet! Sweet, sweet, how do you like Culver?
14 00:03:12.730 ⇒ 00:03:15.530 Amber Lin: I moved here, like…
15 00:03:15.860 ⇒ 00:03:23.060 Amber Lin: a few months ago, it’s been really nice. Yeah. It’s very different than where I lived before. I used to live in downtown, so it’s very different.
16 00:03:24.040 ⇒ 00:03:28.079 Jasmin Multani: Very different, yeah. Are you near, downtown Culver?
17 00:03:28.600 ⇒ 00:03:31.040 Amber Lin: I’m right across from Sony, actually.
18 00:03:31.250 ⇒ 00:03:31.950 Amber Lin: Have you been here?
19 00:03:31.950 ⇒ 00:03:40.289 Jasmin Multani: Cool, okay, okay, that’s really, that’s a really good location that you got. So you can walk down, you can, like, you’re walking distance to all the shops, right?
20 00:03:40.290 ⇒ 00:03:41.420 Amber Lin: Yeah.
21 00:03:42.100 ⇒ 00:03:42.730 Jasmin Multani: Sorry.
22 00:03:42.730 ⇒ 00:03:45.909 Amber Lin: Cool. Have you always been in LA? Been in Santa Monica?
23 00:03:45.910 ⇒ 00:04:04.129 Jasmin Multani: So I grew up in Orange County, went to New York for a few years, and then moved back, so it’s been, like, just like, taking a slice of different parts of LA, but I feel like, Santa Monica’s been convenient for me. The nightlife’s not great, but…
24 00:04:04.910 ⇒ 00:04:09.899 Jasmin Multani: I feel like it’s safe for a single woman living in her own apartment, you know.
25 00:04:09.900 ⇒ 00:04:12.060 Amber Lin: There are things happening at night.
26 00:04:12.650 ⇒ 00:04:25.049 Amber Lin: I see. Cool. I… we only have 30 minutes, and then we… I have so, so, so many questions for you. Robert, I think you already talked to Utam, and…
27 00:04:25.290 ⇒ 00:04:38.940 Amber Lin: Robert knows you, so I know from Robert that he already trusts you in front of clients, and he sees you as a very strong analyst and individual contributor, and mostly what we want to know is,
28 00:04:38.940 ⇒ 00:04:49.999 Amber Lin: engagement ownership of… are you able to… are you able to run the work? Are you able to own it in front of clients and own the roadmap? On that aspect?
29 00:04:50.310 ⇒ 00:05:00.260 Amber Lin: So, today I mostly want to ask about questions in that realm, and see if you have past examples, see how you, approach
30 00:05:00.460 ⇒ 00:05:05.430 Amber Lin: Essentially leading an engagement, leading a small team to handle that.
31 00:05:06.220 ⇒ 00:05:06.580 Jasmin Multani: Yay!
32 00:05:08.790 ⇒ 00:05:17.969 Jasmin Multani: Just a quick question, is this, like, a formal behavioral questions interview, or is this… I thought this was more of a copy chat, from what I gathered with them.
33 00:05:17.970 ⇒ 00:05:24.089 Amber Lin: It is, though I’m… those… that is the main focus area that I want to focus on.
34 00:05:24.090 ⇒ 00:05:24.930 Jasmin Multani: Alright, yeah.
35 00:05:24.930 ⇒ 00:05:31.850 Amber Lin: Of course, I’ll savor at the end, if you have questions for me and questions about the company, like, feel free.
36 00:05:32.170 ⇒ 00:05:35.920 Amber Lin: Feel free to ask me, I’ll make sure to save time at the end.
37 00:05:36.380 ⇒ 00:05:49.749 Jasmin Multani: Sure thing, yeah, yeah. Just wanted to, prime it correctly, because I kind of walked in just thinking this is a coffee shop. But, to answer your question, a little bit about me is that I started off in research.
38 00:05:49.750 ⇒ 00:06:01.370 Jasmin Multani: And pivoted to DoorDash as a, strategy and ops associate. Worked a little bit with Robert before coming over to TikTok as a data…
39 00:06:01.570 ⇒ 00:06:08.870 Jasmin Multani: data science analyst, many reorgs later. So, to answer your question, how do I lead
40 00:06:09.460 ⇒ 00:06:19.690 Jasmin Multani: engagement in a small group, truthfully, because I’ve had, engagement across such different industries, the way I’ve…
41 00:06:20.380 ⇒ 00:06:30.430 Jasmin Multani: try to, lead small groups is first by understanding individual strengths, as part of… within the… within the small operating group.
42 00:06:30.680 ⇒ 00:06:35.739 Jasmin Multani: So, let’s take an example from DoorDash. DoorDash, I was a…
43 00:06:36.730 ⇒ 00:06:49.910 Jasmin Multani: leading projects where I was, leading the strategy for, product redesign and actually launching those product redesigns. So… this is a project that took
44 00:06:49.960 ⇒ 00:07:05.739 Jasmin Multani: about 3 quarters, and had different evolutions based off of which stakeholders I needed to work with, according to the phase of, project completion. So, because it took 3 quarters, the first quarter was based off of,
45 00:07:06.180 ⇒ 00:07:08.949 Jasmin Multani: Creating proof. Proof of concept, right?
46 00:07:08.970 ⇒ 00:07:28.199 Jasmin Multani: So, because I started off at new verticals within DoorDash, that means that the main portion of DoorDash was already built, restaurant side, and we were emerging, new verticals, but using the main product that was formerly fitted for Restaurant Side.
47 00:07:28.260 ⇒ 00:07:40.899 Jasmin Multani: To… and making small tweaks so that as we launch new verticals, like groceries, alcohol, and so forth, we didn’t have to recreate the wheel.
48 00:07:41.260 ⇒ 00:07:59.779 Jasmin Multani: But of course, there are points where we realized, okay, because we’re using formal policies that were initially, designed for restaurant side, we were actually bleeding out a lot of money. And I’ll give you an example, right? So, for restaurant side.
49 00:08:00.180 ⇒ 00:08:07.119 Jasmin Multani: So this formal project was based off of, redefining credit and refund policies.
50 00:08:07.470 ⇒ 00:08:09.780 Jasmin Multani: When an order goes wrong.
51 00:08:10.650 ⇒ 00:08:15.130 Jasmin Multani: Because we just opted into whatever restaurant side was.
52 00:08:15.310 ⇒ 00:08:25.899 Jasmin Multani: Restaurant side, if a complete order is missing or incorrect, we end up giving, the full basket, the full subtotal back.
53 00:08:26.250 ⇒ 00:08:36.150 Jasmin Multani: Which makes sense, operationally, right? Like, if something is in… is missing from your order, that means…
54 00:08:36.500 ⇒ 00:08:42.869 Jasmin Multani: That the restaurant didn’t prepare it correctly, and the dasher, all they had to do was pick it up and leave.
55 00:08:42.870 ⇒ 00:08:58.190 Jasmin Multani: So the onus was on the restaurant side for doing things correctly, and the way our pay model worked was for missing items, we would give a full refund back to customer, but take it out of the restaurant side’s paycheck.
56 00:08:58.860 ⇒ 00:09:07.220 Jasmin Multani: Now, the model is a bit different as we, evolved into new verticals, right? For new verticals, let’s say groceries.
57 00:09:07.400 ⇒ 00:09:21.260 Jasmin Multani: The honest is actually on the Dasher, and therefore DoorDash, to get things right, because the Dasher is the one who’s picking the order and getting your vegetables, groceries, as accurately as you wanted it to be. So that original model
58 00:09:21.480 ⇒ 00:09:28.300 Jasmin Multani: of saying, hey, Amber, you’ve noticed that your bananas are missing from your order.
59 00:09:29.250 ⇒ 00:09:33.179 Jasmin Multani: Because of the way we opted in, now we’re giving you back
60 00:09:33.450 ⇒ 00:09:38.709 Jasmin Multani: $50, total subtotal, because you just happened to miss bananas.
61 00:09:39.100 ⇒ 00:09:40.010 Jasmin Multani: Right?
62 00:09:40.160 ⇒ 00:09:53.449 Jasmin Multani: Not only are we bleeding out, from, our personal pockets of overpaying you, but we’re also mischarging the wrong person, aka the merchant, which hurts our…
63 00:09:53.450 ⇒ 00:10:00.789 Jasmin Multani: Bottom line revenue promises to our merchants, to convince them to stay on the platform for a longer time.
64 00:10:01.370 ⇒ 00:10:09.300 Jasmin Multani: So, in order to… so this was a really big initiative, but because of so many moving parts and moving,
65 00:10:10.560 ⇒ 00:10:26.270 Jasmin Multani: people outside of the company that we had to coordinate with, we had to come up with a strong proof of concept. So, the first quarter was all based off of, reviewing 15,000 Salesforce tickets to size out how, A, how often is this happening.
66 00:10:26.830 ⇒ 00:10:30.249 Jasmin Multani: B, what is the,
67 00:10:31.040 ⇒ 00:10:39.539 Jasmin Multani: The dollar price that we’re burning through in terms of, like, overcharging, our merchant and paying too much back for the customer.
68 00:10:39.840 ⇒ 00:10:40.710 Jasmin Multani: C?
69 00:10:41.460 ⇒ 00:10:54.040 Jasmin Multani: tactfully saying, hey, we know this is a problem, but where in the app is this a problem? Is this a UX design, or is this a backend logic problem that we have to coordinate with back-end software engineers?
70 00:10:54.140 ⇒ 00:11:03.840 Jasmin Multani: And in this specific case, it was double. It was both the work. We had to not only redesign the, user interface for the support
71 00:11:03.970 ⇒ 00:11:07.150 Jasmin Multani: Drop-down list, but we also had to coordinate with…
72 00:11:07.620 ⇒ 00:11:16.620 Jasmin Multani: Not just the front-end engineers, but the back-end engineers, and also a separate payment, engineering team to ensure, hey.
73 00:11:16.660 ⇒ 00:11:27.879 Jasmin Multani: The logic may be sound, and the user design may be sound, but we also have to troubleshoot whether we’re coordinating with the bank accounts, that the accurate charge is being displayed for the customers.
74 00:11:28.470 ⇒ 00:11:29.710 Jasmin Multani: our wallets.
75 00:11:30.050 ⇒ 00:11:44.200 Amber Lin: I see. Very cool. I have a lot of questions that came up as I was listening to you. I guess first is that, how big was the team that you were leading, and who were the stakeholders that you had to coordinate with?
76 00:11:44.990 ⇒ 00:11:51.099 Amber Lin: So, the team was small and lean in terms of.
77 00:11:51.560 ⇒ 00:12:00.109 Jasmin Multani: coordinating with people. I was the sole representative of consumer support… consumer support…
78 00:12:00.620 ⇒ 00:12:08.860 Jasmin Multani: credit and refund, so it was me, my boss, and I, really just leading the strategy. But we were trying to influence
79 00:12:08.940 ⇒ 00:12:20.719 Jasmin Multani: all of new verticals, right? We were leading with grocery vertical examples, but we were opting… we were changing the policies for groceries, alcohol, flowers.
80 00:12:20.960 ⇒ 00:12:21.929 Amber Lin: I see.
81 00:12:21.930 ⇒ 00:12:22.590 Jasmin Multani: Yeah.
82 00:12:23.140 ⇒ 00:12:38.350 Amber Lin: like you’re communicating with a lot of the business stakeholders. I guess my… then my question is twofold. One, you present… you told me about a great strategy, how you designed the work, and I think that’s exactly what they’re looking for.
83 00:12:38.350 ⇒ 00:12:51.310 Amber Lin: how were you involved in designing that work, or was it more of, your boss had an ideation, and then you carried it out? How involved were you in setting the direction of the work?
84 00:12:51.970 ⇒ 00:13:01.409 Jasmin Multani: I owned it end-to-end. So, the first thing we had to do was review the Salesforce inbounds, which are, pre-filled.
85 00:13:01.520 ⇒ 00:13:09.819 Jasmin Multani: And sizing out, how often do we… a half… How often are we,
86 00:13:11.380 ⇒ 00:13:18.129 Jasmin Multani: Summarizing inbounds that are not found in the pre-filled selections.
87 00:13:18.960 ⇒ 00:13:24.589 Jasmin Multani: So, that was a very straightforward strategy, and, thankfully.
88 00:13:25.260 ⇒ 00:13:42.339 Jasmin Multani: I got to use my manager’s clout, because he had been there for years, so I was able to convince Andrew, hey, I can’t… this is a lot of work, 15,000 Salesforce tickets, can you convince the other managers and the other directors to opt-in and do,
89 00:13:42.340 ⇒ 00:13:52.620 Jasmin Multani: groundwork themselves as well. So he was able to coordinate with the other… and this is, like… New Verticals at that time was around, like, 60, 70 people, including directors.
90 00:13:52.620 ⇒ 00:14:07.649 Jasmin Multani: And we were existing at a time where even the directors would do groundwork. So that’s including, like, calling up the dashers, reading through someone’s forces. Of course, their volume is much less, like, I would just ask them, hey, can you do 10, review 10 inbounds?
91 00:14:07.810 ⇒ 00:14:17.990 Jasmin Multani: The higher up you are, and then, you know, the more you’re at my level, you’re gonna be doing higher volume. But that part was necessary so that they understood,
92 00:14:18.290 ⇒ 00:14:28.789 Jasmin Multani: the… the rhetoric I was trying to build, right? Having higher-up people doing ground research as the research is unfolding helps them understand
93 00:14:29.370 ⇒ 00:14:33.600 Jasmin Multani: The… the actual analysis.
94 00:14:34.280 ⇒ 00:14:38.060 Jasmin Multani: So… Does that answer your question, or.
95 00:14:38.060 ⇒ 00:14:52.200 Amber Lin: Yeah, yeah, I just wanted to know of, were you the one that came up, when came up with this plan that sounded awesome? So, I guess the next point, since you’ve talked about talking to the execs and convincing them.
96 00:14:52.200 ⇒ 00:15:10.490 Amber Lin: to buy into your plan, how do you conduct a business review like that? How do you conduct these meetings with these executives? How do you have a regular cadence of reviews? How do you structure them? What is the narrative that you take? I want to understand that.
97 00:15:11.430 ⇒ 00:15:22.510 Jasmin Multani: Yeah, so thankfully, some of this work was… some of the cadence work was already structured out for me when I moved in. So, we often have… we had various weekly reviews.
98 00:15:22.510 ⇒ 00:15:31.709 Jasmin Multani: According to pods. So every Tuesday, we had an all-hands with all of the 60-plus stakeholders within new verticals.
99 00:15:31.910 ⇒ 00:15:41.139 Jasmin Multani: And we would just review each other’s work, right? And we would provide project updates as things are being unfolded. And that is staying accurate.
100 00:15:42.090 ⇒ 00:15:48.699 Jasmin Multani: And linking all of your information for those weekly project updates is essential to build trust.
101 00:15:48.930 ⇒ 00:15:50.539 Jasmin Multani: And each week.
102 00:15:50.690 ⇒ 00:15:57.779 Jasmin Multani: like, reminding… I’m reminding you, this thing took 3 quarters, and along the way, I had to convince the exec…
103 00:15:58.090 ⇒ 00:16:04.710 Jasmin Multani: Our senior directors, what direction we’re going in, why we’re going in that direction, and the current project status.
104 00:16:04.900 ⇒ 00:16:09.950 Jasmin Multani: So, it’s a matter of…
105 00:16:11.560 ⇒ 00:16:28.110 Jasmin Multani: part of it is quarterly planning and saying, hey, the work that I’m doing, it’s going to unlock growth, right? The VP wants us to focus on growth and net new customers, so I had to use rhetoric in my credit and refund.
106 00:16:28.190 ⇒ 00:16:35.510 Jasmin Multani: Project to say, hey, we’re gonna be able to, retain all this money and this bleed out.
107 00:16:35.710 ⇒ 00:16:46.629 Jasmin Multani: So we’re bringing back money to the, rush, to the business, but we’re also deepening, trust with the customer. And trust is being measured by CSAT,
108 00:16:46.840 ⇒ 00:16:49.800 Jasmin Multani: Customer satisfaction and retention.
109 00:16:50.230 ⇒ 00:16:53.359 Jasmin Multani: Those two are very helpful metrics.
110 00:16:54.250 ⇒ 00:17:07.459 Jasmin Multani: That are in line with the VP’s overall vision at that time to unlock growth. And week over week, I would describe, okay, this is the retention, this is how customers have
111 00:17:07.470 ⇒ 00:17:15.369 Jasmin Multani: Reported their NPS scores, their CSAT scores, this is the week-over-week trends,
112 00:17:17.609 ⇒ 00:17:19.489 Jasmin Multani: I’m trying to remember what else.
113 00:17:21.390 ⇒ 00:17:29.480 Jasmin Multani: And every week over week, just being honest with, hey, what are the highs of this project, what are the lows of this project, and what are learnings?
114 00:17:29.480 ⇒ 00:17:32.440 Amber Lin: That’s something Rash especially emphasized.
115 00:17:32.440 ⇒ 00:17:39.169 Jasmin Multani: How do we take learnings from this use case scenario from grocery and adapt it so that,
116 00:17:39.460 ⇒ 00:17:50.909 Jasmin Multani: Other verticals, like alcohol, flowers, and convenience, can opt in with… without having to do as much robust work, that we were doing for grocery.
117 00:17:52.420 ⇒ 00:18:00.890 Amber Lin: Gotcha. That sounds really… that sounds really cool, and I was thinking… you said on this project it was, you…
118 00:18:01.150 ⇒ 00:18:11.959 Amber Lin: mostly just you, and then you had to convince all these VPs to buy in. Did you have all the resources you need, or did you have to fight for them along the way? What was that like?
119 00:18:12.840 ⇒ 00:18:16.920 Jasmin Multani: Let’s see, trying to remember…
120 00:18:17.100 ⇒ 00:18:20.059 Amber Lin: Because of the bleed-out was so…
121 00:18:20.580 ⇒ 00:18:29.629 Jasmin Multani: I’d say the buy-in to redesign was pretty straightforward, because…
122 00:18:30.930 ⇒ 00:18:47.549 Jasmin Multani: at the mid… the way Strategy and Ops talks to data science and engineering is that we start our quarter saying, hey, these are the projects that we want to do. Mid-quarter, we’ve… we’ve done the experiments and we’ve done the analysis to say, hey, this is our proof of concept.
123 00:18:47.740 ⇒ 00:18:57.240 Jasmin Multani: And then from mid-quarter to late quarter, we, evaluate with the data scientists and kind of compete.
124 00:18:57.960 ⇒ 00:19:17.320 Jasmin Multani: why you should take my idea, my experiments, or other people’s experiments, and it’s just a matter of, proving through experiments, A-B testing, working closely with product managers week over week and saying, hey, what is preventing you from, accepting my, my projects?
125 00:19:17.400 ⇒ 00:19:25.590 Jasmin Multani: That’s not to say that things are perfect, there have been times where I’ve handed things off to data scientists.
126 00:19:25.730 ⇒ 00:19:45.570 Jasmin Multani: And they’ll run a two-quarter retention analysis, and at the end, they’ll say, oh, there’s been no impact to retention. And I’d have to be like, that’s BS. There has to be. And in that specific scenario, I had to push back to the data scientist. I’m like, how’s that possible?
127 00:19:45.820 ⇒ 00:20:01.110 Jasmin Multani: If someone gets blueberries… if you get rotten blueberries, and the original credit and refund is 25% off, or 25% back, you’re only getting $1 back for your blueberries, whereas our alternative,
128 00:20:01.210 ⇒ 00:20:19.489 Jasmin Multani: alternative scenario is you get full money back for your rotten blueberries. You’re telling me that the customer doesn’t care enough to retain? That is BS, and I had to push back and ask, hey, how did you measure retention? Did you measure retention based off of
129 00:20:19.510 ⇒ 00:20:25.439 Jasmin Multani: Is the customer returning to the platform, aka also doing restaurant orders?
130 00:20:25.440 ⇒ 00:20:31.010 Amber Lin: Or is the… is the, customer coming… returning back to the grocery line?
131 00:20:31.130 ⇒ 00:20:42.350 Jasmin Multani: And I immediately called it out, and they were able to admit, hey, we actually just did retention based off a platform. We didn’t care to slice it up between.
132 00:20:42.350 ⇒ 00:20:42.910 Amber Lin: Oh, no.
133 00:20:42.910 ⇒ 00:20:47.099 Jasmin Multani: retention on grocery, alcohol, so on and so forth. So that…
134 00:20:47.730 ⇒ 00:21:02.320 Jasmin Multani: of time, and I was very upset. I was like, I gave you guys this information, I trusted you guys to scale this out, and you guys spent two months, like, two quarters doing this analysis, running this experiment without checking in.
135 00:21:02.380 ⇒ 00:21:15.800 Jasmin Multani: So, at that point, they were like, okay, well, our analysis said it’s null, so we’re gonna revisit this, experiment, at the start of the year. So that part was very frustrating.
136 00:21:15.800 ⇒ 00:21:16.360 Amber Lin: Huh.
137 00:21:16.360 ⇒ 00:21:19.840 Jasmin Multani: But I very much publicly called it out, like.
138 00:21:20.010 ⇒ 00:21:29.949 Jasmin Multani: this is a nice experiment to prove that, like, the rest of our businesses don’t suffer. However, this doesn’t, mean that we should not change this policy. This is, like.
139 00:21:30.720 ⇒ 00:21:40.590 Jasmin Multani: a very obvious business decision. We were supposed to use this experiment to prove bleed-out, additional bleed-out. This was not supposed to be, like.
140 00:21:41.660 ⇒ 00:21:52.050 Jasmin Multani: a strong no for a policy. So I just publicly called it out. I was just like… I kind of used business logic as evidence, saying, this is a no-brainer.
141 00:21:52.190 ⇒ 00:21:54.160 Jasmin Multani: There are certain…
142 00:21:54.590 ⇒ 00:22:02.369 Jasmin Multani: things that we need to decide on, based off of empathy and customer empathy, rather than, oh, the numbers said no.
143 00:22:02.370 ⇒ 00:22:15.110 Jasmin Multani: So thankfully, that following year, they were able to run, an experiment, with the correct conditions and the correct analysis, and eventually they said, okay, for rotten blueberries.
144 00:22:15.170 ⇒ 00:22:24.180 Jasmin Multani: you will get full refund. But at that point, I was placed on a different project, and the main engineering efforts were done.
145 00:22:24.320 ⇒ 00:22:32.160 Jasmin Multani: with the, product rollout, and re-changing the UX design. So at that point, it was just a matter of,
146 00:22:32.900 ⇒ 00:22:44.369 Jasmin Multani: changing the percentages. And that… at that point, I got to be hands-off, I got to focus on other things, but there was enough documentation, to say, hey, you don’t need me to lean in.
147 00:22:44.780 ⇒ 00:22:52.290 Jasmin Multani: to, roll out this NPs, just change the percentages and do this A-B test for.
148 00:22:52.880 ⇒ 00:22:53.440 Amber Lin: Gotcha.
149 00:22:53.440 ⇒ 00:22:54.949 Jasmin Multani: check, and it was fine.
150 00:22:54.950 ⇒ 00:23:08.889 Amber Lin: Okay, it sounds like you’re operating at a pretty high level, a very strategic direction, and then you’re able to hand off some of the tasks today as scientists, and then point them in the right direction if they’re not doing well.
151 00:23:08.890 ⇒ 00:23:17.220 Amber Lin: How do you… I guess my question is, how do you think about what you do yourself, and what you should
152 00:23:17.230 ⇒ 00:23:31.669 Amber Lin: where you will hand off, like, how do you find the line between there? Because you only have so much time, and your experience is really valuable, so I want to know how you handle that line of what you let other people do, and what you have to do yourself.
153 00:23:32.520 ⇒ 00:23:42.229 Jasmin Multani: Yeah, so I… guess I can give an example from TikTok itself. One of my project leads tapped me on the shoulder and said, hey,
154 00:23:43.040 ⇒ 00:23:51.440 Jasmin Multani: because you’re tracking for… Jasmine, because you’re tracking for who’s selling guns on the platform illegally,
155 00:23:51.850 ⇒ 00:23:59.670 Jasmin Multani: you can scale this project out through an LLM. I’m like, cool, okay, didn’t ask for this project, it’s a very interesting scope, I’m happy to lean in.
156 00:23:59.920 ⇒ 00:24:02.540 Jasmin Multani: At that point.
157 00:24:03.550 ⇒ 00:24:11.930 Jasmin Multani: It was very messy, I will say, and I’ve given that feedback, to the engineers. Project scope was very messy.
158 00:24:12.540 ⇒ 00:24:21.459 Jasmin Multani: me, understanding what my role is was very messy, and where do I lean on engineers was very messy. But the way…
159 00:24:23.190 ⇒ 00:24:27.740 Jasmin Multani: I leaned in was trying to ask.
160 00:24:28.910 ⇒ 00:24:48.730 Jasmin Multani: trying to just first monitor what are the types of questions that are being asked, and at what point do I get pushback? And, for example, I was told, hey, the golden set that you’re curating has to be a certain way, but my pushback was, well, bad actors don’t always use
161 00:24:50.840 ⇒ 00:24:59.100 Jasmin Multani: proper grammar to signal out, whether they have a gun to sell or not. Given our LLM model,
162 00:24:59.640 ⇒ 00:25:08.859 Jasmin Multani: And what I’ve been noticing are being successfully… what’s being successfully scored, golden sets with, like, proper grammar.
163 00:25:08.970 ⇒ 00:25:19.600 Jasmin Multani: Text with proper grammar is what should be opted in into this golden set, and the rest of the junk should be assessed, through our high-level detections.
164 00:25:19.890 ⇒ 00:25:30.140 Jasmin Multani: And the pushback I got was, well, Jasmine, you shouldn’t be deleting, Golden Set… parts of the golden set, you need to assess it with, your teammate, who’s another analyst.
165 00:25:30.330 ⇒ 00:25:42.190 Jasmin Multani: And so that’s my first guess. It’s like, okay, so the engineers do not want to be, involved in the data purity level. They just want to have,
166 00:25:42.740 ⇒ 00:25:51.120 Jasmin Multani: a high level of how the model is working. They want feedback of how the model is working, and they want feedback as to what,
167 00:25:51.360 ⇒ 00:26:01.749 Jasmin Multani: how do I generalize aspects of the golden set to tell them, hey, when you pair, this golden set with the Dolphin LLM, you get this type of success rate.
168 00:26:01.750 ⇒ 00:26:19.039 Jasmin Multani: So, that’s something I noticed, from my engineering stakeholders. For my analyst stakeholders in that specific example, I went to him and said, hey, I think this should be dropped off, this should not be included in the golden set for XYZ reasons, and he said, sure, that’s it.
169 00:26:19.320 ⇒ 00:26:28.030 Jasmin Multani: So that part was frustrating for me, where I had to be told, hey, Jasmine, you’re not allowed to sign off on things on your own, you need
170 00:26:28.880 ⇒ 00:26:37.049 Jasmin Multani: another analyst, you need to brainstorm it with another analyst, before moving things forward. However.
171 00:26:37.220 ⇒ 00:26:45.350 Jasmin Multani: My issue was, this is very bureaucratic, B, it’s slowing me down. C, the other analyst does not have an opinion.
172 00:26:45.450 ⇒ 00:26:53.820 Jasmin Multani: And that’s especially… that’s because my scope is my scope, his scope is his scope. His scope was based off of violent comments.
173 00:26:54.110 ⇒ 00:27:10.349 Jasmin Multani: inciting… inciting murder, my scope is people selling guns. Because we’re use… we’re working on different data fields, he’s working on comments, I’m working on videos, and because we’re working with different bad actors,
174 00:27:11.210 ⇒ 00:27:20.490 Jasmin Multani: A… it’s not like our brainstorming sessions were very robust, so that was my feedback to the engineers, saying, hey, this is…
175 00:27:20.910 ⇒ 00:27:36.940 Jasmin Multani: this rule is actually slowing me down. We need to find a better reason, for us to collaborate, to get sign-off from each other. Otherwise, just let me do this on my own, and I’d much rather do,
176 00:27:38.480 ⇒ 00:27:48.569 Jasmin Multani: I guess, like, not so much A-B tests, but, like, I’d rather, size out the impact for each difference in golden set that, that I have.
177 00:27:48.570 ⇒ 00:28:01.339 Jasmin Multani: especially for meeting twice a week, especially if we have other projects going on. Coordinating with someone who’s on the East Coast while I’m on the West Coast, and waiting for their sign-off is… is actually an added friction point.
178 00:28:02.620 ⇒ 00:28:07.650 Amber Lin: I see. That was all of my questions. I… we have, I think.
179 00:28:07.970 ⇒ 00:28:11.690 Amber Lin: 5-ish minutes, can I answer any questions you have?
180 00:28:11.890 ⇒ 00:28:15.620 Jasmin Multani: Yeah, of course. So, I mainly wanted to ask,
181 00:28:15.840 ⇒ 00:28:20.970 Jasmin Multani: At a high level, like, what are your… what’s your work scope?
182 00:28:21.260 ⇒ 00:28:25.530 Jasmin Multani: What type of clients do you work with, and what does your day-to-day largely look like?
183 00:28:25.530 ⇒ 00:28:27.880 Amber Lin: My day-to-day, or the company’s?
184 00:28:28.580 ⇒ 00:28:30.020 Jasmin Multani: Your day-to-day.
185 00:28:30.020 ⇒ 00:28:43.630 Amber Lin: Cool. So, I joined as a project manager, and I eventually went back to being a data analyst, because when I started, they had no project manager resources. So right now, I cover…
186 00:28:43.950 ⇒ 00:29:03.030 Amber Lin: two to three clients at a time, and my analysis scope… I would say they’re still quite early, because we started analysis for clients after we set up the data and helped them do the data modeling and all that. So, analysis is relatively new for this company, and so a lot of it is in the…
187 00:29:03.080 ⇒ 00:29:18.069 Amber Lin: diagnostics, finding problem stage. We only recently started doing experiments, because the clients just barely reached that stage after finding insights. So, my day-to-day would look like we have a meeting in the… in the morning.
188 00:29:18.340 ⇒ 00:29:27.660 Amber Lin: Cover stand-ups, cover responsibilities, and then throughout the day, it’s either I do my own analysis, or I meet with clients to sync with them on…
189 00:29:27.720 ⇒ 00:29:40.430 Amber Lin: findings I found to help them to follow through on items that they need, and also internal revisions looking at decks. So my analysis scope,
190 00:29:40.890 ⇒ 00:29:51.819 Amber Lin: My clients right now are mostly e-commerce CPG, so, and then we have other clients that’s in SaaS,
191 00:29:51.940 ⇒ 00:29:53.420 Amber Lin: That’s more…
192 00:29:53.790 ⇒ 00:30:10.209 Amber Lin: more digital products and services, but I mostly do analysis, on the marketing, and a little bit of… tiny bit of product, tiny bit of supply chain, but mostly marketing analytics for the CPG e-commerce companies. And so…
193 00:30:10.430 ⇒ 00:30:14.350 Amber Lin: Usually a short cycle, or a short,
194 00:30:14.550 ⇒ 00:30:19.869 Amber Lin: Not an epic, a short story would be, we have a question that the client’s interested in.
195 00:30:19.970 ⇒ 00:30:32.960 Amber Lin: We do the initial analysis, we prepare a deck for an end-of-week review, and then during that review, the client might ask questions, and then we find out the next steps of what we’re interested in, and then we continue down that route.
196 00:30:33.150 ⇒ 00:30:34.610 Jasmin Multani: And…
197 00:30:34.960 ⇒ 00:30:40.649 Amber Lin: I think at a higher level, I think the The company mostly.
198 00:30:40.770 ⇒ 00:30:41.960 Amber Lin: has…
199 00:30:43.220 ⇒ 00:30:53.360 Amber Lin: I would say every quarter or so, or every two months, we brainstorm with our clients of, okay, this is the direction that we want to
200 00:30:53.500 ⇒ 00:31:08.629 Amber Lin: head in, these are some possibilities we recommend you to go explore more in that direction. That’s usually what Robert handles, and I help support him to plan out some roadmaps, do the slides, make sure the client knows what
201 00:31:08.630 ⇒ 00:31:21.150 Amber Lin: where we are able to go, and then that helps inform new scope, if we want to introduce more analysis work, more engineering work, more stuff on the marketing technology, so…
202 00:31:21.200 ⇒ 00:31:22.080 Amber Lin: I think.
203 00:31:22.250 ⇒ 00:31:29.819 Amber Lin: On the higher level, that’s the cadence of the company. On my level, my cadence is pretty much every single week. There’s a new analysis to do.
204 00:31:30.310 ⇒ 00:31:33.650 Jasmin Multani: Alright. Do you have one more minute?
205 00:31:33.650 ⇒ 00:31:34.270 Amber Lin: Yeah.
206 00:31:34.490 ⇒ 00:31:43.349 Jasmin Multani: Okay, what type of analysis do you perform, then? Are they usually RCAs? Are they pre-post analysis? Ad hoc?
207 00:31:44.120 ⇒ 00:32:01.870 Amber Lin: Whatever comes up, because right now we are limited in analyst resources. We just got someone in product analytics and just got someone that’s more skilled in forecasting. So, I would say my… my scope has been…
208 00:32:02.630 ⇒ 00:32:08.159 Amber Lin: very different. It… it mostly covers a lot of stuff.
209 00:32:10.390 ⇒ 00:32:28.139 Amber Lin: I’ve done, on the market analysis, I’ve done a lot of cohort analysis, life cycle analysis, did a bit of pre-post event comparison, and then… and then also was assigned to some of the product analytics stuff, pricing analysis, so…
210 00:32:29.090 ⇒ 00:32:40.299 Amber Lin: I think when you come on board, you’ll have to cover the reviews or cover the planning for a big amount of scope, so you will cover more… you will be reviewing
211 00:32:40.500 ⇒ 00:32:44.750 Amber Lin: We’re planning out more than the scope that I would be covering.
212 00:32:46.470 ⇒ 00:32:47.090 Jasmin Multani: Okay.
213 00:32:47.320 ⇒ 00:32:50.169 Jasmin Multani: Maybe I can follow… who would be the best person to follow up?
214 00:32:50.910 ⇒ 00:32:51.550 Jasmin Multani: In that.
215 00:32:51.550 ⇒ 00:32:53.709 Amber Lin: In terms of scope that you’ll be covering?
216 00:32:53.710 ⇒ 00:32:54.470 Jasmin Multani: Yeah.
217 00:32:54.540 ⇒ 00:33:07.500 Amber Lin: That will be Robert. I think what I heard from Robert, what he wants you to be, or to become in this company, someone that can plan out the engagements, handle,
218 00:33:07.500 ⇒ 00:33:15.029 Amber Lin: The relationship with the clients, think about if there’s potentials for upsells, and also be the person who…
219 00:33:15.190 ⇒ 00:33:19.179 Amber Lin: Reviews or assigns the more junior analysts’ work.
220 00:33:19.320 ⇒ 00:33:20.410 Amber Lin: So…
221 00:33:20.870 ⇒ 00:33:35.359 Amber Lin: someone junior or someone mid-level, they would report to you before they report to Robert, because his time is very, very swamped. So, essentially, just like how you assign work to the data scientist at
222 00:33:35.680 ⇒ 00:33:55.269 Amber Lin: At DoorDash, we would be assigning work to some of the people more junior than you, and then checking if that makes sense, checking if that pushes on the outcome for the clients. So it’s essentially whatever scope that the client would end up needing, which means it will cover a very broad base of
223 00:33:55.490 ⇒ 00:33:56.760 Amber Lin: analysis.
224 00:33:57.030 ⇒ 00:33:58.600 Jasmin Multani: Okay. Okay, okay.
225 00:33:58.720 ⇒ 00:34:01.849 Jasmin Multani: Maybe that’s something I can follow up with Robert further on. Yeah.
226 00:34:01.850 ⇒ 00:34:09.900 Amber Lin: Yeah. I bet they’ll… I bet you’ll be talking with him. If… if not on your own, then also a last call will be with him.
227 00:34:10.270 ⇒ 00:34:12.720 Amber Lin: Okay, sounds good! Yeah.
228 00:34:12.719 ⇒ 00:34:21.920 Jasmin Multani: I know we’re at time, but thank you so much for… for these questions, giving me insight, as to what your work was. Let me know if you have any other questions for me as well.
229 00:34:21.929 ⇒ 00:34:37.619 Amber Lin: Of course, I appreciate it. Sorry I didn’t know that this was meant to be a coffee chat. I came in full blast, because I just got very short notice. I was like, this is an interview, and I need to do this, so… but I’m really glad I got to know more how you think about work.
230 00:34:37.619 ⇒ 00:34:43.999 Amber Lin: Because I only gather so much from your LinkedIn and your conversation at UTAM, so I’m glad to think…
231 00:34:43.999 ⇒ 00:34:45.619 Amber Lin: To know more about your brain.
232 00:34:45.620 ⇒ 00:34:55.570 Jasmin Multani: Of course, yeah, I feel like LinkedIn… I just keep it light and sweet, so that that leaves more questions for people to discuss in person. So, I think it’s worked out great.
233 00:34:55.820 ⇒ 00:35:01.729 Amber Lin: Awesome! Alright. You have our contacts, and I know you’re talking to Damlade later.
234 00:35:02.450 ⇒ 00:35:04.210 Jasmin Multani: Sounds good. I’ll talk to you soon.
235 00:35:04.210 ⇒ 00:35:05.510 Amber Lin: Yeah, have a good one. Bye.
236 00:35:05.510 ⇒ 00:35:06.070 Jasmin Multani: Bye.