How Photon Used Automations to Change E-prescriptions | Michael Rado, CPO, Co-Founder, Photon Health

In this episode of Ops I Did It Again, Michael Rado (aka Rado), Co-Founder and CPO at Photon Health joins the thinksquad aka Danielle and Nikhil to break down how they built and scaled products to support over 20,000 e-prescriptions a month. Rado shows a live demo of their build (on YouTube version) and candidly shares his learnings on how to scale with optimization and automation in mind.
Hosted By:
Nikhil Krishnan & Danielle Poreh
Michael Rado

Show Notes

In this episode of Ops I Did It Again, Michael Rado (aka Rado), Co-Founder and CPO at Photon Health joins the thinksquad aka Danielle and Nikhil to break down how they built and scaled products to support over 20,000 e-prescriptions a month. Rado shows a live demo of their build (on YouTube version) and candidly shares his learnings on how to scale with optimization and automation in mind. 

This episode is sponsored by Out of Pocket, because no one is prouder than us than us:

To register for the upcoming Healthcare Call Center 101 crash course course visit:; Use code: ANSWERS for $100 off; Next cohort starts 4/16- 5/2

To register for the upcoming Healthcare 101 crash course course visit:; Use code: IBELIEVEINME for $100 off; Next cohort starts 4/23-5/9


Danielle Poreh (

Nikhil Krishnan (


Michael Rado (


(00:00) Intro

(01:45) Michael Rado and his journey to Photon

(03:14) How Photon Health works

(07:36) Courses by Out Of Pocket!

(16:49) What photon automates and measures

(24:23) The approach towards product development at Photon Health 

(29:25) Practical experiments for orgs to try

Podcast Transcript

Nikhil: [00:00:00] () So we just wrapped up our podcast with Rado at Photon Health. They're building some very cool stuff in the e-prescribing space. When people use apps that are really simple and intuitive, they basically take for granted all the janky stuff that happens the back end to make that happen.

Nikhil: So in pharmacy, he talked to us through e-prescribing, through faxing automation, through setting up phone queues for their call center agents. It was pretty fascinating stuff.

Danielle: Yeah, and their evolution is really extraordinary. I mean, they went from like 0 to 20,000 real quick. So, their scale is massive.

Danielle: And we got this, view into Rado, who's the co-founder, their leading product, how they thought about both product design and their whole backend architecture from day 0 up until now. We get a demo, so if you're watching on YouTube, you'll literally see what they're building at Photon. And he gave us a bunch of frameworks that they use that really like drives how they make product decisions, [00:01:00] which to me were all very different from what I've been indoctrinated in.

Danielle: I think any product ops or early stage person will gain a lot from this one.

Nikhil: Take a listen, Try out some of the stuff he suggested and let us know what you think. So we're here with Michael Rado who prefers Rado. So we're going to call him that going forward. So Rado, is one of the co-founders of a portfolio company of mine, so obviously I get to shill my own wares on this one, which is fun. you know, I've been keeping in touch with the Photon group for a long time, and they've been doing some really cool things on the e-prescribing side, which, requires a lot of, working with very clunky stuff in the back end to make a pretty seamless process on the front end for patients. So we're going to go a little bit into what Photon does. but I'm very excited to have Rado on the pod. Thanks for coming.

Rado: Yeah, thanks for having me.

Rado: So I'm Rado. Most people call me Rado because the proliferation of Michaels in any office setting. We already hired our 1st, 2nd Michael, I should say. So I come from e commerce background. And then the last 10 years in tech, I spent my 1st startup was out of beta [00:02:00] works. And then a whole host of different stars, but I spent about three years at Thirty Madison. So you'll know Thirty Madison mostly, from Keeps, their kind of flagship brand for a hair loss.

Rado: And so I started, very early, like 30 employees, 10 million in revenue, and then left that, 200 million in revenue, classic high growth startup story. I saw a lot of shit, as they say. I helped launch five brands and moved to the clinical side, did a bunch of deep health integrations where I met my co-founder, the beautiful Otto Seip.

Rado: He is not here today. We did the six month SureScript integration. We try to build custom clinical tools around prescribing. We try to increase provider to patient ratio and think about efficiency a lot. And the thing we walked away from was prescriptions are awful in America. That is the main thing we're trying to solve now.

Nikhil: So maybe just like set a baseline for the people listening. I know we've sort of alluded to what Photon does, but just, I'll give the two second version, which is, doctors are basically able to send a prescription to a patient's wallet. They can then take that prescription and send it to whichever pharmacy they want. So even if one [00:03:00] pharmacy is out of stock of something, instead of needing to go to the doctor, have them represcribe it or whatever, the patient can just send the script wherever they need.

Nikhil: I'm curious can you just tell us a little bit about like, what is actually happening in the background to enable something like this?

Rado: I mean, the high level is basically we sell prescribed tools to doctors to meet the need of folks that are building custom clinical tools, right? Most traditional EHRs aren't meant for high volume growth stage startup type of patient to provider ratios where you just have one provider treating a thousand patients in a day, whether it's a message or not.

Rado: that's our business model. The patient experience came out of the fact that, Ultimately, providers are making a lot of administrative decisions, right? They're deciding what pharmacy to go to. And our general take is a doctor should never be proxying a decision for a patient.

Rado: So in our world, you get a text message. Doctor wrote you a prescription. You can open up a pharmacy selection flow. You can give us your expected ready time. You can search by hours, all things that are really painful. If you make a bad decision, when you go into a hospital or a brick and mortar clinic and [00:04:00] choose a pharmacy upfront, and then it's the wrong pharmacy, right?

Rado: And then it's stuck at that pharmacy. You got to call the doctor back yada, yada, yada. And so what's happening underneath the hood is that we've effectively separated a prescription order into two separate constructs. So in everywhere else in the United States, there's a standard that says not a regulatory standard, just a standard that was invented.

Rado: This says a pharmacy has to be a part of the prescription. We've said that's a fulfillment choice. That's not the doctor's choice. and there's a whole lot of conspiracy theories and things I can go to about why this is, but at the end of the day, it's just a bad consumer experience and so what we said is we'll separate the prescription pathway. That's for doctors. It's immutable. It's the treatment pathway based on the diagnosis code and the condition that the, patient has. . And then the fulfillment decision can be triggered by anyone, whether it's a medical operations person that wants to do a check in before they pick a pharmacy, whether it's the patient themselves, or whether you want to do a programmatically via, an API so that you can fulfill your specific finasteride at your vertically integrated pharmacy to the Keeps reference earlier.

Rado: So underneath the hood, what we're doing is we are sending a prescription to a [00:05:00] pharmacy via fax. Most of the time, about 55 percent are e prescriptions, and this was our big bet that as we gained traction, as we redefined consumer expectations, that we would get more and more pharmacy integration. So that's been proven with massive retail chains all the way down to your sort of like regional mail order pharmacies.

Rado: what we said is that fax is not a prescription, it's a notification and the pharmacist can log into a pharmacy application using the same, security credentials that you need to pick up a prescription. Fun fact, in the United States, all you need is the pharmacy, the name, and the date of birth of the patient, and you can pick up even a controlled substance, which is a famous lawsuit in California that went down. That is the level of security we're dealing with generally. And so we mimic that. If a pharmacist has that digital prescription digital e facts, I should say, they can log in, verify the patient's information, data, enter it and fill it. In the meantime, we have to make a phone call.

Rado: And so we make a phone call and we say, Hey, [00:06:00] did you get this prescription? When's it going to be ready? Do you have it in stock? Is the patient's insurance on file? All of the things we're doing normally happen by a, the patient getting to the pharmacy or be calling the pharmacy ahead of time, which is massively painful, especially for folks that are, that have multiple chronic conditions and or multiple prescriptions or folks that have kids that are sick.

Nikhil: And just to clarify for my understanding, so if you are not integrated with the pharmacy, then the way that you guys send the prescription right now is through fax or phone call. And then when you are integrated with the pharmacy, you send a direct. Basically, I send a direct e-prescription.

Rado: That's right. The clarification I would make is the prescription, the fax is, our general take on this is the fax is a notification for that prescription. It says, Hey, here's the data. You can access the digital prescription here. And that meets the broader requirements around digital prescriptions in the United States.

Rado: Either way, we make a phone call, right? Because we need to get to the [00:07:00] consumer level that we've outlined which is basically saying, Hey we're gonna. We're going to build this like we would Uber, DoorDash, Instacart, Airbnb, fake it till you make it. And we're going to build the consumer experience that should exist.

Rado: And then we're going to work backwards on the tech. Famous Steve Jobs quote. Sorry, I have to quote Steve Jobs. This is the only time I would do it, actually, to be honest.

Danielle: I wish you were wearing a black turtleneck. It would really elevate the delivery.

Rado: But you see most folks in healthcare, I will say that I see, are trying to shim in or live on top of the already existing infrastructure. I think it's a disaster.

A.I. Speaker: The following message is brought to you by out of pocket. People who have been tasked with building a call center, have many questions.

A.I. Speaker: How did I get this job? What are my options? How will this affect my loved ones? Will I ever be happy again? You need answers. That's why we made a course led by call center veterans who've been there and recovered. [00:08:00] Join nearly 100 others who have not learned the hard way. To enroll, call toll free today to reach no one or go to outofpocket. health. Hear from Stephen, a student who said, quote, "This course was hugely validating and packed in many invaluable insights that we would have otherwise had to learn the hard way. Five out of five stars." You can be Stephen. Save money. Start learning. Enroll today. For $100 off your seat, use code ANSWERS.

A.I. Speaker: That's code ANSWERS for $100 off.

Nikhil: This episode of Ops, I Did It Again is brought to you by me, because no one believes in me more than me. I'm teaching a Healthcare 101 crash course that starts soon. You can find it on the Out of Pocket website at

Nikhil: We go over all the major stakeholders in healthcare. Payers, providers, pharmacy, all those in between acronyms that you've been too scared to ask people what they [00:09:00] mean. We'll talk about how the money flows, major laws in the industry. It's the fastest way to get up to speed on U.S. Healthcare. You can use the code IBELIEVEINME to get a $100 off.

Danielle: Walk us through the life cycle of a prescription. But really what I want to know is very detailed, right? So like within 30 minutes, we have to call the pharmacy and then what happens from there? Or they get a text if they don't answer within three minutes, like I want to know your whole crazy formula behind the scenes.

Rado: Yep, okay. So a doctor writes one or more prescriptions, sends that order to the patient. Patient can select a pharmacy. They can pick an urgency time, which is something we're testing now. And then, so basically what time does the patient want to pick it up? and then they go into the pizza tracker for prescriptions, right?

Rado: So just imagine DoorDash or any other sort of modern consumer marketplace or tech enabled service business, right? And so basically a patient will get a text, go through the pharmacy flow, pick a pharmacy, give us an urgency time, and then get a series of text message updates. In the same time, [00:10:00] basically within 10 minutes of that order being created, we call the pharmacy so it enters into a queue. And it is a timer starts and we measure this based on order agent and we have an SLA that's effectively 80 percent of prescriptions are confirmed, meaning we check that the pharmacy is inventory. We get a ready time from the pharmacist, a promise time effectively, and we check if there's any insurance gaps at the pharmacy.

Rado: 80 percent are in under an hour, 70, 60 to 70 or under 30 minutes. And I would suspect 40 percent or under 15 minutes. And so our SLAs are wild. And so part of what we did with the urgency collection from the patient is say, when do you actually want to pick this up? Because right now it's first in, first out we treat every order as the same. And when orders start to age, that's when things are going very wrong. And that's generally bad for us and our customers and the patient, right? We try to limit that as well. We have a team of external agents and internal agents handling escalations and then folks that are just hammering the queue, getting through it as quickly as possible, about [00:11:00] 50 percent of orders are one touch.

Rado: This is a critical metric for us, right? Because every time we touch it, that's the same average handling time, the same average downtime. that it takes for a real order that's going to be a happy path order. And so then you have an agent spending that time on an Oregon firm. So there's lots of conditions that go into why that might be.

Rado: The big one is out of stock in an inventory issue, but most of the time medications can be ordered. We deal with a lot of pediatrics customers. And we love them and they care about us the most because we help most with the most painful issues in terms of like parent has two sick kids. We see things like parent has three sick kids.

Rado: They all have the same condition. Mom needs medication at two in the morning and is driving around Austin trying to find a pharmacy that's open. So we just cut through all of that. We will call nearby pharmacies. If a medication is out of stock, we'll find it in stock. We'll offer that to the patient and they can confirm it.

Rado: There's a whole ton of things we can do to automate and streamline those workflows internally, which we're [00:12:00] doing. But at the end of the day, it's when, we joke internally, right? It's when, The pharmacy experience is awful, but the Photon experience is great, and the responses we get from patients are pretty wild because of that, things that they don't expect from a normal pharmacy experience.

Rado: And from that point on, if an escalation happens something goes wrong internal agent can't handle it, or I should say an external agent can't handle it, it goes to an internal agent. So these are, pharmacist has a clinical question about the prescription, and we need to get in touch with the doctor.

Rado: What we do is basically triage all the pharmacy orders and say, is this a real issue? That can't be handled by anybody but a doctor, then we forward that and triage that to the clinical folks for our customers. If it's not, we try to handle it. So that's out of stock. That's a question about the patient's name yada, yada, yada.

Nikhil: I think maybe, just like switch gears a little bit, because one of the things that I think was interesting, we were prepping and talking about this a little bit is just to show people kind of the scale of what we're talking about was when one of your customers out had an outage with their e-prescribing and you had to make [00:13:00] thousands, I think tens of thousands maybe of calls. I'm just, can you just like, tell us a little bit about that story and what ended up happening and how you got out of that?

Rado: Yeah. For historical context, when we first started this, we were working in Slack.

Rado: With Slack integrations that would pipe orders into Slack, and they would open them into a really not a queue as much as Slack was the queue effectively, but just an order of you in retool. And then we were everybody on the team very much like Paul Graham things do things that don't scale every single person on the team, whether it's an engineer, designer, founder was taking calls and understanding what was going on at the pharmacy. And so we were in this transitional phase where we knew with the upcoming scale that we had to transition away from this and build a true kind of queue and optimize for scale. And so right after that got done, and we were testing it, one of our customers who is a pediatrics based company had an outage with their prescription provider. And the benefit of being in healthcare is it's all IT companies. No, no one's [00:14:00] outside of digital health and outside of like true innovative startups. Everybody else is like an IT company no one's product oriented. No one's tech oriented. And they're just not, they're not great in terms of service level. So. We felt this at Thirty Madison, right? We had outages all the time. We would not be able to send prescriptions for a day. That would be like 50,000 orders. We'd had to, we had to delay it was insane. This customer had an outage. They asked us to accelerate our rollout plan that we had started. And I was about to go upstate to the Catskills for my one vacation this year. And I bounced and we made a founder level decision to that. We were going to move forward and it would be okay and our volume projections and capacity planning were, let's say. Real bad. So I in the Valley of Hunter mountain, where just get a sliver of internet service would come through or cell service would come through. would just get 18 Slack messages. So I saw this happening from afar and the first weekend was rough.

Rado: But we learned more than we had learned thus far. I think as product people and engineers and ops [00:15:00] folks, I think we all try to theorize and plan and try to de risk as much as possible, but there's nothing like. learning from doing when you increase your volume by five X. And so we felt that pain.

Rado: And we learned more in that one week than we had learned in probably two months. And that's partially why we're here today.

Nikhil: That's awesome. From there, today now you guys have obviously handled a much larger volume of everything, right? Can you give us like a sense in the last month, like how many phone calls, how many faxes, like what are some of the, how many messages are you sending?

Nikhil: What are some of the stats around like the volume you guys are doing now up from, obviously that one, Slack obviously is not meant for the volume you're doing today.

Rado: Yeah. I felt Slack was breaking down, which was a sight to see like literally from a technology standpoint, the product was crashing and so I think we really tested the boundaries, but right now we're doing 20,000 orders a month, about 15,000 faxes, 10,000 legacy ERX, and about 26,000 phone calls. For that 20,000 orders, we get about [00:16:00] three- no, I'm sorry, 3,500 text messages. That's inbound and outbound is roughly the same.

Rado: About 30% of those are thank you's. So I don't know when the last time is that you responded to a very clearly automated message, but our patients do a lot, which is awesome. So the challenge is now, once again, classic, growing startup is we'll 5x that in the next two months. And so making 100,000 phone calls, 200,000 phone calls. That changes things for sure.

Nikhil: One thing I found myself doing by accident is sometimes I'll do the like Apple, like thumbs up reaction to an automated message that comes in. Like it's an SMS, right? So I'm like, Oh no, did I just like send an SMS back with a thumbs up thing?

Rado: We get those, we get liked. Yeah.

Rado: Liked in their own message quoted to us. And we're like, okay, cool.

Nikhil: One of the things we had asked you a little bit again, beforehand is what are some of the first things you automated when you began this whole process? It sounds like you have some pretty janky beginnings of probably Slack [00:17:00] bots or, Zapier piping into Slack and all this kind of stuff but, can you tell us like what first thing you were like, "Okay, we need to definitely automate this?"

Rado: Yeah. I mean, So now, that this is quarter our strategy in general, I've broken this up into two buckets, right? So we think about it in terms of automation, which is just wholesale or placing human workflows. And then optimization, which is making humans faster for automation.

Rado: We technically really haven't done anything in terms of true replacement of workflows. We're exploring a ton and that looks like trying to get an AI bot, wait on hold to wait on a hold and then transfer to a human in terms of like optimization. The first thing we did was get out of Slack. And so we built a.

Rado: Q that we now call the push Q. There's a big distinction here. This is a paradigm I've seen at every telehealth company in terms of product folks that I've talked to and how they handle their clinical workflows is that you have a Q where orders could be claimed. That would be a poll. So I'm an agent and I have to, make the decision.

Rado: What's the next order. And then the other side of that is you just push an order to someone [00:18:00] and then they take it. And so we just released that and I can talk to the results there. But the big thing for us is we have these KPIs that we look to and we have handle time, which is. It's difficult.

Rado: Handle time is basically how long it takes to work in order. What's an agent doing? Are they answering a text message? Are they reviewing the context of that order? Are they making a phone call? Are they waiting on hold? Because that is an uncontrollable variable for us. And then the other piece is downtime.

Rado: And so downtime is like. what's happening? An agent is not claiming an order. And so we have reversed the we've first pushed out the push queue and saw the poll queue effectively. And we saw a pretty massive increase in our capacity. We saw a lot of downtime and we don't know what's happening, right?

Rado: We're not sitting with agents. We don't understand what they're doing necessarily. And so now we've started to say the real thing to automate here is claim an order. So that's arguably automation, right? We basically said you are no longer the human making a decision. We're going to do this for you and we're going to do it programmatically or algorithmically, and that will help to reduce [00:19:00] downtime.

Rado: So we just released that downtime in our peak time went from four and a half minutes to a minute and a half in a week.

Danielle: Wow. It's interesting, your definitions of handle time and downtime actually really speak to how producty you are, because in the call center world, handle time and downtime would just be for a single interaction.

Rado: Right.

Danielle: It wouldn't be for a cumulative order, but actually, that makes a lot of sense that you consider handle time is a full order and makes your life a lot easier as you're trying to figure out where there's room for optimization and automation to your framework. I'm going to add that to my next slide on the course.

Rado: Give you real definitions, yeah. We're just parodying it over here, right? We read like one blog post. But the way I think about it, the way I calculate it actually is like, the end to end completion time is downtime and handle time. And that's true. How many times can you do that per hour? And then you look at, we look at touches per order and we say, okay, now each order has to get touched twice to be completed.

Rado: And on average, overall downtime plus handle time takes [00:20:00] 20 minutes. That means you can really only touch three orders and half of those aren't going to get completed. So you start to compound these things and you look at it and you're like, what's happening. And so it does become glaringly obvious about where the opportunities are.

Rado: And the products I have for me is like, what do I think is actually in my control? What can I really do? And so a lot of it's incremental, but some of it is big swings. And so that's the difference between saying, okay I can optimize how long it takes an agent to send a message template, but I can wholesale replace them, picking an order and might shave off three minutes per order of completion time overall.

Danielle: What happened internally where you decided "Now is the time for us to move from this pull system into an automated system?"

Rado: Part of it is what I was just talking about in terms of 20,000 orders to 100,000 orders.

Rado: It changes, it changes the problems at hand from a unit economic standpoint to the business problems, to also our ability to just maintain quality.

Nikhil: We're sort of talking about what generative AI kind of means for companies like yours, right?

Nikhil: Like, Have you tested out things [00:21:00] like, Hey, can we do actually even like robo outbound calling, or any of these other, like more sophisticated. Generally, I kind of use cases for something like talking on the phone or even sending, interacting with the real world in some capacity.

Rado: Yeah, for us, as I mentioned, I think it's a race towards both integrations and replacing a lot of core workflows to really get us the consumer experience that we're trying to give as in here's the expectation when this is going to be ready. Here's it's out of stock. What are the exceptions?

Rado: There's a lot we can automate across the board. And so we're going to do both. We've evaluated probably 20 different vendors 10 to 20 different vendors and all of them, one, if you don't have ABI docs in your site, red flag, we only have found one that seems viable. But what's actually shifted is a general belief in viability.

Rado: So we hear from investors now that yes, enough of a timescale, this will, we believe that, so that's mostly what we care about because it means what we need [00:22:00] is time as any sort of seed to a stage startup and we need support and believe that we can do this over time and we need integration. So we'll get that for sure.

Rado: And so now it's kind of shifted the conversation to not if but when. And the tests that we've run so far definitely have some room to grow, so we ran a bunch of tests to pharmacies and you have whatever sort of middleware transformer that actually accidentally is, converting the text to Mandarin.

Rado: So that's not going to work. You have a lot of drop rates in terms of just that basic problem I mentioned around, can you get AI to wait on hold and tell me when a human is actually talking versus. Repeated voice saying CBS, if you'd like to check out your order online, go here. And so that problem is really hard and we haven't seen a ton of success in terms of hit rate.

Rado: I can't give you a number, but it's in previewing it. It's enough that it's like, well, this probably isn't going to work at scale yet, but it does work. Sometimes [00:23:00] our benefit is pharmacists. Poor pharmacists have such tolerance for like inefficiencies on a phone call that they'll just wait. I've done a couple of tests myself with bland AI, which is one of these companies and done specifically the workflow of transferring to a human, which is a huge increase for us.

Rado: It means probably 40% of the call is waiting on hold and then transferring to a human. It means no one sits on hold. You're just constantly queuing up orders that are ready to go. I had three separate pharmacists wait to get to my voicemail before hanging up, meaning they talked to a bot, they waited for a phone to ring, they waited for a voicemail to say, hey, this is Michael Rado, and then they hung up when the message connected.

Rado: The tolerance is pretty high, but we also want to be careful, right? We're trying to not destroy the trust of pharmacists that are one of our three core users, and so we will phase it accordingly. And like I said, the first thing we'll try to do is get a bot to wait on hold, and then we'll try to do the happy path, right?

Rado: The happy path for us is basic structured data insurance. Did you get this? When will it [00:24:00] be ready? Is the patient's insurance on file? And so that's collecting three data points and then piping it into a separate workflow if there's an exception. So, we have pretty high confidence that in the next six months we'll get to a working prototype that is viable.

Rado: Maybe sooner, but we'll see.

Danielle: So you talked about timeline of this project being six months. That got me curious a bit more on how you're approaching product. More generally, can you share a little bit about how you prioritize which products, like what product builds, builds to work on, excuse me, is some frameworks.

Danielle: I don't know, Paul gramming this, like what's the deal.

Rado: I am ruthlessly idealistic about product and how it works with other functions. So I would say. I like to start with the definition of product, which is basically something that creates value for a business and a consumer. When those things get out of tow, right?

Rado: When they become imbalanced, you have shitty businesses or you have great unviable businesses. And so we're constantly trying to toe that line of how do we create a ton of value for our consumers and our customers and their [00:25:00] consumers? And then how do we trade it? Create value for ourselves. And so much of this is hypothesis driven.

Rado: We look at data, we start Melissa Perry's great on this, the product Cata, right? We start with what is the state of the world today? And we look at our efficiency metrics. We look at our conversion rates. We look at our customer growth and we say, what is our biggest problem? And then we take a couple of bets.

Rado: And so we're a team of 15 mostly engineers but not all builders. And. We say, Okay, this quarter, this next two quarters, automation and optimization are the most important strategies we have. We want to create a better experience around optimization. And I haven't really said this yet, but part of this is being able to provide greater granularity for customers.

Rado: What happened to this? Think FedEx. Your order was delayed because the address on file isn't correct. So how can we really create this end to end consumer experience where you have transparency to every step that happens? And so that's value for both of us. And then automation is reducing, our overall costs, but it's also saying, Hey, we want to mop these orders up as quickly as possible.

Rado: Patients have [00:26:00] urgency. They need to know what's going on. They don't get this from anywhere else. They're generally waiting on hold for an hour to get this elsewhere. And so we need to do this as quickly as possible. Thank you. And so that's the core strategy. Then we define two North Star metrics that are lagging and then we leave it up to the squads.

Rado: So we have two squads, one map to each North Star. They get a key metric that's a lagging metric. We work together to map out proxy metrics. So, for optimization, right, it might be the time it takes. It might be handling average handling time. How can you reduce this? If a PA is required instead of picking a message template, can we click one button that picks a message template and then updates the provider on their side and updates the patient?

Rado: And you just shaved off a minute of our handling. Great. And so very much outcomes focused. Our roadmaps reflect that it's now next later. We are hypothesis driven, so I try to get folks to get to an estimation question. How many hours do you think you can save for 50,000 orders? If we take, you know, so that example of, of the push queue, if [00:27:00] downtime is four minutes, now it's two minutes, we just cut downtime in half.

Rado: What percentage of the total working time for an order is that? Did we reduce the amount of time it takes to complete an order and to end by 20%? Is that the most viable hypothesis you have, then run with it. The reality is, if you get into a world where you've got 15 variables that are inputted into a spreadsheet and you're trying to rank order based on impact, effort and potential revenue,

Rado: you're basically in a world where you don't understand what the biggest problems are for your business. Take a step back and look at the data and understand where the biggest opportunities are. And you probably have a top down architecture or an organization that's pushing you to do things

Rado: and you're trying to meet some goal that you don't even understand. Here's the biggest problem, we understand that optimization automation will get us to the next phase of growth. Now it's up to the team. And so that's where we see empowered folks taking shots.

Danielle: It's up to the squads, actually.

Rado: Exactly, exactly. And so you have 10 bets in mind. You have probably going to take five. It doesn't matter if one or two or three is more important than the other. You're going to get five done. So just take the ones you [00:28:00] have the most confidence in.

Danielle: That's awesome.

Nikhil: My astronomy is rusty, but two north stars feels like uh, something that doesn't really exist.

Danielle: What is the big dipper.

Rado: Fair, fair, fair.

Rado: I mean, for us, the north star is a single metric for sure.

Rado: Nikhil,

Danielle: you're such a troll, ! . Such a troll.

Rado: Come on!

Nikhil: I'm just messing. I'm just messing!

Danielle: One quick thought on that is, there's like this saying in design thinking that if you told somebody, can you design a better spoon, they'll design like three permutations of a spoon, but if you told them, can you design a new way to eat soup, they might not even come up with a spoon.

Danielle: And so in your example, that's really what stood out to me. You're like, how do I reduce handling time? And so that, I think that's what empowers people. Cause they're like, Oh, it could be a spoon or it could be this, or it could be that, and suddenly you create like really creative ideas. So I'm really excited to see what your squads are going to think up. I hope you tell us!

Rado: It's fun for sure and I think honestly it [00:29:00] leads to folks that are generally more driven and happy and feel the weight of their decisions. Like, the worst thing you can do is just build features and ship them and have no idea what. Then your measurement become, "How many features did you ship?"

Rado: Which is classic. "How many story points do you got on your belt today?"

Nikhil: The third North Star, obviously.

Nikhil: Im just kidding!

Danielle: We got to wrap this up, you're getting unhinged.

Nikhil: All right. All right..

Nikhil: So, to wrap it up, at the end of each podcast episode, we ask our guests to basically give three actionable experiments that people can try in their own orgs that you've seen work or things you've tested out, so, or are there some things that, for listeners that you can leave them with? Takeaways?

Rado: Yeah. Yeah. So when I look back on how this thing worked and how we got this far, the true tests to your point is in five years where we're at. So I think about early on when we. Had a bunch of hypotheses.

Rado: We had a bunch of signals, but we didn't really have a ton of users to [00:30:00] roll out a feature and look at the usage or look at engagement. We came, we had this early idea to let patients pick their pharmacies. And our bet was like, doctors would be jazzed about this. And so that's how the patient experience came about.

Rado: We shipped a prototype, we built it super scrappy, didn't connect to any of the system. And it looked pretty real and we went out and sold it and we landed our first five customers. The misconception is it's not the genius in the group coming up with the next best idea.

Rado: It's a bunch of ideas and then going out there and saying, "Hey, would you actually pay for this?" So that's stolen straight from Steve Blank.

Danielle: Nikhil, is that when you invested, when it was, like, connected to nothing. And you're like, "That works. And that's cool."

Rado: Nikhil was earlier than this.

Nikhil: Mine was even worse. Otto barely had a deck together. That was just like a thought.

Rado: Yeah, it was just a piece of paper and like a great gold chain. Uh, so the next, the next one is like, do it with a spreadsheet. The number of companies that we talked to in another number of organizations that I've been a part of [00:31:00] were.

Rado: Someone likely has an MBA, is pitching some idea, enforcing that the real thing has to be built and that you can't sacrifice anything is absolutely ridiculous. Amazon, when it started its small business program, you literally had to type in your SKUs, description, product names into a spreadsheet and they would scrape it once a week.

Rado: And that's how they started that business. So if Amazon can do scrappy experimental testing to validate an idea, anybody can do it. De risk it before you invest six months of time and a lot of smart engineers time. And then the last one I was going to give you like another product thing, but I actually think like do weird shit with your team.

Rado: So the first offsite we had, onsite we had, we flew in the whole team and we had one of Otto's friend Taylor Myers, Tay, who now runs a little bar called number nine. Come in and do basically a workshop. And the first exercise we did literally no one had ever met anybody from the extended team yet.

Rado: We sat cross legged across from each other, stared into each other's eyes, put our [00:32:00] palms on our forearms. And tried to squeeze as hard as we could while, while Tay read a general kind of like poem to us sounds a little hippie dippie, but I talk about that with the team and the team brings it up nearly every time we have an onsite and says, can we do generally says, can we do something like that again?

Rado: The feeling of like closeness getting weird. And my favorite investors long journey and they get real weird on their onsites and offsites to make people uncomfortable. And get people out of their skin. It's a professional environment. You gotta have some boundaries, but like, try to do things that shake people up and force people to like open up a little bit.

Rado: So that's a big ups to Taylor and also Otto.

Nikhil: I will say that like, when I was in college, I was on a dance team and our best choreography ideas would always come from, When we're in the weirdest environments by far, because people are so uninhibited at that point that obviously alcohol helped a little bit.

Nikhil: But people are so uninhibited at that point that they'll just pitch the ideas that are like half jokes, but every [00:33:00] good idea always starts as like the strange half joke where you're like, ha ha, but what if, and then, but you, but to create that environment is very difficult. And so you need a little bit.

Nikhil: Of weirdness to kind of break that ice and like, you know, make it comfortable for everyone to like pitch what they think might be a weird orthogonal solution to a problem. So it, that resonates for sure.

Rado: Yeah, I think the challenging part and what we've done by hiring people we know is like, how do you get people to trust each other and make decisions, take bets on each other and get past that initial.

Rado: Social intelligence that's required to really collaborate at a heightened scale and getting weird helps and alcohol helps. By the way, that was no alcohol, so.

Nikhil: Damn, just using our brains. I respect it.

Nikhil: All right. Rado, I know we're at time. We appreciate you coming, sharing some of your learnings along the way.

Nikhil: Very excited to see what the next 100,000 scripts look like and what breaks and what you end up fixing there, but we hope people a lot from this one. I did for sure.

Danielle: I did too. This was [00:34:00] awesome.

Rado: Great. This was great. Appreciate you both.

Let's Keep In Touch

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
search icon