Recast Incrementality Platform Overview: Plan, experiment, validate & optimize marketing performance

Foreign I'm Mike, one of the co founders and co CEOs of Recast. I am really excited to be giving you a walkthrough of the Recast platform. Today we're going to hit on the unique insights that Recast can provide, all of the really interesting reporting features that we have. We're also going to talk about the planning tools that you can use for building your quarterly or annual operating plan, how you track your pacing against your goals, and how you evaluate and valid the Recast model. So we're going to jump in, we're going to start on the initial homepage of the Recast dashboard and what you're seeing on the screen is exactly what any user that is signed into Recast will be able to see. We're looking at demo data for a brand. You can imagine something like a fast moving consumer good brand. It sells online through their direct to consumer website. They also distribute through third party retail channels like Amazon, Walmart and Target. First things first, we can see the KPI that we're looking looking at here is for D2C subscriptions. But there's a variety of different KPIs that we can look at. So if we want to look at Walmart revenue as our main KPI, we can select that. If we want to look at Amazon revenue as our main KPI, we can look at this. Things to highlight here is that Recast is incredibly flexible. We can look at revenue KPIs, we can look at profit KPIs, we can look at conversion or acquisition related KPIs. Whatever the KPI is for your business, Recast is going to be flexible enough to handle it. You can imagine looking through different distribution channels or at different geographies of our different customers break down their KPIs in different ways and then when they're reporting on them, they can select between different KPIs and you can look at aggregated views. So total revenue altogether, adding up a bunch of geographies. Whatever it is, it's always available for you here at Recast. So coming back, this is a snapshot overview. What we see here is a summary of what's going on with this business. We spent $210,000. We had this amount of our paid CPA. So that's the cost per acquisition driven by our paid channels over the last seven days. And then we see a bunch of interesting summary statistics. There was 2.1 4,000 conversions that came from baseline, 1.7 4,000 that were driven by our paid media. That implies a blended CPA of $57, a paid CBA CPA of $124, et cetera. We can also switch between different time periods. So if we're more interested in looking at the last 30 days instead of the last seven days, we can select this dropdown. If we want to see the last 90 days or the last 365 days, we can see all of this as well. All is very flexible. If we want to see some custom date range that's also available, we can generate custom reports for any time period that we're interested in looking at. So let's look at the last 30 days for this business. And again, we can see they spent almost a million dollars. They had around $120 paid CPA. This is the summary statistics and below this we can see how all of the different marketing channels are performing during this time period that we're looking at.

Right.

And one important thing to keep in mind is that Recast is designed to be updated on a weekly basis or on an on demand basis. And so here we're going to be looking at a snapshot of the last 30 days which is from December 21st back to November 22nd. And next week when the model updates, it's going to get new data, it's going to re estimate all of the parameters. We'll be able to see how are all of our trammels performing during that next 30 day window. The idea here is that the frequent refreshing does a couple of things for us. Keeps the data really actionable because we're reflecting the most recent performance. So if one of your channels has started to increase in performance, say because you released a new creative campaign that's performing really well, Recast is going to respond to that. Or if a channel starts performing less well, maybe because the algorithm has changed and the targeting that we were using before is no longer working as well, Recast is going to respond to that as well. So the idea behind Recast is that it's a very dynamic platform, constantly updating, constantly seeing how things are changing. And that allows you to be much more responsive. So instead of looking at stale, out of date data from, you know, six months ago, this is constantly updating so that you can take advantage of changing trends across your different marketing channels. So let's take a look at some of the reports here. We can see the marketing effectiveness report. So for all of our different marketing channels, meta prospecting versus linear TV versus meta retargeting, what's the incremental cost per acquisition for this channel? And so we can see the incremental CPA for meta prospecting is $191. The incremental CPA for linear TV is $110. The incremental CPA for meta retargeting is $85.

Right.

So fairly easy to understand. We also have an overall here that shows the overall total budget and then the overall paid performance and then the spin column is just showing us how much did we invest during this time period. So again, over the last 30 days in each of these different marketing channels, you'll see here that we have both blue bars and red bars. The blue bar is the average cpa, the average cost per acquisition for the total amount invested, and the red bar is the marginal cpa. That is to say, what was the cost per acquisition of the last customer that we acquired. The idea here is that in some channels we are probably experiencing diminishing marginal returns where as we've scaled up spend in the channel, each additional customer we're acquiring is costing more and more. We can really see this in meta prospecting where the average CPA is $191, but the marginal CPA is $335. So each of those additional customers that we're acquiring. Right. To get a one more customer in meta prospecting, we're going to have to pay over $340 for it. So even though the channel performs pretty well as a whole, we've saturated this channel by expanding it as much as we possibly can. And in order to have the channel continue to perform, we need to either find some new way to get customers more efficiently or scale back spend. So here again we can see apples to apples basis digital channels like meta prospecting, traditional channels like linear tv, non brand search. We have mailers here, podcasts. So this includes both online and offline channels. And so like other media mix models, this is treating all of these of these channels on an apples to apples basis. Which is a really great reason to use a platform like Recast where we can see all of the performance of all of our channels measured on an apples to apples basis in one place. Carrying on below this Recast will highlight some interesting information for you. We can see which channels are relatively over performing, non brand search, influencers, podcasts, meta retargeting, linear TV and mailers. These are all places where we could consider investing more and which channels are underperforming. So meta prospecting and branded search.

Right.

We could take some dollars out of these channels, reallocate them to the green channels over here and overall get a more Performant marketing budget. And finally at the bottom here, this is just another way of viewing the data. We can see, okay, what was the total number of subscriptions that we acquired during this period, how much came from the baseline or were acquired organically, how much came from spikes. And so this is things like a promotional event. So Black Friday, Cyber Monday, whatever it might be, how much came from meta prospecting, how much came from linear tv, how much came from non branded search, et cetera. For all of the different marketing channels, one thing to highlight is that these names that are showing up here, these are all totally configurable, right? The Recast statistical model doesn't know anything about Meta underscore prospecting. This is just the name that the brand had set up in their marketing data warehouse. So however you label channels, whatever channels you have, you might have multiple marketing channels that I've never heard of before. That's all totally fine. Recast is totally flexible to work for your business. If you have data scientists or marketing analysts on the team, there's a bunch of information here that you can download, put into Excel, share with other models. All of this is fully available for you. So there's a lot going on in of things that you can look at in the reporting here. I'm just going to highlight a couple of these things that I think are really interesting and in particular special about the research. So I just clicked on the channel level deep dive where we can see deep dive information about all of the different marketing channels that we're currently operating in. And what you see here is you see a lot of time series curves here, right? And so this is the CPA for meta prospecting and how it's changed over time. This is the CPA for influencers and how it's changed over time. These little red dots are the spend investment in this channel also over time. And what we can see here is that the CPA for influencers has changed a lot over time. We currently have a CPA of around $16. It's been as low as $12 and as high as $20 about a year ago. And so what we can see is that the performance of this channel has really evolved, right? Sometime in around August we had a big drop in the CPA potentially related to product changes that we made, potentially related to how we manage this program. Maybe we decided to really focus on micro influencers instead of macro influencers, whatever it might be. The idea is that we can see see how the performance of this channel has evolved and when we're making Predictions into the future. We want to take into account the evolution of this. We don't just want to say, oh, we think that in the future the CPA will be whatever the average of these numbers is, right? Instead we want to say, okay, well, maybe there's seasonality in the CPA for this channel, right? Maybe it goes up in the summer and down in the winter or whatever it may be. Maybe we're on an upward trend and we expect that that will continue. Whatever's happening, we want to make sure that when we're making projections in the future that we're taking this into account and all of this is built into Recast. And Recast will just handle all of this for you when it comes to making your plans and evaluating your progress against those plans. Couple of other interesting things to look at is we can see the baseline or the organic performance of this business and how it's changed over time aggregated at a monthly, daily or weekly level. The daily view here is a little interesting because we can see that there's day of week seasonality and we can see a little bit better how the trend of this has changed over time. Right? So the organic part of the business has really increased over the last six months. Maybe there's some amount of seasonality here in this business as well. That's potentially interesting. One thing that I want to highlight is the Spikes view. And the Spikes view is what allows us to look at the performance of certain holidays. And so I'm going to, we're going to look at Black Friday start. We're going to look at the most recent Black Friday holiday. And what we can see here is that look, there's a big spike in sales associated with Black Friday. That makes a lot of sense. A lot of you who are in any sort of fast moving consumer good have probably seen this pattern before. But what's interesting about Recast is that recast not only estimates, okay, what was the lift associated with the day of Black Friday, but also what were the distortionary effects of this sale? And what we can see is that leading up to Black Friday we actually had a reduction in sales. And this is very common, right? And this probably matches your behavior right before Black Friday. You're probably not ordering a lot of things online. If you see an ad for something you like, you probably think, oh well actually I'm gonna wait another week, see if there's a sale on Black Friday and if there is, then I'll buy that. Additionally, we can also see that there's a little bit of this happening where people are actually pulling forward purchases and that they were going to make anyway.

Right.

Again, this is often a thing in fast moving consumer goods where people say, oh, there's a sale on Black Friday. I know that I love this product. Instead of buying, you know, this product once every three months over the next year, I'm just going to make all of the purchases today, take advantage of the discount and then, you know, use up my stockpile over the next year. And so you can have this like pull forward effect of running a sale. All, all of these distortionary effects are actually really important to take into account when you're estimating the total effect of the promotion. And so Recast will show you what this shape is, show you the distortionary effects and actually estimate what's the total incremental effect.

Right.

And the part of the incremental effect could be negative. And so this is really, really important for you to estimate as you're thinking about, okay, which sales do we want to run, when do we want to run them, what do we here? Black Friday is largely positive for this particular brand, but you can definitely imagine situations where it will be negative. And this is a thing that we've seen a number of times. Last thing that I want to show here are the context variables. And context variables can be really important depending on your brand, your company, how it works, what are the things that really impact your business. So for this brand there's two context variables. One is brand awareness and one is price. And so what we're estimating here is what are the effects of brand awareness and price on the business, but not in addition to marketing effect, but actually interacting with marketing effects. And what we see below is, okay, how has brand awareness changed over time? So this brand is measuring brand awareness on a monthly basis. We can see how this is evolving and then we can see how is this affecting both the baseline sales, which here actually not very much, but marketing performance as well. And what we can see is that as brand awareness has changed over time, it has increased our marketing effectiveness.

Right?

And so this allows us to estimate, okay, what is a 5 percentage point increase in brand awareness worth to the business? And what this is saying is that for every five percentage points that brand awareness increases, marketing effectiveness actually increases 25%. So increasing brand awareness makes everything else more effective. Price has a similar but negative relationship, right? As we reduce price, this increases our sales. And so we can estimate, hey, what's the ongoing relationship here? If we increase price 10%, it reduces our marketing effectiveness and then backwards, that is, if we reduce price 10, it increases our marketing effectiveness. So all of this is available here. And again, very useful for thinking about, okay, what do we want to do next year? How are we going to plan in order to take advantage of these effects? The last thing that I want to show before we wrap up on the insights page here is the experiments view. So recast is fully compatible with every different type of lift, testing and experimentation platform that's out there. In order to calibrate your MMM results with experiments, all you need to do is run the experiment, load that information into recast. Where that information is, hey, what dates was the experiment run? What was the incremental estimated lift from the experiment in terms of either cost per acquisition or return on investment? And then what was the standard error or the uncertainty associated with this experiment? Once you include these experiments, these will automatically be used to calibrate your recast results. So you can actually use experiments to calibrate the recast media mix model. And really advanced companies actually get into building and incremental mentality system where all in the recast platform, they're going to plan, experiment, validate and optimize. And so that is to say they're going to plan for, you know, what experiments do we want to run? How do we want to adjust our marketing budget? They're going to execute those experiments, they're going to validate, see what worked. Did we see the lift or not? They're going to plug the results from their experiments into the recast platform. They're going to optimize their marketing budget and then go back and plan again. What are we going to do next month, next quarter, and then just continuously repeating that process and using that to really optimize their whole, continuously optimize their whole marketing mix. So let's talk a little bit about how planning actually works. And what I'm going to do is I'm going to click on the optimizer tab. And the optimizer tab allows us to create an optimal budget based on the results from the marketing mix model. And there are a number of different ways that we can think about building an optimal marketing budget. The first thing that we need to do is we need to set an objective. And there's a couple of different objective options here. The first is reach a target outcome. So this might be something like, okay, you know, next quarter we want to acquire 50,000 or 500,000, you know, new users or again, whatever the KPI is. Or we can say we want to maximize profit, right? So we Assume some value per acquisition. And we say, hey, recast. Find the profit maximizing point of investment for every different marketing channel. And then last, we have a reach, a target efficiency. So we might say, okay, what the business wants us to do as a marketing team is acquire as many customers as we can, as long as the blended CPA is less than or equal to some number. So let's say here it's blended CPA of less than or equal to 50. And there's a couple of different things that we can do here. And so maybe we have an overall spin constraint and we say, okay, we want to hit this goal and we want to spend exactly $10 million. Or maybe we don't have that and there's no overall spend constraint and instead we just have by channel constraints. And so here we're going to say, okay, we're going to do this through the end of February. We're going to say we have some overall constraint for most channels, but then some channels are special. Linear TV. So here we're saying we can spend between zero and $10 million on these channels. Linear TV, we've committed to spending $500,000 with some agency. And they've also told us we couldn't deploy more than $2.5 million in this time period. Mailers may be right or special. Maybe we can only send a mailer every 45 days. And I don't know, maybe we could only deploy $250,000 worth of mailers during the time period that we're looking, looking at. And I don't know, maybe podcast also has some, you know, constraint on it. We know we couldn't deploy more than $50,000. Whatever these numbers here are that are relevant to your business, that's just what you're going to plug in. And then what we do is. Let me move my face here. We're going to come down here to the bottom. We can give this optimization a name. So we can call this through February with podcast limit, right, or whatever the relevant name here is, and then hit this Run Optimizer button. And what this is going to do is it's actually going to run a whole bunch of different simulations and try to identify the optimal budget that hits this goal within these constraints. Taking advantage of everything that the MMM has learned and all of the experiments and lift tests that you've run in the past. This takes about 30 to 90 seconds to run. So I'm going to pop back in here once this is completed. So this optimization has completed. You can see we had the Goal of getting to a 50 CPA. We can see that we came in under that we're going to generate this many subscriptions. And here we can see the optimal breakdown of investment. Linear TV influencers, non brand search, brand search, Meta retargeting mailers. You notice we've pulled basically all of our budget out of meta prospecting, which is really interesting. Here we can see, okay, what's the schedule? So it's basically saying, look, get a mailer out as soon as we can. Send another mailer on February 5th. This is the daily plan for how to invest on terms of influencers, which are spiky, linear TV, which is also going to be spiky, podcast, etc. And then here we can see, okay, if we follow this plan, this daily budget with all of these different investment levels, this is the forecast for the number of conversions that we're going to get on every day. And this is the forecast for the cumulative number of conversions we're going to get by through the end of February. And so here we can see that not only is there a midpoint estimate, but there's also an uncertainty range. And this is actually really, really important for doing forecasts. You want your forecast to give you back the true uncertainty. And so when you're using this to make goals, right, maybe you don't want to use the midpoint and you want to say, actually our goal or what we're going to report out to our investors or our board is going to be at the bottom of this, right? A number that we're very sure that we can hit. And so the rest of this is upside. If we come in high, we're going to blow the goal out of the water. You might think about different ways of deciding, hey, where do you want to be on this spectrum? What is our chance of hitting the goal if we follow the plan and then use that to inform, hey, what are we going to communicate to leadership? What are we actually going to lock into our plan? And how do we want to think about high case, low case, mid case, depending on exactly how you operate your business. So this is all reported out to you and of course you can download all of this data. So here we have the optimized daily budget. We can download this to Excel, maybe plug this into another spreadsheet that the finance team uses for operational planning, the optimized outcome. So the number of conversions by day, cumulative. All of this information is available here for you so that you can plug this into your other systems. Okay, so now we've run our budget optimization. And we have the optimal budget and the plan that we're going to follow. Now what we want to do is we want to tell Recast about the plan so that it can automatically track our pacing against it. And what I've done here is I'm creating a goal. And so this goal is for Q4 + Q1. So the six months between October and March, we here have a goal of let's actually do this. 120,000 DTC subscriptions. We've uploaded a budget, right, the budget that we created from an optimization or that we generated in some other way. And then what we're going to do here, we're just going to save this goal and what Recast is going to do. And here we can see on this page, of course, all of the different goals that we've created is what Recast is going to do is it is going to track our pacing against this goal. And this goal is being processed and in about 30 to 60 seconds we will actually be able to to see what is the forecast with the plan, what is our chance of hitting this goal? And we'll see some recommendations that Recast is going to make for us. So I'm just going to pause the video real quick while we wait for this to finish loading. Okay, cool. So here we've got the goal. Goal is to get to 120k subscriptions. We can see that we are forecast to come in just a little shy of this with our current plan. Assuming that we follow the plan exactly, we have a 12% chance of hitting the goal. And here we can see how much have we spent so far? Far, how much are we Planning to spend D2C subscriptions so far? How much are we forecasted to get? And then what's the cost per the overall cost per acquisition, both so far and then by the end of the period and here below we can see what is our pacing against this goal. And we can see, okay, what are the actuals to date, which is the orange line, what is the straight line goal, which is this blue line. Let me zoom in here. See, we're pacing against it. And then what's the forecast from today forward? And here we can see that the goal is within the forecast range.

Right.

So there's definitely a chance that we do get this goal. It's at the top of the forecast range. And so, you know, we might not hit it. And this is going to be a thing that we're going to want to pay attention to. And the really nice thing about Recast is that every week when Recast updates, it's going to refresh this. So we're going to get new data, we're going to get another week of actuals, we're going to re estimate the parameters, see how all of our different marketing channels are performing. And then this goal will reforecast, all of these numbers will update and next week maybe this will say, oh, actually we're only 6k off and now there's an 18% chance of hitting this goal or whatever the number is. And anyone that's subscribed to this goal will get an email updating them on, hey, what's our status? How are we pacing? Are we on track to hit the goal or not? And the reason why this is valuable is because it allows you to get out in front of potential issues, right? Maybe you see this and you say, oh, actually at this point, now's maybe the time to go talk to the CEO or the board and let them know, hey, we're not quite on pace to hit our goal for the quarter. Or maybe we need to go find additional budget. Be like we could get back on pace if we can just make some additional investment or run a different promotion. All kinds of different things you can do to respond to this. At the bottom here, Recast is actually going to produce some recommendations for how we can alter our budget to get us closer to hitting the goal. These will populate in another about 20 or 30 seconds. So I'll just wait for those to come in and then we'll jump back into the video. Okay, and here the recommendations have populated so we can take a look at them. So this is the same goal that we were just looking at. And now we can see, okay, Recast is going to look at the plan, the target and try to figure out is there a way that we can get there more efficient. And so what this is saying is, look, we can make these optimizations and there's conservative view of this which makes less changes. There's a an aggressive view of this which makes more changes. And this is what it's recommending. It's saying, look, pull all of that money out of meta prospecting, reallocate it to linear tv, non brand search, meta retargeting, podcast, et cetera. If you do this, we could get an additional 4,000 subscribers using the same budget. So keeping the overall dollars flat, that would reduce our expected CPA blended CPA from 55 to $51 and increase our probability of hitting the goal from 12 to 29%. All just by making some of these small budget changes. So this is all here. You can take a look at this. You can see what the optimal allocation of these dollars would be. And then, of course, you can download this optimized budget. If we want to say, okay, let's go make these changes, let's share this with the finance team and get it rolling. So this is all available here for you. And again, Recast will automatically track your pacing against this. So you're always staying up to date. Are we on track to hit the goal or not? Now, the last question you might be asking yourself is, well, it's making these forecasts. How do we know if this is going to come true or not? And this is one of the really cool features of Recast is that we actually build backtesting into the platform itself. And so here we can see a bunch of backtests that have been run. And when I say backtest, you know, what does that mean? The idea behind a backtest is we want to test the model on how well does it do at predicting data that it hasn't seen before.

Right.

And that's the problem that we're facing today when we're trying to make a prediction about what's going to happen in Q1 2025. And I'm sitting here in, you know, the end of December 2024, we haven't seen what is going to happen in 2025 yet. And so we have to make a prediction without having seen any data.

Right.

That's a forecasting or a prediction problem. And in order to evaluate that, what we want to do is test the model in the same situation, but where we have seen the data. And so what we want to do is we want to say, okay, for the model that we trained in, number of days ago, 7 days ago, 35 days ago, 70 days ago, 91 days ago, that model hadn't seen the last 90 days of data because it was trained in October or September. But we have. And so what we can do is we can go back to that model and ask it to make a prediction about the future and then see how did that prediction come true or not. And that's built into this view itself. And so we can see for the last seven days what was the forecast of what was going to happen for the 90 range and the 50 range, and then what were the actuals. We can see that by day. Cumulatively, we can see this for the last 35 days, for the last 70 days, for the last 91 days, etc. And all of this is built in so that you can track and say, okay, you know, given how this forecast has done in the past, how do we expect that it's going to do in the future? When we're looking at our goals tool or the optimizer or the forecasting tool? So all of this is available here for you to be able to validate and really understand. Is this model doing what we expect? Can it predict the future or not? If it can't predict the future, then it's not really very useful. And so we build this into the platform so that you yourself can validate it. You don't have to trust us.

Right.

You can see on your own exactly how well the model is doing. You're predicting data as well seen before, and that can be very informative for thinking about, okay, when we're making this prediction for next quarter, how much accuracy can we expect?

Right.

Where can we expect to end up given what we've seen historically in the past? Hopefully this was helpful for you. This was a really brief, lightning introduction to the Recast platform. There's a whole ton of features, reports, workflows that we didn't touch on today. But if you have any questions, definitely reach out to us. We would love to give you a personalized demo of the platform and talk about how you might be able to use Recast to build an incrementality system at your business. Have a great rest of your day. Always staying up to date. Are we on track to hit the goal or not? Now, the last question you might be asking yourself is, well, it's making these forecasts. How do we know if this is going to come true or not? And this is one of the really cool features of Recast is that we actually build back testing into the platform itself. And so here we can see a bunch of backtests that have been run. And when I say backtest, you know, what does that mean? The idea behind a backtest is we want to test the model on how well does it do at predicting data that it hasn't seen before.

Right.

And that's the problem that we're facing today when we're trying to make a prediction about what's going to happen in Q1, 2025. And I'm sitting here in, you know, the end of December 2024. We haven't seen what is going to happen in 2025 yet. And so we have to make a prediction without having seen any data.

Right.

That's a forecasting or a prediction problem. And in order to evaluate that what we want to do is test the model in the same situation but where we have seen the data. And so what we want to do is we want to say okay for the model that we trained in number of days ago, 7 days ago, 35 days ago, 70 days ago, 91 days ago, that model hadn't seen the last 90 days of data because it was trained in October or September. But we have. And so what we can do is we can go back to that model and ask it to make a prediction about the future and then see how did that prediction come true or not. And that's built into this view itself. And so we can see for the last seven days what was the forecast of what was going to happen for the 90% range and the 50% range and then what were the actuals. We can see that by day machine we can see this for the last 35 days, for the last 70 days, for the last 91 days, etc. And all of this is built in so that you can track and say okay.