Recast Incrementality Platform: Key tools, features, and reports

Hi, I'm Tom Vladek, co founder of Recast, the media mix modeling platform where the world's most sophisticated brands measure and optimize their marketing spend. Today, I'm going to take you on a grand tour of the Recast platform. I'm going to cover every major feature and let you know how it works and, and what it can do to help you manage your marketing budget just a little bit better. So you can see we're on the demo page today. We're on the demo client, and they're modeling total sales. We've just updated their model through December 9th. Now, just so you know, we update models every week, so every single week, you'll get a completely new dashboard with updated data. And I'm going to walk you through how it works right now. So the first feature that I want to talk about is the goals feature. This is to help our customers set and hit their marketing goals, whether it's for the next week, the next quarter, or the next year. So here you're looking at a quarterly goal. The goal Is to hit $20.1 million in sales. Our current model is estimating that they're going to just barely beat that goal by about $100,000 in revenue and that there's a 75% chance of hitting that goal. Now, as I mentioned, every week we're going to get an updated estimate of this. In addition, we don't just tell you what your likelihood of hitting the goal is. We also give you automatic recommendations that tell you how we think you should change your budget to give yourself a better shot of hitting that goal. All of these recommended changes are budget neutral, so not a dollar of additional spend is recommended. And they range from conservative, very close to your original budget, all the way to aggressive. In the aggressive case here, we're saying that there's a 98% chance of hitting that goal. Now, I want to take you back to the Insights tab here. And so now we're on the last week snapshot. This is where the majority of our customers are going to go when they log into the app first. And so what the last week snapshot is, is it's just that it's a snapshot of your last week's performance. The new data that we've seen since the last model. Now, the killer plot here, well, there are actually two. But one of the killer plots here is this waterfall chart which breaks down where your sales came from. So here we've got the intercept. Base sales organic, about 16%. Facebook prospecting, about 67%. Some other channels rounding it out. Here, we don't just break down where your sales came from. We break down the ROI and marginal ROI of every channel and which channels are accounting for more of their relative effect, punching above their weight and which channels are punching below their weight. This is all, of course, very easy to download at the bottom of the page. So the last week snapshot is just that, a snapshot. But if you want to really dive into the details of how your channels have performed over time, we have that in the channel level deep dive. And the deep dive has information about how performance has evolved over time. Right now I'm just showing you the graph of roi, but we also include the marginal roi, how much spend was driven, and when that effect landed after you account for timeshift. Now, in addition to that, we have details on the shift curves for every channel, whether they move fast or whether they take time to have their effect, and the saturation curves, meaning that for every additional dollar, how much additional revenue are we expecting? And we have this by impact, by ROI and by marginal ROI as well. Next up, we have the intercept, which is the level of baseline or organic demand. This answers the question, if you were to turn off your marketing spend, what would remain? Would 50%, 30%, 10%, or 90% of your business still be there? This is critical in figuring out what the efficacy of your paid marketing program is and how much of your business is organic versus paid. So whereas the intercept tells us the sort of baseline level of organic demand, the spike summary tells us how that demand changes around holidays and promotions. I'm going to zoom in on the Black Friday case because I think that one is super interesting because it illustrates the power of Recast's modeling of holidays and promotions. So you can see here that this is exhibiting what we would call the pushback and the pull forward effect simultaneously. So we're saying that in the days running up to Black Friday, customers are delaying their purchases and they've also accelerated their purchases. That would have happened after Black Friday, but now land on Black Friday Black Friday itself instead. We know that other kinds of holidays and promotions can have this sort of effect. And Recast models all of this out of the box. And it can tell you whether or not your holidays and promotions are actually incremental or if they just look that way because they're piling a lot of cannibalized sales onto a specific day. Now we know that it's not just your marketing spend and spikes that affect your sales. Things like your price, your brand awareness, competitor activity have major effects as well. We include those effects in something we call the context summary, which are variables like price and brand awareness that can have a differentiated effect both on your organic and paid marketing. Here we can see that we have both price modeled and brand awareness modeled, and we can see the effect that it's having both on your organic sales and paper paid sales. We can see this effect both in a static snapshot view, but also how this effect has varied over time. So we can see that brand awareness, as it has gone down and then up, has raised both organic and paid sales by between 110 and 125%. Finally, why should you trust any of this? Right, you should trust it because Recast is super rigorous about running backtests. What are backtests? Back tests are we take a model that we deployed seven days ago, 30 days ago, 90 days ago, and we say to that model that hasn't seen the most recent data, how would you have predicted sales over this time? Knowing only what was spent, nothing more. And we publish these backtests to our clients and most of the time we do pretty well, although sometimes there are misses. In this particular demo client case, almost all of the models are nailing it spot on. This one has a slight miss by about 8%, but the more recent ones are within 1% error. These forecasts not only show how accurate those models were at predicting future data, but they also show how the model has changed in light of the new data. How have the estimates been revised? In this particular case, we think that the intercept went up by about $20,000 in that time period and that sales from non branded search declined a small amount. So we're publishing not just the predictive performance of our model, which we would stack up against any other, but also how the model has revised in the meantime. That's great looking into the past, but we care about planning the future and for that we have the forecaster. The forecaster is a tool where you can upload your marketing budget for the next week, the next quarter, the next year, in fact the next two years, if you want to go that far. And then Recast will run it through the model and produce a range of predictions about what the model expects that that budget will produce in terms of revenue or conversions or whatever you're modeling in the future. This is what a forecast looks like. This budget has already been uploaded. This this forecast has already been run and we can see that it is forecasting somewhere between 6.9 and $7.2 million over the next roughly 30 days. And not only that, but we break it down channel by channel and we say, you know, this channel, we expect to have an ROI of about 6. This channel has an ROI of about 9. This channel has an ROI only of about 0.6. These forecasts take about a minute to run and they're super easy to do. So. So it makes it dead simple to run a range of scenarios and to see what the trade offs are. Now, whereas in the forecaster you give recast a budget and it tells you what's likely to happen, the optimizer goes the other way. You tell recast what you want to happen and then it will tell you what the budget should be. Now, in this particular case, this has already been run. The objective was to reach a target outcome. So this user wanted to reach $10 million worth of sales you in the period between December 10 and January 31. This isn't the only objective. In this objective we're trying to reach $10 million worth of sales with a minimum amount of spend. But we have other objectives as well. A very common one is to reach a target efficiency, say a 5x ROI. Now that will try to maximize your spend while staying above that efficiency threshold. We have the objective in this case reach $10 million worth of sales. But we don't want the optimizer to just give any answer. We want to give it constraints as well. The constraints are super powerful. In this case we're saying that we want to spend between zero and a million dollars in total across these five channels. And we have some additional constraints for linear TV and mailers. We hit run and it produced a recommendation. Our goal was 10 million. We expect to slightly exceed that with a recommended spend of 1.4 million, which was the minimum amount of spend that the optimizer was comfortable recomm while being fairly confident that you would exceed this $10 million goal. You can see how spend is broken down both by channel and by day. So for example, this spike here is when the optimizer is saying that mailers will come online and be spent. Now we can see the optimized daily outcome and the optimized cumulative outcome which has us landing right above the 10 million dollar mark by the end of the period. And of course we also break down performance by channel and tell you how to download that optimized budget. As I was walking through the Insights tab, if you came across one of those reports that you wanted to customize for any reason. So for example, if you wanted that really great waterfall summary, but instead of by Last week you wanted it covering, say a couple different months. You could run that here in the reporter. This gives us a range of different custom reports that, that we can run with different parameters to get custom reports anytime you want. Now you're probably thinking at this point, wow, this is so amazing. Must take months and months and months to produce. Absolutely not. In fact, once you're up and running, this runs automatically. Every week you give us the data, we set that up with your data team and it just runs automatically. New dashboard, new model produced on a week by week basis. To get it up and running the first time, there's a couple different steps. So the first step is that we work with you to get the data. For some of our clients, this takes about a week. For some, it can take a little bit longer. But the beauty of Recast is that we only work with first party data. So the only data that we require from you is stuff that you already have. What did you spend on which channels, on which days, when did you run promotions and which holidays are the most important ones for your business at a very high level? That's it. Now, of course, we're in the real world. We know it can sometimes be slightly more complicated, but we've never had any issues with clients putting that together. The second thing that we need to do is have about a 30 minute conversation with you about your going in. Expectations around the business is organic, 10%, 30%, 75% of your business. Those rough ranges are all we need to get the model started. Then we take it back to the lab. There's a whole lot of checks that we do to make sure that the model is up to our standards before we deliver it. That can take anywhere between one and three weeks. At that point, we schedule our first initial call with you, we deliver it to you, and now you have a dashboard up and running, ready to go. Recast would not be possible without leveraging the state of the art in statistical computation. As much as I would love to get into the weeds with you about how it works, that will have to wait for another time. But suffice it to say that Every week about 40 large servers are running to update every single client's model. What's happening is that Recast is running millions, billions, actually much, much larger than that number of simulations to find the right combination of parameters for your business. Now, if you're interested in applying the state of the art to your business, Simply go to getrecast.com and hit book a demo. Thank you very much. Anywhere between one and three weeks. At that point, we schedule our first initial call with you, we deliver it to you, and now you have a dashboard up and running ready to go. Recast would not be possible without leveraging the state of the art in statistical computation. As much as I would love to get into the weeds with you about how it works, that will have to wait for another time. But suffice it to say that Every week about 40 large servers are running to update every single client's model. What's happening is that Recast is running millions. Billions. Actually much, much larger than that number of simulations to find the right combination of parameters for your business. Now, if you're interested in applying the state of the art to your business, Simply go to getrecast.com and hit book a demo. Thank you very much.