What Finance Got Wrong About Revenue Forecasting (and How Marketers Can Fix It!)

[00:00:00] Richard Gaffin: Hey folks, welcome to The Ecommerce Playbook Podcast. I'm your host, Richard Gaffin, Director of Digital Product Strategy here at Common Thread, and I'm joined here bright and early on, actually we're recording this on Halloween, by Mr. Taylor Holiday, newly celebrating his World Series champion Dodgers. Taylor, what's going on today. 

[00:00:16] Taylor Holiday: It's a great day. World Series champion Dodgers. My puppy turns one year old today. So all things are good in my house. Got a nice Halloween party with my kids friends tonight. So life is good.

[00:00:27] Richard Gaffin: Wow, beautiful. What are you dressing up?

[00:00:29] Taylor Holiday: Yeah, we're going family costume this year. We're the Sandlot. 

So, yeah, my son is one, one twin is Scotty Smalls. The other is Benny the Jet Rodriguez. My wife's Wendy Peppercorn and I'm a squints.

[00:00:40] Richard Gaffin: Nice. Oh, wow, you're going to

[00:00:41] Taylor Holiday: Yeah. Yeah.

[00:00:43] Richard Gaffin: Alright well, let's let's get into something else exciting. I, maybe even, maybe even as exciting as a Dodgers series, who knows? But we've, we've added something called the Event Effect Model. To the profit system, specifically to some of the forecasting models that we build for our clients.

And so we wanted To sit down here And talk through exactly what the event effect model is today. So Taylor, why don't you give us sort of the short summary definition of what, what this particular edition 

[00:01:08] Taylor Holiday: Yeah. and if you're new here, the profit system 

is CTCs 

sort of 

operating plan that we build on behalf 

of our customers to help unite finance and marketing into one cohesive 

daily plan that you can execute against. And if you hear inherent in that message is this idea of unifying these two seemingly disparate practices, 

finance and marketing into one cohesive 

strategy and one of the ambitions that I've had about really where that 

actually happens is.

To increase the value and importance of the marketing calendar as the mechanism that actually affects your revenue in the 

future. So if you think about it, 

the, the 

the primary key, 

When you think 

about any forecasting exercises, you are extrapolating the past. So any data model looks at performance in your past and uses it to build an expectation of the future.

Okay. Well, that's useful. Only in so much that the past is like 

the 

future. In other words, you behave the same, you run the sales at the same moments, you release the products at the same time. And a great illustration 

of this is like the chaos that a black Friday, cyber Monday being split between November, December causes to year over year.

Like when the, when the future deviates from the past, a lot of your models become not that useful. 

So what we've wanted to do is to figure out how we could begin to tie together your future planning and allow the. Actions and marketing initiatives that you have planned in the future to affect the forecast 

by connecting qualitative data around your marketing actions to quantitative outcome.

And so that's what the event effect model is all about. It's creating an effect from the events that you plan and execute on the marketing 

calendar. And so this is a really exciting way that we're improving our ability to forecast, improving our ability to plan and allow e commerce businesses to start to root around a very important principle.

Which is if you want to affect the 

future, if you want to change what your forecast says, which everybody does, everybody wants their forecast to be better. 

You don't just get to edit the numbers in the spreadsheet. That is not a real way to 

build a plan. Instead, what you have to do is build building blocks, new ads, new campaigns, new emails, new PR moments, new product releases, and those have to have a corresponding effect on your forecast, such that as you plan them, you see the outcome change based on the action, 

but you don't get to create changes in revenue, devoid of action.

And so this is a really important it's both ideological in that actions 

And that is a core tenant of our sort of system development, but also allows us to be more specific in the way that we plan and flow daily revenue expectations 

for our customers. 

[00:03:52] Richard Gaffin: Gotcha. Okay. So you make an interesting point there about how models essentially can only, can only model what's happened in the past and sort of, you know, Translate what would happen if the same thing happened in the future. But of course, in this particular, like, what we've done with the event effect model is what we're trying to do is model what would happen if the future was actually different from what happened in the past in some ways.

Right. Because the idea being is like, you have your, your model of what might happen, but then you say, I'm going to do, let's say a sale mid month, and then we're going to model what's going to specifically happen with that sale based on maybe how some sales happened in the past, but maybe talk a little bit more about specifically how this differs from what we 

[00:04:29] Taylor Holiday: Well, so yeah, I would actually say that what we've done is get more detailed about the composition of the past and use it 

to tie it together better to the future. So a lot of times people look at the past related to sort of the macro indicators. Revenue over time, spend over time. Maybe they break that down into a channel level or the customer file level, but we're actually going to deeper building blocks.

What happens every time you launch an individual ad? What happens every time you launch an individual email? What happens when you have a promotion? What happens when you do a product release and we're taking those individual actions and assigning a corresponding expectation of value to them 

so that, and the way I think about this is like, 

your, your business has a bunch of Lego blocks.

Right. And an example of a Lego block is an ad you launched in meta. It's an email that you send. It's a promotion that you run. These are all building blocks that have corresponding expectations of value based on using that building block in the past. Maybe you've run a 20 percent off sale every year on your anniversary sale.

And you have some reference to the effect that doing running that sale has. And so if you were to grab that building block out of the past, 

And put it into the future. You could expect some corresponding impact relative to its historical performance. That's what we want to do is we want to start to take all these individual little building blocks and assign corresponding impact to them so that when we put them into the plan in the future, it has an effect on the forecast.

And also really important. If we take it out of the future, such that that sale that you ran last March, isn't happening anymore. We can also point to why we're actually expecting the year over year results to be worse. And so I think that's really the key here is not to, to ignore the past necessarily, but to break it down into even more detail so that we could use the building blocks from it to affect the future.

[00:06:19] Richard Gaffin: right? That makes sense. So, So what types of building blocks then are available for us in the event in the 

[00:06:25] Taylor Holiday: So right now we have we have a pretty small set of groupings. We have about 10 different labels that we're using to define action. So it's product release promotion, a seasonal event. That would be like black Friday, cyber Monday and influencer release A VIP, we call like a VIP moment so that we have a lot of brands that do like, specific sales to their existing customer base.

And then we have a PR hit. So we have some customers that go on TV, they go on Fox news sometimes or whatever it might be. And they get effects from 

that. And so we're just sort of building this library of categories of actions. We're not product release. I think I might've said that, but and those are the labels.

So the way to think about this is in the metaphor I use is like building a chart of accounts. For your accounting software. So what final loop is trying to do is to look at every transaction that occurs in an e commerce business on your line item and assign a category to it in your chart of 

accounts in the same way, we want to look at every marketing action that happens across every business in our entire database, every kind of email that gets sent, every kind of ad that gets launched, what are the headlines?

What are the subject line? What do they contain 

new product releases, new things? What causes that big revenue spike? What could we categorize that as and create. 

The 

largest library of contextual marketing events that you can choose from. And then we both want to assign your individual historical value of those different kinds of labels, as well as the broader set of actions that occur across all of CTC.

So we can sort of understand what marketing moments are the most effective kind of marketing moments across all of our businesses. 

So this idea. Of sort of building a chart of accounts of marketing events, and then the corresponding impact for your individual business and brand is really what we're after.

And so the labeling groups are just going to grow over time. As we see more things that drive peaks and drive moments, you know, like, one might be a sweepstakes that you run, or like, we might come up with more and more ways that we identify different kinds of actions that we can start to group them into.

[00:08:16] Richard Gaffin: well, and I was thinking like as, as time goes on, you could also get more and more granular within specific groupings as well, I would imagine, like, let's say, because, because with product

[00:08:25] Taylor Holiday: it. Yes. Okay. Exactly. 

[00:08:27] Richard Gaffin: there's a pretty wide range of, of, outcomes depending on how good or bad the product is. 

[00:08:31] Taylor Holiday: Well, a lot of that. So that's, that's a great question. So product really says, so a lot of times when brands do this, they'll identify this as a tier one launch, tier two launch, and the, the, 

the question there is really usually about the amount of inventory that exists and how big of an opportunity the brand believes 

that to be. And so, yeah, the other data point we're going to bring into the event effect model for something like product is inventory position. So if a brand only has 50 SKUs of something, the expected effect relative to a drop that has 

5, 000 SKUs. Could be wildly different. Also, you have things like the progression of their email database and how the effects returning customer revenue versus new customer revenue.

That's a big part of what we're trying to model with the event effect model is how does this affect new customer acquisition versus returning customer revenue? Cause those percentages change a lot relative to what the moment is, meaning the percentage of revenue each day that comes from new versus returning customers, 

Isn't like perfectly.

Symmetrical all month long. So these are all different components and inputs that the key here is to start tying these things together and allow machine learning to sort of apply its it's magic of understanding more and more when you go to choose that event in the future, what's it going to do to your forecast?

And so that's where the hope is like the longterm. View of where we're headed is that the way you build your forecast is actually through the marketing calendar. You build it by telling the machine what you're going to do, and then it tells you what you can expect. And so that I 

think is really the key to getting forecasting even more, right.

Is to not do it top down from finance is to do it. Bottoms up from marketing by saying, what are the actions we're going to take in the future to get to where we want to go?

[00:10:08] Richard Gaffin: Yeah, and actually, so maybe let's dig into how, not, not even how forecasting happens generally now, but how how forecasting was happening for us prior to the event effect model. 

So, so if we weren't going off of basically the actions that we were going to take the next month, like, what types of things were 

[00:10:23] Taylor Holiday: Well, so we were, we were, we were, it was just human, right? So what would happen is, 

Let's say let's Black Friday, Cyber Monday is a good example where if you were to do a forecast for Black Friday, Cyber Monday, you might get to your spending revenue, but if you need to break that down to a daily level, well, then you have to go through and understand when am I launching the sale?

Am I launching it on the 

15th or the 20th? Black, when is the date of black Friday? And we would have a growth strategist go in and manually make adjustments to the daily flow of revenue and 

spend. And so we're trying to find all these places where we're asking people to do this sort of like, go look at the historical analysis.

What would I do and their application of it and try and apply 

it? So it's not that the calendar wasn't 

affecting our forecast in any way, I'd say we were already very much invested in that idea because we have to forecast daily revenue expectations but now we're just applying an initial. and, in reality, it's still a starting point. Because if there's still an imperfect 

relationship between the past and these actions in the future, but it definitely creates the flow much more consistent and gives you a good data analysis of here's this, all these events in the past, here's what happened because the other beautiful thing is like, not only.

Can we see the effect of the event? We can also see the performance to expectation, 

right? So you, because we have now histories of forecasting for some of these brands every day for many years, the 

database will show us like, here's what you expected AMER to be on that day. Here's what it was you were ahead of behind of your own expectation.

Here was the new versus returning 

revenue split here was the effect. And so you can go back and look at. It sort of surfaces that data that you would have gone to look for on your own into a very clear way, and it shows you how those actions are affecting the expectations in the future.

[00:12:10] Richard Gaffin: So this, this kind of feels like it's, it's part of a longer journey that we've been on towards almost redefining what a, what a forecast even is, because I think like, All too well, we've talked about this before all too often, like a financial forecast is treated like, like the weather forecast or something like that. It's something you have no control over, but you have certain, you have a certain sense of what dynamics might be like, I don't know, temperature and pressure or whatever. And so you can kind of guess and, but then once the forecast is built, it's kind of like, well, I, this will probably happen. And if it doesn't, then, you know, there's nothing we could really do about it, but we're trying to move it towards something along the lines of like, what if you actually could have controlled the weather the 

[00:12:44] Taylor Holiday: Well, so, yeah, you're 

going down some conspiracy rabbit holes now 

here, Richard, but, but we are interested in, we are interested in controlling the weather. That's sort of exactly it. You've probably heard me say this. If 

you listen to my content at all that I think forecasting is an exercise in execution way more than it is in modeling.

And that's why I think the key here is to really get clear on what actions you're going to take to execute against that. Because a lot of times what I see is like, I see businesses with these numbers that are just sort of devoid of how. 

It's like, Oh, there's the number, but it's like, how is it going to happen?

And if you start actually going, well, I'm going to do this. And then this, and this, and this, and then you add that up and you go like, Oh, well, that's not going 

to get you there. So how, and a lot of times, like we're in a conversation right now with a very large customer. And it's like, well, can you tell me where the spend and efficiency expectations for this month are coming from?

And it's like, that's the gap to the budget.

[00:13:35] Richard Gaffin: hmm. Mm hmm. 

[00:13:36] Taylor Holiday: Oh, okay. So you just, you just took the gap and you just said this much spend at this much efficiency equals. Gap solved. And that's what, when I say like spreadsheet forecasting, like that's spreadsheet forecasting that it's really easy to edit the numbers in the spreadsheet to make it equal to budget.

But that is not the same as being clear on. I'm going to spend this much money in this channel. It's going to generate this much revenue. I'm gonna spend this much money here. It's going to generate this much. I'm gonna send these emails that'll make this much. And when you add it all up, you realize like, Oh, it doesn't, it doesn't get there.

So what am I going to do to solve for that deficit? And I think that the amount of times I see brands start with a plan that actually has no chance of achieving the thing they say they want to do is all the time. And it's not to say that just cause you have a plan, even with estimations that it's going to get there, probably not.

To start, but the key is how quickly can you find out where you're wrong and course correct. And so there's more and more. We're just trying to be thoughtful about the plan. 

But the plan is not the point. The plan is the boundary that begins the journey. It's the map that tells us when we're off course and on course so that we can course correct because the actions are what matter, not the plan.

The plan is step one. But the real magic is in how you interact with it. 

[00:14:47] Richard Gaffin: Yeah. Well, I kind of love the idea that like we're talking about forecasting being an exercise in execution and not modeling, but it does sound like kind of what we're doing is moving forecasting towards modeling execution, 

[00:14:57] Taylor Holiday: right. 

That's a good. That's and that's what we say, that's why we use the phrase operating system, because this isn't a plan that we hand to you and walk away. This is how we 

work every day, 

right? And that's really the key here is, you're exactly right, is that modeling execution. And the 

other thing that I think it helps to do is some of these ideas, like if 

we talk about like four peaks theory or these different ideas that we have, it actually helps to bring to life.

Like, okay, let's look out at your forecast for 2025. Okay. March sucks. 

Like we don't like 

March. 

Let's talk. What are we going to do to change March? Because Left alone. This is the result of March. So what thing could we come up with? All right, let's brainstorm. Is there any product release that we could come up with into that moment?

Is there a big story we could put together? Is there a campaign initiative? Is there some cultural event? You know, like, what could we do? To take that crappy point in the forecast and drive it upwards. But the only answer to that, it won't happen on its own. You have to go design the action to create the peak.

And so I think it also helps to sort of really, again, connect to this idea of you don't like the numbers. Cool. What are we going to do to change them?

[00:16:07] Richard Gaffin: So what, here, here's a, maybe a broader question. Why do you think that, that forecasting Is has been the way it has been, I 

guess, like, why is modeling so disconnected 

[00:16:15] Taylor Holiday: Because 

the 

people responsible for it aren't responsible for the actions. That's why, so the finance team doesn't control the events that drive the revenue. This is why I don't believe that finance should be doing forecasting. 

They don't control the actions. And so if it's the same thing of like, why do media buyers start like going crazy and ad accounts if they don't have access to creative, it's like, cause they don't have the things they need to actually affect the results.

And so they'll start tinkering like crazy. And that's why It's like, you need these things to be connected. And so I think it really just comes down to the finance team can't change the actions. They don't get to build the marketing plan. And so all they have is the modeling. That's all they only are allowed to edit the spreadsheet, just like a media buyer is only allowed to edit the campaigns, but without actually controlling the building blocks, the things that changed the numbers.

That's all they're left. That's all they have control over. And so these things just become a natural outlay of the authority of the person creating the plan. And so that's where I think the real trouble is.

[00:17:13] Richard Gaffin: Gotcha. Cool. Well, so actually, is there anything else that you want to hit on this 

[00:17:17] Taylor Holiday: Yeah, it's funny because like, just on this last point, as I'm saying this, I think we've got something flipped, which is that, you know, Finance should be responsible for marketing measurement and marketing should be responsible for financial planning. Like, I think we've got this thing flipped in terms of what finances is really capable of doing is measuring financial effect of the actions, but not necessarily of planning and designing the expectations of them on the front end.

And I think that can be collaborative. But I just think that there's yeah. We have to understand that marketers create revenue. Like that's, that's where it comes from. And I, you know, I think in partnership with product and inventory, those are actually the units of revenue. So that group is what actually goes out and makes things happen that create money.

Finance has no control over that. That's not, that's not because they're not useful and helpful, a really important side of an organization. They're just not the creators of revenue. And so I think that that's it's really important that we assign This idea that we create revenue by the things we do.

And so we have to really make sure that those things are associated with one another.

[00:18:17] Richard Gaffin: Well, actually, so, so on that last point, so would you say that's fair to say that, like, the utility of finance, of the finance team is to Identify what has happened in the past, or maybe even identify things like, well, obviously, like, Hey, we're spending way too much on on the ping pong table or

[00:18:34] Taylor Holiday: Yeah, yeah,

[00:18:35] Richard Gaffin: in the

[00:18:35] Taylor Holiday: Control costs. And, but

[00:18:37] Richard Gaffin: about what's going on in the

[00:18:38] Taylor Holiday: yeah, yeah, I think you're exactly, they should they should provide the effect model, right? Like in the sense that what they should be saying is, okay, every time you send an email. So this is the value of that unit. So plan accordingly. Every time you launch an ad, here's the value of that unit.

So plan accordingly. So like in our world, the creator of the events effect model would be data science and finance. It's an analysis of the financial impact of the marketing actions that have been taken. And it's provided to the marketing team as a resource for their planning exercise. Right? So here, here's what an email is worth.

Here's what a product release is worth. Here's what a promo is worth. Now build your plan to get to the best possible result based on what things you can execute in the correct time horizon. So I think in that way, it's like these models, the same thing with like our spending AMR model or our cohort specific LTV forecast, their resources that finance creates for marketing to use to understand how to build the best possible plan to accomplish the best possible financial result.

I think they can also provide transparency into areas where costs are becoming are, are not where we want them to be and can put pressure on the team to go solve gross margin or to say the, Hey, we've got a rising OpEx or, Hey, there's this our, you know, Duties and VATs were higher than expected because we sold more internationally than we plan to.

Can we get that back in alignment? And that's where it's like, they're providing this reporting and boundary against the measurement expectations of the plan that's created, but they can't create the plan. They don't, that's not their competency. So it's a dance. They're both so essential to the organization, but I think more and more, we need to move this idea of the revenue creation model into marketing. 

[00:20:12] Richard Gaffin: That's right. Cool. All right. Well, folks, I think that'll do it for us for this week. But if you're looking for a team. To help you build a financial model that puts you in the driver's seat of how to bring it to life. Please just hit us up commonthreatcode. com. Hit that hire us button, drop us a line.

We would love to chat with you. All right, folks. Appreciate you joining us. Take care and we'll see you next week.