Science in marketing - modeling growth with data

Science in marketing – modeling growth with data

Growth modeling is an absolute must in every marketer’s go-to tool stack. In this article, you’ll explore why you should use it and how to make the most of it.


Table of Contents


What is the one equation that describes our business?

Tomasz Tunguz, Venture Capitalist at Redpoint

Answer this question and a whole new universe of possibilities opens before you. Find the vital metrics for your business and take them apart to learn what influences them. This alone will show you what to look for in your analytics and what to focus on in your marketing activities. Thus, the data will become your powerful friend.

Find the vital metrics for your business and take them apart to learn what influences them.
Credit: tomtunguz.com

There’s over 84% of websites that use Google Analytics. However, this isn’t proof that all of them are set up correctly or track any specific goals, or the marketers who configured them knew the business’ key metrics. That may be the reason why in 2018, basing more decisions on data analysis was the objective for over 50% of marketers.

Companies already were aware of the value data has for them. However, they didn’t know which metrics to focus on, which ones will bring the most growth. It’s no surprise then that growth modeling comes up more often.

Modeling growth – basics

As all or at least some decisions should be based on data, you should become friends with analytics. Those numbers show all the growth possibilities out there and allow you to build powerful strategies focused on attaining your goal. So, you will be the one who will let the data model business’ growth.

To use the numbers gathered in the analytics tool and prepare possible scenarios of growth, you can use either Excel/Google Sheets or a modeling tool that does all the calculations and builds a growth model for you.

What a growth model is? Here are definitions from a few articles that look at this matter from a marketing point of view:

A growth model is an equation that tells you what are the different variables in your business and how they work together and translate into growth.

Growth Hackers

This indicates the similarity between a growth model and a business model. However, the latter is a more strategic document, while the growth model goes into much greater detail. It tells us what particular elements contribute to achieving specific goals. Also, it can indicate what areas of the funnel should you concentrate on.

It’s a new concept for an old idea about creating a simplified high-level model in order to make better business decisions. You’re trying to explain “how does this company grow?” What levers exist as inputs that contribute to “growth” as an output (however you define growth)?

Alex Birknett

In other words, you’re trying to explain how growth works in your company. What channels bring the best results? How do they contribute to the bottom line? Pulling which growth levers will result in the biggest output? And all other questions you can think of, business-wise.

Growth models… are feedback loops that project how one cohort of users leads to the acquisition of the next cohort of users. Viewing growth with this reinforcing model simplifies a complex system with tons of moving pieces to a set of functions and assumptions.

Segment, found on alexbirknett.com

That sounds just great but only if you have data on the first cohort. Usually, the best would be data gathered from at least several weeks. Otherwise, when you’re starting anew and have no data to look at, you have to rely on third-party benchmarks. You can find those in various researches available online or sources like Adjust that provides you with the overview of the vital metrics.

Remember that if you want to focus on the full funnel or just on its specific stage, it’s important to understand the entire setup to know how each part of the funnel affects another. Depending on how many stages and platforms your funnel includes, the model will reflect its complexity.

The importance of growth modeling

When you think about growth, how do you perceive it? Is it a straight line? A picture-perfect situation where everything runs smoothly, and you see improvement each month?

Or is it more of a bumpy road you need to work hard to get through? With ups and downs, surprising turns, and unexpected dead ends?

From my experience, it’s the latter.

It’s crucial to understand that a growth spike isn’t the result of good luck. It’s something that starts long before that boost.
It’s crucial to understand that a growth spike isn’t the result of good luck. It’s something that starts long before that boost. Read our customer’s growth story.

And here comes growth modeling. To help you predict those dead ends, highs, and lows.

Even though it’s only a prediction based on generated data, it shows you how much growth you can expect with presented benchmarks. It’s not 100% right, but at least it will be something to compare real performance to.

You need to remember that like in every economy, a business has limited budgets and resources. Using them in the best possible manner to reach business goals means saying Yes to a few ideas and No to most.

Like in every economy, a business has limited budgets and resources. Using them in the best possible manner to reach business goals means saying Yes to a few ideas and No to most.

And the more plausible the specific action is, the more likely it is for the stakeholders to agree on it. Despite it being a forecast, a growth model is more credible than improvising with various ideas that aren’t supported with data. There is no guesswork here. It will present you with arguments for or against focusing on a specific metric. You may even find bigger opportunities out there for you to jump at. Thus, you will be able to use those limited resources well.

Though to use the full potential of growth models, you need to update them regularly with actual results, once you put it to work. This will help you validate what worked and what didn’t. Plus, you will provide a solid base that will make future predictions more accurate, taking you closer to attaining your goals.

Why should you go for it?

Capital efficiency and understanding the levers in your business model needs to become the top priority to make sure you have the best shot at long term success.

Alex Oppenheimer on Medium

Besides all those things mentioned above, what Alex Oppenheimer said should be a sufficient answer. However, there’s much more to what you will gain from growth models than that.

  • You’ll understand the levers in your business model, as you will see what works and what doesn’t. It’s just another way of getting to know your target audience and their preferences. Because, as mentioned before, a growth model will reflect the complexity of your funnel – either full or any specific stage you want to focus on.
  • It’ll give you valuable insight into the credibility of your data. Not only will it point out what you should measure, but it will also show if analytics are set up correctly.
  • Like all other marketing techniques, it will show you what part of your funnel needs fixing as obstacles your prospects meet on their journey through your funnel will prevent them from driving up the specific metric.
  • As the growth model works on the accumulated data, it shows the benchmarks and growth levers, as well as uncovers gaps in the plan.
  • It indicates what would be the outcomes with a specific budget, what to fix or test, which channel to focus on (budget- and work-wise), and which elements are the main factors in attaining the growth.
  • It helps in determining the priority of various initiatives with a simple simulation of their overall impact on your business’ growth.
  • You’ll be able to predict what happens if and play around with possibilities.
  • It puts reverse engineering to work – showing if the goal is attainable by working backward from the final goal through campaign metrics to the budgets.

Overall, you’ll be able to build a strategy based on the model and provide your stakeholders with a credible forecast of what your plan can give them. Thus, you show them your train of thought, how deliberate is your plan, and build trust on that basis.

However, you need to remember that no matter how much effort you put into your marketing activities, there’s nothing you can do if a product (for example a mobile app) isn’t developed well. You can only steer the developers into a problematic area.

No matter how much effort you put into your marketing activities, there’s nothing you can do if a product isn’t developed well.

Basic ingredients of the model

There’s a couple of things that are included in the growth model:

  • Actuals – data that exists on the account, or the third-party benchmarks if you have no data. They’re the baseline for your forecast, and you can’t tweak them. As time goes on, you have more and more of those;
  • Inputs – estimates provided by you. From analyzing existing data, you can come up with an appropriate budget range or range within which you’ll be able to improve certain metrics, i.e. click-through rate or customer value proposition growth percentage. You can play around with them;
  • Calculations – equations that transform inputs to outputs;
  • Growth levers – answers the question: “what gives you the most results for the least spend?”, presents the best growth opportunity for the lowest costs;
  • Outputs – final forecast;
  • Probability forecast – the range where your results can fall in.

In Ladder, we recognize two types of growth models:

  1. Based on third-party benchmarks – usually applies to startups, but may also apply to a company that launches a new product for a different market or industry. There’s no previous data you can base your forecasts on, so you need to use industry data.

    This type of growth model is super theoretical. You use third-party benchmarks to come up with hypotheses. You test channels, ad formats, audiences, and others, and see what results it will bring. Once you have enough results, you can identify the best performers. These become the growth levers you push on.

  2. Based on generated data – those who have data on their account. You build a model based on those actuals, enrich them with inputs, and calculate what will work best for you.

It’s good to get data from campaigns that have been running at least several weeks, preferably a few months or even more than six months. This way, you’ll have enough data about such metrics like cost per click, budget impact on lead generation, conversion rate, and others.

As you progress through the campaigns, there are a lot of forecasts that will turn into “actuals”. It’ll update your model to how things look at the moment and show you how this newly generated data impacts your forecast.

You can either use Excel or Google Sheets to write down all the calculations. Or, you can use a tool like Causal that will do those calculations for you and present them on diagrams.

As Excel and Google Sheets may be something familiar to you, let’s go through the process of building a growth model in Causal.

A step-by-step guide to Causal growth models

  1. Assess the business model and goal:
    1. Different business models have a different funnel setup, so it’s vital to understand the step by step events that lead to the conversion and metrics that influence it, for example for Facebook Ads you have CPM, Impressions, CTR, conversion rate on-site and conversions. There’s a whole journey to go and your model should show what are the steps.
    2. The goal is different from the conversion itself, we must understand what is the goal of the project, is it more volume, better efficiency, gathering data or something else.
  2. Mapping out the funnel:
    1. You can do this either from the start of the user journey to the conversion or the other way around – reverse engineering from the conversion back to the beginning of the journey,
    2. It is important to map out as many measurable conversions and micro-conversions as possible, for example in a mobile app funnel we have an ad impression, ad click, traffic in the App Store, Installation, Registration, and then real usage.
  3. Choosing the channels:
    1. The most commonly used channels for Ladder clients are Google Ads and Facebook Ads, as they are the most powerful ad platforms and the best fit for most businesses. We can add for example Bing Ads or LinkedIn Ads on top of that or instead of the main channels, depending on the project,
    2. You may also start with only one channel if the budget is low and not worth spreading over multiple channels,
    3. Pick the channels that fit your budget and are the most relevant for your project to find the best prospects there,
  4. Coming up with the data, you can use either or a combination of these:
    1. Your previous data that has been recorded in the chosen channels and/or on the website/app,
    2. Benchmark data from third-party sources to foresee what numbers are possible;
    3. The metrics we use most often are CPM, CPC, CTR, CVR, CPA. I recommend using ranges like 0.5% to 0.8% CTR instead of an exact number. This gives more wiggle room and you’re not promising anything other than a range. So, pick one that you feel comfortable with;
    4. What we like to do with these data points is predict that they will be improving over time through optimization and testing, so we add for example a 3% expected improvement on CPC month-on-month;
    5. It’s important to input the budgets that will be used for the campaigns.
  5. Inputting the data into the model. Now, that you have the funnel, channels, and data, you have to plug in the numbers into the model.
  6. Plotting the data onto graphs:
    1. Showcase the model through a series of graphs that visualize the predicted outcomes of the campaigns;
    2. Plot the data onto graphs that show the level of CPM, CPC, CTR, CVR, CPA over time. It’ll show how they’ll be evolving given the predicted month-on-month improvements;
    3. Since you’ve picked ranges for the metrics we are using in the model, you’ll be seeing a range of outcomes;
    4. For most clients, the most important graph is the one showing the number of expected conversions over time. Especially important here is that you’ll be given a range within which the model predicts the total number of conversions. For example, the graph will show 500-580 app installs per month;
    5. It’s much better to give a range than a specific number to not be held accountable for not delivering that particular metric.
  7. Once you have that model built out, you can implement additional scenarios:
    1. Present 3 scenarios to prepare the person on the receiving end for the possibility that things might go bad or good,
    2. Create an optimistic, default, and pessimistic scenario,
    3. The differences are usually with the levels of one, some, or all of these metrics – CPM, CPC, CTR, CVR.

Here’s how the built growth model will look like:

Best practices

Whenever you build a growth model, it’s best to remember to:

  1. Define the core metric first.
  2. Know your customer and your funnel.
  3. Create three versions of the same model (default, best-, and worst-case scenario) to be prepared for every possibility.
  4. Update Actuals regularly with new data.

And other than that – be ready to grow.