What separates the top 5% tech companies from the rest? If you have looked into this answer on the web, you’ve probably found the result in: low customer churn, a sticky product and superior customer experience. But one thing that is often ignored is distribution. In this blog post, I’ll focus on the marketing processes great companies use behind the scenes to build data-driven organizations.
Peter Thiel, Co-Founder of PayPal and writer of the book Zero to One famously said:
“Most businesses actually get zero distribution channels to work. Poor distribution — not product — is the number one cause of failure.“ If you can get just one distribution channel to work, you have a great business.”
Let’s continue Peter’s trail of thought:
Having no distribution channels, i.e. marketing channels means you go out of business.
So if you get one channel to work, you have a great business.
Get several to work, you may become unbeaten in your market.
Now the advertising market is super loud, noisy and in many ways saturated. Traditional methods of advertising have been repeated so often that audiences start to ignore them.
So it’s only logical one would ask how to get a distribution channel to work in today’s age.
You could wait for new apps or websites to arrive, such as Slack, Medium, TikTok, Zest or Messenger at their inception, then use them as distribution channels while the competition is still sound asleep.
But this does put you kind of in “Reaction Mode”. You’ll only be able to act when something new is launched. And you’d have to be among the first on there, and have enough scalability, to actually reap most of the benefits.
So what happens?
The best companies start doing new things on existing channels.
Experiment (variable) on Channel (fixed) in order to optimize Revenue.
Many companies here optimize on the channel level instead of revenue which is a problem in itself, but we’ll get to that in another post.
Basically you run a number of experiments in the hope to optimize the outcome.
You then find something that works and do it for just long enough until everyone else figures out it works. And then you start looking for something new again that works. And the cycle repeats.
So in summary, how the best companies are making distribution work in today’s age is in two ways:
1) doing new things on old channels
2) or doing new things on new channels
Once in a while, you may find something that has been repeated so often, but it still works.
You, sir, are extremely lucky and the Gods are with you.
But in general, most companies are forced to constantly experiment with new things if they want to achieve the highest output.
And so marketing has to become obsessed with data. The more data you have, the higher your chances of finding what works.
Ok but let’s continue by analyzing what great companies do.
“They foster this growth mindset inside the company.”
The “let’s optimize everything perspective”.
They create a refined, agile process to find, prioritize, test, implement and measure experiments quickly in the search of seeing what works best.
Before we expand further and take you through some of the actual processes; let’s take a look at what sort of organization could handle data-driven processes.
No matter how much I personally like to praise the concept of speed and agility, there is a minimum quality and viability needed to put things on the market.
You may have heard of the MVP (minimum viable product) and the MMP (minimum marketable product) but Brian Balfour coined a new term specifically for running marketing experiments.
The “minimum viable test”.
It’s simply a fact that you’re not really testing an ad, channel or product if it’s so bad, misunderstood or not aligned with your audience that nobody is going to be genuinely interested.
All stats will be insignificant so drawing conclusions on that data won’t bring you far.
This brings us to the first part of producing performance:
1. Human capital
“The skills, knowledge, and experience possessed by an individual or population, viewed in terms of their value or cost to an organization”
This measurement is your best friend in building a data-driven organization from scratch.
In the end, the unit economics don’t really care about how caregiving you’ve been to your employees.
Hire people who can produce high ROI by:
- understanding your market and its problems
- having the skills to execute
- being able to run minimum viable tests
There’s only one advice I have here when hiring your top candidate: “Don’t look at the money.”
It’s not that I myself am such a baller – I’ve just learned that money influences rational decision making.
So make sure your person checks off the criteria and then check the price-tag.
2. Role distribution
The bigger you become, the greater the need for structure. Especially when it comes to experimentation.
Startups below 20 employees are basically exempt from this rule, but if you really want to build a dedicated data-driven marketing organization, you need to assign posts (fixed positions).
When we do a bit of everything, we end up doing or impacting nothing at all.
You could have made more optimizations or discovered new ideas if only you had spent more time on LinkedIn Ads than on Google Ads. And vice-versa.
Your work starts to compete with each other and it gets in each other’s way.
You do everything at a rate of 30% effectiveness but nothing really at 100%. So a lot of effort, energy, money and opportunities gets lost here.
The only way to be efficient in terms of a data-driven approach with experimentation is to have people assigned to posts. And to have them focus on one or very few things.
You can improve data to the degree that you’re responsible for it.
Basically, if I’m the boss and I need to sign off 50 people’s decisions on whether they can run experiments on their channels- that means that I’m responsible for all data, which is simply too much to handle.
Whereas people get confused if they have too much to take care of on an operational level; managers get confused if they’re responsible for too many variances of in-depth data.
You simply don’t have the exact know-how or deep insights to be making decisions over fragments of data you see 1x per week during a reporting session.
Give people the responsibility to make decisions, to fail and optimize all on their own.
A person asking you permission is not autonomous. As long as they’re learning in the long term and are sticking to known team policies, they should have complete freedom while experimenting.
Basically in a data-driven team, you want people to tell you what they’re doing and when something is good to go, rather than asking you if it can be done or if it’s good to go.
Make sure that you have policies set up in your team so everybody knows what is expected and what the modus operandi is supposed to be.
Not doing this is the quickest way to fail.
Write down everything that is important to you and your team and how you would like to collaborate. And simply avoid the thousands of miscommunications and mistakes that would be made otherwise.
- What kind of tests will you do?
- Where is all data stored?
- Who is responsible for updates?
- What are communication procedures?
- How do you track time?
- What is the definition of done?
- What tool stack will you use?- How do you calculate ROI?
- What are your short-term and long-term goals?
- What is our north-star metric?
Sit down with your team and create mutual agreement and understanding of how things will go.
If everyone agrees and understands the policies that are set, you’re all headed in the same direction and become an unstoppable force.
If you don’t, people will run experiments with different outcomes, goals, measurement sticks etc. all creating opportunity cost.
5. Knowledge gap
If you’re missing knowledge or expertise in your team to make things successful, make sure you add this to your team ASAP.
Hiring a consultant or agency might have a higher price tag, but the ROI of setting things up correctly to increase the knowledge and experience of Full Time Employees for years to come doesn’t compare.
Setting up data-driven marketing processes
Step #1: Decide what you are trying to maximize for
In most companies we’ve worked with, this will probably be something like:
- increasing registrations for our product
- getting more qualified leads / demo requests
- getting more SQLs (Sales-Qualified-Leads or SALs (Sales-Accepted-Leads) or PQLs (Product-Qualified-Leads)
While these are all valid metrics and KPIs (Key-Performance-Indicators) to track, we recommend always adding a second benchmark being
- How did it contribute to revenue?
- What was the non-monetary contribution?
Using revenue as part of your data-driven system will not necessarily help you qualify or disqualify channels, since we’re talking about integrated marketing here, but it will rather help you uncover which parts of your funnel you need to still work on. Where does the conversion break down?
And using non-monetary contribution as a metric will help you understand that marketing is not only about Lead Generation but also about brand awareness and demand generation.
Step #2: Understand the fundamentals of your business
Now, As you get started, you need an anchor point. The field of marketing by itself is way too big to just start doing things randomly.
That means, narrowing the possible things down to the ones that make the most sense for your product, market and business model.
At Zengrowth, we use this framework created by Brian Balfour in which we don’t only look for Product-Market Fit, but also for Product-Channel Fit.
Now understanding the fundamentals of your business means understanding:
- value proposition
- customer segments
- target audience
- knowing your ARPA (Average Revenue Per Account)
- knowing your CPL (Cost-Per-Lead) or CPR (Cost-Per-Registration)
- ideally knowing your CLTV (Customer Lifetime Value)
- Buying Unit
- customer journey map (who in the Buying Unit is involved when)
Once you’ve got all of this noted and a deep awareness, let’s visit the next step in the process.
Struggling with data tracking and you’re looking to raise series A? Check if you’re ready for growth.TAKE A QUIZ
Step #3: Analyze the ARPU (or ARPA) – CAC spectrum
As VC Investor Christopher Janz described, you need to know who you are targeting.
Are you targeting elephants, deers, mice or flies?
However, knowing who you should target also changes the price tag of how much you can or should spend for a customer. Also often measured through CAC (customer acquisition cost).
If we’re talking about a company that is going to pay $250,000 a year and we are aiming to use ABM (Account Based Marketing), and we know that we’re dealing with a sales cycle of 6-18 months, we’re going to be spending significantly more money per lead than if we were generating a lead for a self-service product priced at 10$/user/month.
So you’ll need to take a look at which channels fit your product, model and market and your max CPL based on your average ARPA.
From here you can deduct your scope of channels to experiment within
I do have to say that although low CAC channels generally don’t work as “conversion” channels for high ARPA products. They might and sometimes do work as brand awareness channels for simply generating contact data after which you’d normally have to nurture them for a very long time.
Step #4: Create your backlog
Now that you’ve deducted which channels make most sense, sit and brainstorm with your team about all ideas. Every single thing to be tested or which to be run an experiment for.
List everything down and choose your focus.
There are two ways you can go here:
1) Be as detailed as possible. No high-level categories such as: “Run LinkedIn Ads”. Yeah, we know that, but what are you trying to test exactly?
And by that approach you’ll easily be able to come up with 200-500 experiments and in-depth experiments to list down.
2) Create high level categories and only break down the multiple experiments in a dedicated experiment doc by that same category. So for example, if you run LinkedIn Ads, you may have different variables to test, which you then track in the experiment doc.
You want it to be looking somewhat like this:
The point is to use a template to specify each experiment in much more detail.
Here’s a growth experimentation backlog template you may use from Sjoerd Handgraff:
It depends whether you chose approach 1 or 2, but both should work.
Now set a date or have a regular interval period within your team where you add new experiments to the backlog so you can foster a culture of “Growth and Optimization”.
Step #5: Prioritize your backlog
Ok, now that you have jotted down plenty of ideas, you need to prioritize them.
At Zengrowth we use two measures:
- Quantitative predictions – What are the experiments for which you in some capacity have previous data? Would this previous data indicate or hint that a new experiment could be more successful than another?
- Qualitative predictions – Do you have any other data such as surveys, expert opinions, consultancies who have run such experiments which would help you prioritize experiments?
The main factors which you want to optimize your backlog are:
- metric impacted
- prediction of outcome
- probability of outcome
- resources required (man hours or team members involved, e.g. design, engineering etc.)
- costs required (marketing spent needed for significant data)
You’ll want to sort your list using data filters so you can prioritize by predictability, $ value added month and resources required.
Each organization should choose what is most important to them when prioritizing experiments.
Other things to consider
Foster a culture of autonomy, decision-making and learning by having your team fill out predictions and metrics for the experiments they are responsible for. This way, your team stays accountable for their prioritizations and can compare their predictions with actual outcomes which helps future decision-making.
Step #6: Experiment and implement
Take a maximum of 1 or 2 experiments to focus on and keep complete control of these.
Make sure you keep a log document where you note down all specifications of the experiment.
Here’s our Zengrowth doc template you can use:
In the end, you don’t want to create an experiment and in 6 months have forgotten what exactly you did that was so successful.
Step #7: Track and optimize
Logically, you’ll want to know if something worked or not. This comes down to each individual tracking system that is being used at your company.
If you’re optimizing for leads, then it’s usually easy to read this off from channel data or via your website form fills.
If you’re doing revenue tracking, you’ll need a more sophisticated solution which we can help you with.
Once you have that data at hand, you need to add it back to the experiment doc so you have a full record of what happened.
Step #8: Create playbooks and feedback loops
If something works well, do it for as long as you can until it gets outdated.
But most importantly, you want to use it for two reasons:
1) Turn it into a playbook containing all data and complete how-tos so the knowledge is retained within your organization.
2) Create a Feedback Loop to Sales and Customer Success so they have access to the experiment. This can provide valuable feedback and data so sales and customer success can optimize their operations based on the data that is gathered from marketing experiments.
Of course, we can go much more in-depth on each individual point, but I hope this helps you get started.
Written with 💙 by
Marco van Bree