Building a Data Strategy to Drive Better Decision Making
Data is everywhere. In the Mad Men days, the only data you had on your customers was whatever you learned over a three-martini lunch. Today, every interaction a customer has with any aspect of your company can generate data, from how much of your video they watched before getting bored to how many times they consider that one splurge before purchasing.
Data is meaningless—unless you have a purpose for gathering it and a plan for using it.
We believe in a tactical approach to data. Ask yourself: what’s the simplest, smallest data point I need in order to make a decision and move on?
The wealth of data available today, at relatively low costs, provides enormous opportunities to drive continuous improvement for your company. But you can’t just collect data for data’s sake. You need a data strategy.
There are 3 basic components to any data strategy:
1. Data collection and storage.
- What do we observe, what do we ask, and where do we store it all?
2. Data processing and analysis.
- How do we aggregate and calculate metrics, and what are those metrics?
3. Data reporting and utilization.
- How do we aggregate and calculate metrics, and what are those metrics?
It might sound logical to start with #1—collecting and storing data—and build from there. But that’s data for data’s sake. You’d end up with more “interesting data” than you could parse through in a lifetime.
To avoid data paralysis when your team is small and resources are scarce, start with #3: the use-case for your data. Start by thinking about the types of decisions you can only make with good data. Only when you know why you’re collecting data and what you’ll use it for will it be truly useful.
Of course, not all decisions run on data. Sometimes data only serves to reinforce a decision you were always going to make, or to report on what you’ve already done. It’s absolutely fine to make those kinds of decisions without trying to attach a data-point to it. Don’t let data paralysis stop you from making crucial calls!
Once you know how you’re going to use your data, you can build your data collection and processing systems strategically to generate the specific, actionable metrics that you need to make good decisions.
Understanding when and where data will be used will help you avoid situations where data becomes isolated and contained to side conversations. “Reviewing the data” should never be its own agenda item. Data is not your job. Data exists to help you do your job.
How to design your data strategy
Remember, your goal is to extract value from your data and let it help you make decisions. That means you should:
- Limit the breadth of data gathered to areas where you actually have a “lever”—a decision to make that can be influenced by data;
- Know at what level of complexity you need to analyze your data;
- Make sure all your reported and reviewed metrics are optimized for actionability.
One of the best ways to get a handle on actionability is to map out what kinds of decisions are happening, who is making them, and when, where, and how they’re made. Start with decisions made on a regular basis, and expand to categories of one-off decisions after that.
Let’s look at two examples.
Decision #1: How do I allocate next month’s marketing dollars?
First, walk through when, where, and how this decision needs to be made:
- Who decides: Chief Marketing Officer
- When: Monthly
- Where: Monthly Paid Marketing Review Meeting
- How: Review cost per action (CPA) across all channels and try to allocate the budget primarily to the best-performing CPAs, while maintaining a distributed presence.
Next, ask: is there any additional data that would help us make a better-informed decision?
Possible answers include:
- “I want to understand customer lifetime value (LTV) better so I can optimize for long-term ROI, not just acquisition.”
- “I’m not sure how to value indirect channels like unbranded ads, content marketing, banners, etc.”
- “I want to figure out the right message for the right channel, but I’m not sure how to test.”
- “Do some channels deliver customers with higher costs in terms of support and returns?”
Decision #2: How much should we discount our subscription offering compared to our normal prices?
Walk through how this decision gets made:
- Who decides: CXO
- When: Once upfront, re-evaluate quarterly
- Where: Quarterly Business Review
- How: Review percentage of purchases opting for subscription and evaluate against target.
Now ask: is there any additional data that would help us make a better-informed decision?
- “I’d like to see the rate of re-orders for our ad-hoc purchasers vs. subscription customers.”
- “What’s the average order amount for these two groups?”
- “Who’s buying more distinct products?”
- “How many subscribers drop out after their first refill/billing cycle?”
- “Do some channels deliver more ad-hoc purchasers who are willing to buy at a higher price point, versus bargain-seekers who value a subscription discount?”
Once you’ve walked through several decisions this way, you’ll generate a list of additional data needs. For example, using the two decisions above, you might find you want an alternative metric than CPA to measure both customer LTV and the value of different marketing channels. You might find you want to track customer behavior in more detail for both subscribers and ad-hoc purchasers.
The key is that you’re approaching data strategically—you’re using it as a tool to make better decisions, not as an end in itself.
The best data tech stack for startups
We recommend 3 tools as the gold standard for new data capture and storage. We’ll summarize them below; for more detail, click here.
1. Google Analytics
There are many reasons that an estimated 30-50 million websites use Google Analytics: It’s reliable. It can tell you how people found your website, where they’re coming from, what they looked at, and how long they lingered on different areas of your site. These are the key data points you need to generate buyer personas and start to optimize your marketing.
Google Analytics is even more powerful in combination with a tool like Intercom, which allows you to track all customer interactions with your help and chat functions. With those tools combined, you’ve got all the data you could need on who your customers are and what they want from your website.
Business intelligence software is an essential tool for any modern company. It gives you a comprehensive view of what your customers are doing at every point on your sales funnel. The best products combine data from multiple sources, so you don’t have to check on your AdWords campaign separately from your other marketing efforts—it’s all right there in one dashboard.
We like Looker for its top-notch data modeling. It makes connecting disparate datasets very easy, and the end user experience is intuitive and clean.
With oceans of data coming at you from all sides, you need a way to pull all that information together into one useful application. Segment is a customer data platform that collects data from multiple sources, including your customer relationship management (CRM) software and your web and mobile apps, and creates a unified view of your customer.
We like Segment because it integrates well with the other tools we recommend, and it makes data accessible and usable for every team in your organization.
What you really need to measure
Now that you’ve got data flowing into these excellent tools, what should you try to measure with it? Here are some of the most crucial KPIs your early-stage company needs to track.
If you’re subscription-based:
Some experts have argued that every business will be subscription-based in the future. We’re not sure subscriptions are the way forward for every business, but there are good reasons for this model’s increasing popularity. Namely, subscription customers are almost higher-value (justifying a higher cost per acquisition).
Early-stage companies should be experimenting with different pricing options, and subscriptions and bundles should definitely be part of that mix. We’ll share more about data-driven pricing in a future deep-dive tutorial. This forEntrepreneurs article is an excellent read about KPIs for cloud software companies, now nearly always sold on a subscription basis. We’ll highlight some key points below.
If you’re trying out a subscription model, you need to be tracking:
- The subscription cash flow trough. Acquiring customers costs money upfront—that creates your trough. As those customers start generating revenue, the trough turns into a cresting hill. But that’s often the moment when you should step on the gas and spend more on customer acquisition. That deepens your trough again - hopefully temporarily.
- Unit economics. Ultimately, your customers need to generate more revenue than it costs you to acquire them. You should keep an eye on two questions: First, is your customer lifetime value (LTV) at least three times your customer acquisition cost (CAC)? And second, does it take you 12 months or less to recover your CAC with revenues? The answer to both questions should be yes.
- Churn rate. This measures how many existing customers/revenue you lose versus how many new customers/revenue you acquire. Some churn is inevitable, but too much churn will limit your ability to grow.
- Cohorts. You need the ability to track customers in groups based on the month they signed up. That gives you visibility into how effective your churn-reduction tactics are, the value of customers who come in through different marketing campaigns, etc.
- MRR movements. Changes in your monthly recurring revenue (MRR) due to reactivations, expansions, new business, and churn.
- Average sale price. This establishes the first dollar value for a new customer. It’s crucial to track this figure because if you only look at a measure like average revenue per account, you’ll get a view skewed by customers who’ve upgraded over time. You need to know what your new customers bring in on day one.
How to turn positive churn into negative customer churn.
Basically, you need expanding revenue from your existing customers to exceed the lost revenue from churning customers. There are 2 ways to do this:
- Build some variability into your pricing scheme, so that when customers use your product more, they pay more—for additional seats used, leads tracked, etc.
- Upsell customers into more robust versions of your product, or cross-sell them additional modules.
For more on metrics subscription businesses should be tracking, check out this cheat sheet from ChartMogul.
If you’ve got an ad-hoc purchasing model:
Whether you’re all ad-hoc purchases or a mix of subscription and ad-hoc, you’ll need to track some additional KPIs specific to individual purchases, too.
If you’re a DTC business, you need to be tracking:
Average Order Value (AOV). Total revenue / number of orders. Driving up AOV will improve your margins.
- Strategies to increase AOV include: Making shipping free slightly above your AOV; offering discounts at different total thresholds; offering a gift with purchase.
Cart abandonment. You can calculate this percentage as follows: 1 - total number of complete purchases / total number of shopping carts created
- The number one reason customers abandon items in their cart is a lack of free shipping. Of course, sometimes people are just researching or browsing with no real intention to buy. Strategies to combat this problem include: Showing limited inventory to create urgency; automating abandoned cart emails and retargeting campaigns.
- Gross margin. (Revenue - Cost of Goods Sold) / Revenue. This is typically higher for DTC businesses than for B2B or subscription businesses. But it’s just as important to track…
- Contribution margin. (Revenue - Variable Costs) / Revenue. This number includes things like shipping costs, payment processing fees, and fulfillment costs, which tend to impact DTC businesses more.
- Conversion rate (CR). Number of Conversions / Number of Sessions. This measures the percentage of people who made a purchase (or signed up for your newsletter or otherwise entered your funnel) after visiting your website. Conversion rate offers a window into both how effective your website is and how attractive your value proposition is overall. For even better results, measure…
- CR x AOV. This basically gives you the dollar value of a session on your website. You can evaluate this number against how much it costs to drive a session through marketing. These numbers are key for optimizing your website and, ultimately, creating a test-and-learn culture that’s obsessed with the customer journey from first touchpoint to final purchase.
For more on metrics DTC businesses should be tracking, check out this post from AllyCommerce.
Beyond the basics: how do you know when to up-level your data game?
Here are 3 signs it’s time to expand and improve your data strategy:
- When your company has 20-30 employees who all need data access on a daily basis to make decisions.
- When data becomes integral to how your business functions, as with Netflix and its recommendation engine. Few businesses reach this level of data-intensive needs, but those that do rely on very specific tools. More on this in a minute.
- When you’re ready for a Series B fundraise.
Let’s take a closer look at the data needs of companies at the Series B stage and beyond. SeatGeek has a great deep dive into how they built out their data pipeline. We’ll highlight a few key points here.
Most of the data problems companies face at this stage fall into four buckets:
- The data isn’t there.
- The data is there, but it isn’t formatted well.
- The data is there, but it isn’t organized well.
- The data is too accessible (and we’ve got too many people poking around in it).
If you’ve built your data tech stack on the 3 gold-standard tools we recommended above - Google Analytics, Looker and Segment -- you’ll avoid problem #1, and you can focus on problems #2-4, which inevitably creep in as a company grows.
SeatGeek has some more tech stack recommendations for this stage, which we endorse:
- Luigi for data processing;
- Redshift for data storage;
- and Looker for data visualization and access.
At this stage, you need to separate out those 3 pillars as distinct responsibilities for your data engineers (or whoever’s responsible for your data quality and availability). Once you’ve reached the point that your business is thinking about its own data ETL (extract>transform>load), you’ve progressed beyond what off-the-shelf data platforms can offer, and you’re ready to take more ownership of your data in-house.
If data becomes integral to your business, as it is for companies like Netflix, Pinterest, Zulily, and Spotify, your needs will be much more intensive. Looker has a great article on the considerations these companies face as they build out world-class data tech stacks to drive original business solutions.
Common misconceptions about data strategy
M13 works with a lot of founders. Here are some of the most common misconceptions we see when it comes to working with data:
- “If I’m not looking at all the data, all the time, I’m missing something.” We often ask founders to tell us about the last 5 decisions they made—and how data could have helped them make a better decision. And yes, sometimes this exercise surfaces a data point that’s worth capturing and analyzing for future use. But a lot of times, the decisions they mention didn’t depend on data. You don’t have to look at data for every decision. Sometimes it’s OK to look at less data.
- “I have to automate all my data.” If you’ve selected the right KPIs to focus on, you usually don’t need to automate anything. Solid data analysis can be pretty low-fi. You can make most decisions with math you can do on a napkin.
- “Every department needs its own custom data stack.” For raw data capture, you need a single source of truth. Don’t build your data stack with silos. You want to avoid a situation where your CFO claims you acquired X customers last month, but your CMO’s data gives a different figure. The raw data stack we’ve recommended allows you to capture everything, so you can always build analysis on top of it or go hunting for a specific answer. But our data philosophy is capture everything, focus on almost nothing (at first).
Remember, data is meaningless on its own. Data analysis shouldn’t be seen as a super-wonky dive into the weeds. Rather, it should be a guided fact-finding mission. You’re going looking for a specific data point that will help you answer a specific question or make a specific decision.
If you don’t know why you need a piece of data, then you don’t need it. It’s so easy to capture data today that it’s easy to fall into the trap of overloading yourself with information. That’s why you need a data strategy that guides the design of your data tech stack.
Data exists to drive better decisions. Period.