Data Operations: How to do data analysis for market growth!

Written at the beginning:

This article is translated from the foreign marketing website Ladder. The most important reason for choosing to translate this article is that for most people, even those who have worked for a few years, have not fully understood the meaning and value of data analysis for market growth.

Even though many people in the market have shared related articles, their perspectives are limited, which prevents people from directly seeing the overall picture. Most importantly, they are not concise and easy to understand. However, this article has a comprehensive perspective and is easy to understand, so it is worth reading in detail.

01. “Tell me how the marketing side of things is going.”

As a grassroots marketing manager, I once received a late-night call from our CEO, asking a seemingly simple question.

“Well, Google ads are good, and we’re still getting a lot of users from ‘organic search.’ We, uh, wanted to test Facebook ads… They seem to be working. I mean, it’s getting a lot of clicks, but our revenue per user is down, and I think that’s probably due to the Facebook traffic.”

Pro tip: If you want to keep your job, don’t make a habit of giving your CEO vague answers.

A better answer would be:

“Google AdWords and organic search bring in 80% of our new users, but year-over-year growth is flat, so we’re exploring Facebook ads as a new channel. We’ve spent $600 so far, and we’re acquiring new users at half the price of our $2.50 CPA target. That means these users are also buying half as much, so they’re getting the same ROI.”

marvelous. How do you do that now?

How do you look at the data and calculate it? How do you know what is important?

It starts with a good tracking and tagging setup, but assuming that part is done, how do you make sense of large amounts of user data?

Market analysis. Specifically: Marketing funnel analysis.

02Marketing Analysis Process

Last year, I gave a presentation at GrandCentralTech on market analysis, for which I created the following chart. They’re a lot simpler (and prettier) than what you see when you log into Google Analytics, so hopefully they’ll be less intimidating and easier to understand.

Suppose you have a mobile app with 10,000 users.

Not all users should be treated equally; some are new to us, some are returning users, and a small number may have already purchased something to become our customers.

Let's separate them:

What we have now is a basic marketing funnel. But what does it tell us?

Well, if it looks like the one above, one thing is for sure we are getting a lot of new users, but not a lot of them are staying or becoming customers.

Knowing this, we now want to focus on improving our conversion and retention rates.

This is an important step in marketing analytics: simply by breaking down the data by stage in the buying cycle, we are already able to use this data to determine our strategy.

If we had the opposite situation — a large customer base but few new users — we’d be much happier. This shows that we have a strong product offering.

However, this information begs the question:

If the product is so strong, why aren't we investing more in growth? Are we fully capable of attracting new users and completing conversions by increasing our marketing budget?

But before we start investing, we want to be really sure that we can earn back our investment. Let's break down our customer base a little bit more.

This is great, the vast majority are multiple purchases. It’s a marketer’s dream.

However, it's more like a nightmare; they buy it once and never come back for more. This insight may cause us to revisit our product to address any issues before doubling down on the market.

Of course, for most businesses, there is one more step before buying anything - registration. Now that we have 1,000 email addresses from people who didn’t buy anything, maybe we should set up a CRM campaign to drive more sales?

CRM is a no-brainer at this point; drop everything else and think about how to make money from your list of 4,000 email addresses.

Now we have segmented our data horizontally (i.e. the marketing funnel). What other ways are there to look at this data?

By segmenting our data vertically by marketing channel, we immediately saw how heavily we relied on Facebook ads for customer acquisition: it drove over 70% of our new users. It also tells us how insignificant Google Adwords or the App Store are in comparison.

Now, if in your experience the App Store is bringing in a higher percentage of your users than normal, you might want to focus on that channel. If 7% of new users is a normal contribution to the App Store, you will feel good because you know you are doing well in this channel.

What, you want all channels to have the same conversions? This is basically impossible to happen. In this case, although the App Store only brought in half of Twitter’s new users, the conversion rate of returning users was very high, ultimately becoming comparable to Twitter.

Likewise, while Facebook is still king, the number of returning users has dropped significantly, suggesting that the quality of its users is getting worse.

However, it’s only getting to the end of the funnel that’s truly revealing. Despite bringing in new users, low conversion rates mean Facebook barely brings in more users than Twitter and the App Store. Meanwhile Google Ads did not bring in any customers.

Of course, at this stage, it is important to introduce the concept of "marketing expenditure". If you invested thousands of dollars in Facebook, you would probably lose money on those 354 users. If Twitter is relatively cheap, you might consider shifting some of your budget to that channel.

Based on my experience with marketing analytics, it’s also normal to see one channel performing better than the others. That’s why it’s so important to do a thorough funnel analysis — so you know which channels to shut down and where to redirect your budget.

In addition to segmenting by channel, you should also segment by time. If each of the funnels mentioned above represented one month’s worth of data, then this business would look dead in a matter of months.

This is not bad growth. Not only is it trending upward, but all funnel stages are growing, which looks good and means that this growth is likely to be sustainable over a longer period of time.

This was even better; our customer base was growing exponentially over time; a strong signal that we could start increasing our investment in marketing.

But if we are still far from the goal, it is nothing. Failure to meet targets could mean failure to raise a new round of investment, layoffs, or even bankruptcy.

In this case, we would be very happy; it looks like we can easily meet our goals on time as predicted.

Note: Past performance is not always a reliable indicator of what will happen next. Maybe we took our foot off the accelerator because we got too comfortable? Maybe a major competitor started poaching customers, or maybe there were huge quality issues with our product?

This is a reminder that data is not magic. The graph will not go up or to the right as you expect. It’s just an abstraction of real people’s activity, and it’s important to understand what can be measured outside of an analytics platform.

It's useful to remember that even though we're looking at simplified charts here, market analysis is messy and patterns are hard to find. Don’t get discouraged when you don’t get a smooth graph, most real data I see looks like the graph above. Sometimes interpreting data is as much an art as it is a science.

Despite all the irregularities we observe in the data, we still have to make our best guesses about what is going on, otherwise we can't run this business. Look at the example above, where I depict what the “Pirate Metrics” (acronym AAARRR) might look like for a business.

Once you have your conversion rate and the final revenue amount your customers paid you on average, it’s easy to calculate the “Cost” column. Simply multiply the cost of the previous stage by the conversion rate and work backwards. That is, if one registration is worth $13.33 to us, and 5% of our visitors register, then one visitor is worth 67 cents (13.33×0.05=0.6665).

In the real world, inadequate data tracking solutions mean that getting accurate funnel numbers is nearly impossible. However, if you don’t at least have a hypothesis in place (or data that’s as close to real as possible), you’ll be plagued by funnel analysis paralysis.

For example: How do you decide whether to run a brand awareness campaign if you don’t know how much an “impression” is worth to you? Or, if you’re not sure how much you can afford to pay per click, how do you know if a new channel is viable? If you don’t know the conversion rate between each stage, how do you know which part is the weakest so you can focus your efforts?

Take a look at your funnel analysis results. What statistics surprised you? Which data did not meet your expectations? Only by continually segmenting and interpreting your data can you build a good mental model to predict what will happen. Once you have this model, you will be able to make the right decisions without even having to crunch the numbers.

Making the right decisions based on mental models built on solid assumptions and in-depth data is what market analysis and growth strategy are really about.

Insight, Context, Action

At Ladder, when conducting funnel analysis, we developed a framework called ICA, or Insight, Context, Action. Here’s how it works:

Insight : “My cost per click on Google Ads is 80 cents”

Context : “On average, a click costs us over $2.50.”

Action : “It’s time to increase our spending on Google Ads.”

Without context, insight is worthless.

Knowing my Google cost per click means nothing to me if I don’t know how much a single click is worth to me. Even knowing this doesn't help me take action.

If context doesn't lead to action, it's useless.

Quoting from SEO agency Distilled,

“If we don’t take action to improve our marketing campaigns, then any analysis we just did is largely wasted.”

Benchmarking

An important source of context is benchmarking.

Knowing that the global average cost per click on Facebook is $0.48 will tell you whether your campaign is successful or if there is room for improvement.

Knowing that exit capture popups convert 1.2% of your visitors will help you calculate how quickly you can increase your website traffic if you leverage your email list.

Knowing that 57% of marketing experiments fail will help you set realistic expectations for the success of your testing program.

You should take these industry benchmarks with a pinch of salt; many of them are simply averages for industries that may be completely different from your business.

I’ve seen campaigns with a $30 CPC that had a good ROI and campaigns with a CPC of less than 12 cents that lost money. It all depends on your business model, you may have strengths in one part of the funnel that make up for weaknesses in another part.

It's up to you to decide whether the benchmark is relevant.

The ability to set good benchmarks is one of the main advantages senior marketers have over junior marketers. They’ve seen this before, so they know what’s possible and what’s likely.

At Ladder, we’re working to level the playing field: Check out our Strategy Playbook, where we share over 800 proven growth strategies so you know what’s possible, and each with an expected ROI so you know what outcomes are likely.

Of course, the best benchmark is against yourself: this is where market segmentation comes in.

Segment

Segmenting your data simply means that you break it down into one or more variables and then see how each “segment” compares to each other.

You can segment your data by market channel to see what’s working and what’s not. Then it will tell you how to arrange your budget.

Segmenting your data by time is another powerful way to segment your data. Are you doing better marketing this year compared to last year? This month or last month? This week and last week? Understanding the situation can tell you whether you can sit back and breathe a sigh of relief or go into panic mode.

Other popular sections are audience (do teachers or stay-at-home moms buy more?), location (do New Yorkers spend more than Texans?), product (is the increase in sales due to raincoats or hiking boots?), and employee (is Sally outselling Jane?).

The simplest form of segmentation is A/B testing. You run the test with one variable and another at the same time to see which one works. Does this new ad copy perform better than your existing ad copy? Does this new landing page perform better than your existing landing page?

As long as you’re consistently outperforming old marketing campaigns, you’re growing.

Ultimately, this is your job.

Not to produce reports or shiny charts, but to drive growth.

Source: Yan Zhou