6 common mistakes made by smart people when it comes to data analysis!

6 common mistakes smart people make when it comes to data analysis :

After helping a large number of companies sort out their data analysis processes, Porterfield summarized six common detours that companies take.

1. Walking too fast and not having time to look back.

People in startups seem to be constantly being told: "Go fast or die, go fast or die." They are so eager to develop products that they often don't think about the specific details of how users use the product, how the product is used in what scenarios, which parts of the product are used, and what are the main reasons why users come back for a second use. These questions are difficult to answer without data.

2. You're not recording enough data.

It’s no use just presenting summarized data to your team. Without detailed changes down to the day or even the hour, you cannot analyze the invisible hand behind the data changes. If there are only rough and intermittent statistics, no one can interpret the impact of various subtle factors on sales or user usage habits.

At the same time, data storage is getting cheaper. Doing a lot of analysis at the same time is not a high-risk thing. As long as you buy enough space, there will be no risk of system breakdown. Therefore, it is never a bad thing to record as much data as possible.

Don’t be afraid of quantity. For startups, big data is still a relatively rare thing. If you are in the startup phase and you are (fortunately) facing this problem, Porterfield recommends using a platform called Hadoop.

3. In fact, your team members often feel like they are groping in the dark.

Many companies think they can just throw their data into Mixpanel, Kissmetrics, or Google Analytics, but they often overlook which members of the team can actually interpret the underlying meaning of that data. You need to frequently remind every member of the team to understand the data more and make decisions based more on the data. Otherwise, your product team will just blindly develop products and hope to hit the hot spots, and will be confused whether they succeed or fail in the end.

For example. One day you decide to use viral marketing methods commonly used in the market to attract new users. As you wish, the number of users has increased dramatically. But at this time you will encounter new confusion: you cannot measure the impact of this marketing method on old users. People may be attracted, sign up as new users, then get bored and stop using. You may be paying too much for attracting a bunch of worthless users. And your product team may still be complacent, thinking that this marketing method that damages the product is successful.

This silly mistake happens all the time. However, if your company builds a data platform that everyone can use by themselves from the beginning to answer the most important questions in their work, you can avoid the tragedy mentioned above.

4. Storing data in inappropriate places.

Let's take a look at a correct demonstration first. Porerfield mentioned that one of his clients integrated NoSQL, Redshift, Kitnesis and Looker resources to create a data analysis framework. This framework can not only capture and store its own data at a high level, but also withstand millions of clicks per month, and allow everyone to query the data they want. This system can even allow novice users who do not understand SQL language to truly understand the meaning of data. And in the world of data analysis, basically if you don’t know SQL, you’re screwed. If you always have to wait for engineers to run the data, you will put yourself in trouble. The algorithms created or software purchased by engineers without understanding the requirements are often a torment for users, because their use of data is often not on the same level as the former.

You need to keep all your data in one place. This is the most critical principle.

Let’s go back to the hypothetical company mentioned above. They did one viral campaign after another, but they didn’t put the user campaign data into the same framework, so they couldn’t analyze how one campaign connected to another. They also cannot perform a comparative analysis of data across daily operations and during events.

Many companies send data to outsourcers for storage and then just sit back and do nothing. However, these data often become other forms when they reach the hands of outsourcing companies, and converting them back requires a lot of processes. These data are often related to your website or product during certain promotional activities. Combined with daily operational data, you can explore which activities lead to user conversions. It is crucial to combine daily operational data to analyze user usage history. But shockingly, despite the critical importance of all operational data at any given time, many companies still do not bother to capture and record it. About half of the companies Porterfield has seen separate day-to-day operational data from activity data. This seriously hinders the company's correct understanding and decision-making.

5. Short-sighted.

Any good data analysis framework must meet the needs of long-term use from the beginning of its design. Granted, you can always tweak your frame. But the more data you accumulate, the greater the cost of making adjustments. And often after making adjustments, you need to record both the old and new systems to ensure that data is not lost.

Therefore, it is best to design the framework on the first day. One of the simplest, crude and effective methods is to put all available data on the same scalable platform. There is no need to waste time choosing the best solution. Just make sure that the platform can accommodate all the data that may be used in the future and can run across platforms. Generally speaking, such an original platform can last at least one to two years.

6. Over-summarizing

Although this problem is more common for companies with big data analysis teams, startups are also advised to avoid it. Just think about it, how many companies only record the average sales per minute instead of the specific sales amount per minute? In the past, due to limited computing power, we could only summarize massive amounts of data into a few points. But at present, the amount of computing power is not a problem at all, and everyone can record operational data accurately to the minute. And these precise records can tell you a ton of information, such as why conversion rates are rising or falling.

People often indulge themselves in making a few beautiful icons or PPTs. These summary statements may seem exciting, but we should not make decisions based on these superficial summaries because these beautiful summary statements do not truly reflect the essence of the problem. Instead, we should pay more attention to outliers.

The 3 Easiest Ways to Avoid These Mistakes

Three simple protective measures to help you avoid detours

Making fewer mistakes is more important than you think, because once a mistake occurs, it is easy to spend a lot of engineering time and resources to correct it. If you’re not careful, your engineers could be spending expensive time decoding data for your sales team, potentially missing out on countless valuable marketing opportunities. Whenever data becomes hard to use or understand, your team’s decision-making speed slows down, and your business progress suffers.

The good news is that if you adopt the following three simple protective measures from the beginning of your user life, you can definitely avoid many detours.

1. Appoint a Chief Business Data Engineer

If you can find an engineer on your team who is genuinely interested in data analysis, you can put him in charge of recording and managing all the data. This will save the entire team a ton of time. Porterfield shared that at Looker, a business data lead engineer is responsible for writing scripts that can record all data so that everyone can always get the information they need in the same database. It turns out that this is a simple and effective method that greatly improves the team's work efficiency.

2. Put data on an open platform

Porterfield strongly recommends that you use an open source platform like Snowplow to record all product-related activity event data in real time. It is easy to use, has good technical support, and can be used in large quantities. And the best part is, it plays nicely with the rest of your data structures.

3. Move your data to AWS Redshift or other massively parallel processing (MPP) databases as soon as possible

For early-stage companies, cloud-based MPPs like Redshift are often the best choice. Because they are cheap, easy to deploy and manage, and highly scalable. Ideally, you would want to write data about your events and operations to Amazon Redshift from the very beginning of your company's history. “The benefit of using Redshift is that the platform is cheap, fast and accessible,” Porterfield said. And, for those who already use AWS services, it (using redshift) can be seamlessly integrated into your existing architecture. You can easily build a data channel to transfer data directly into this system for analysis and processing. “Redshift gives you the flexibility to write massive amounts of granular data without being charged based on hard-to-estimate parameters like how many events are triggered,” he said. "Other services charge you based on how many events you store, so as more and more people use your product, more and more operational data will be recorded, which will cause the final fee to rise like a rocket."

How to use data analysis to seize market opportunities?

The value of data analysis depends on how it can help you seize market opportunities. As a startup, all of this data should be used to inform the goals you set for the different stages of your company.

For example. A courier company will usually measure the average delivery time for each shipment. This may seem like critical data, but it is meaningless without sufficient context (after all, the recipient could be just a block away or hundreds of kilometers away). On the other hand, average delivery time is not as important as the overall satisfaction of the recipient. Therefore, you must ensure that your analysis includes the correct data.

Please list and quantify the results you need: What do you want your customer experience to be like? Some common success data analyses are based on sales or user conversion rates (i.e., if a customer does X things, they will purchase or become a user), the time required for conversion, and the percentage of customers who have a negative experience. You'll want the first ratio to be high, and the latter two to be lower.

Typically, media sites measure performance solely by page views. But now they are also starting to pay attention to an indicator called "attention duration": how long people focus on a certain page, whether they pay attention to certain words, whether they scroll up and down the page, whether they watch videos, and so on. They don't just want to see how long users stay on a page, they need to know which parts of the page users are attracted to and how much time they spend actively and attentively browsing. This can help media sites design new headlines, page designs, and content selections to extend such attention spans. This way, they can innovate the way they design websites to better impress their audience.

Another key focus is monitoring retained users. Successful data analysis can cover both daily operational data and activity data, and conduct horizontal analysis. If you only look at daily operational data, you can tell which people will return to your website and which ones can achieve repeat purchases. But you also need to understand who are the people who return to your site but don’t repurchase: Why are they reluctant to buy again? Questions like these can be answered through analysis of operational and activity data. Activity data will tell you which customers who did not make a purchase browsed the site in what order, what they noticed, what they clicked, and what they did before leaving the site. As you track this path, you can understand how to modify this behavior to increase the likelihood that they will purchase on their next visit.

To design the data basket that works best for you, here are three suggestions:

1. Find a suitable type of user behavior
2. Calculate what percentage of the audience will have this type of user behavior
3. Test whether this type of user behavior contains important information

Sometimes, inventing a new basket of data records can lead to big changes for a company.

Take Venmo as an example. For a while, the company's payment app team heard that many users who wanted to request money from friends accidentally paid the money to their friends instead, because the "Request Money" and "Pay Money" buttons were placed together and it was easy to press the wrong button. However, the company does not know how widespread the problem is or whether it warrants a redesign of the user interface. To make better decisions, they designed a new data system to detect how common this request/payment error is. They found all the strange payment behaviors such as "A paid B and soon after B paid A double the amount." The results show that this happens frequently. So in the next product update, they fixed this issue.

Make your data shareable.

Often the biggest culprit preventing teams from easily sharing data is the definition of the data. Therefore, it is best to fully and completely define your data from the beginning. Consider creating a central glossary wiki page to make it easier for everyone to understand. Porterfield points out that people like to use strange words to describe data. For example, the word "Ratio" is often abused because people often fail to make the numerator and denominator clear when naming.

Data is the lifeblood of most successful companies. Good data sharing can not only increase the company's transparency, but also enhance collaboration between different departments. For example, in many companies, different departments often hire engineers to generate different data to answer the same question. If there is a good data sharing platform, such waste of time and energy can be avoided.

In addition, visualizing data is also something a good platform can easily do. Visualizing granular data as charts allows every member of the team to better interpret the data. For most people, it is much easier to understand a graph than a table, so visualizing the data can help communication flow more smoothly.

A bad data analysis framework will only undermine people's self-confidence. It will invisibly divide the company into two factions: the experts who understand data and the idiots who don’t understand data. This is a common and dangerous mistake. You have to make it easy for even the most ignorant data user in your company to generate the charts they need and understand them. This is a basic principle for choosing a data platform.

Poterfield concluded: Good data analysis can help people go into meetings more prepared, help sales teams ask more relevant questions, and eliminate unnecessary guesswork. People no longer have to guess what their users are looking for, or why they close a sale, or why they don’t come back. People no longer have to guess what colleagues on other teams know or don’t know. And all this is thanks to designing the data framework well from the beginning.

APP Top Promotion (www.opp2.com) is the top mobile APP promotion platform in China, focusing on mobile APP promotion operation methods, experience and skills, channel ASO optimization ranking, and sharing APP marketing information. Welcome to follow the official WeChat public account: appganhuo

[Scan the top APP promotion WeChat QR code to get more dry goods and explosive materials]

This article was compiled and published by (APP Top Promotion) by @董老师. Reprinting this article requires the consent of Top Promotion , and please attach the link to this article!