Message Mining: Don’t Write Great Copy – Steal It!

I first heard about a technique called “review mining” some years back from Joanna Wiebe from Copyhackers. And it was a true revelation.

Today, there’s rarely a project I start that doesn’t use message mining under some shape or form. It’s such a versatile tactic that I apply it to general strategy building, positioning research, landing page copywriting, advertising copy… And it’s one of the tactics that continues to blow my students’ minds whenever I teach it. It drives similar insights to Jobs to be Done customer interviews but it’s quicker and can return results within a week.

So I thought I’ll finally pay it forward. I’ll share with you what message mining is, how to do it, and what you can use it for.

Review mining: how it all started

A couple of years ago, I was listening to a podcast episode with Joanna Wiebe on Everyone Hates Marketers. She was talking about all the things that go into effective copywriting – and the role research plays into all of it. Suddenly, she mentioned a technique called “review mining”. 

In short, review mining is a research tactic based on data that’s freely available online and that can help you understand your target audience better. The way it works is simple:

1. You think about the main pain point or customer job your target audience has.
2. You figure out what types of solutions exist out there to work for that pain point or satisfy the need.
3. You look for online reviews of all/some of these solutions and take note of what people are saying in a voice-of-customer (VOC) data file.
4. You use the same or similar copy on your next landing page or ad copy.

It’s a beautifully simple system that helps you understand what your target market wants and how they talk about their pains, needs, and the situation they are in. You can use that information to inform your actual copy. But you’ll also use it to understand what attributes and product benefits you need to focus on.

Well, I use “message mining” to explain an even broader approach that covers VOC data from other sources. But before we get to them, let’s see what you need to pay attention to when doing message mining.

What to look for in message mining

Since there’s a wide variety of things you can use message mining for, there’s also a lot of different things you can be looking for.

The key focus I usually have is two-fold. First off, I’m trying to understand the user’s situation and desires. What were people trying to get done or what was the scenario they were in. Who are they? What so-called “hiring criteria” did they have – i.e. what were the requirements they were not ready to compromise with? What benefits were they delighted to get out of using a certain solution?

This is the content layer of your research. It is not so focused on the words themselves but on the facts they depict.

Then we go into the messaging/copywriting layer of the research. This bit is all about what words people use to talk about their situation, their pains, and their desires. Here, you’ll be paying close attention to language and you’ll note down specific phrases.

At the end of it all, message mining will tell you what features and benefits of your solution are most important to people. You can basically create the structure of your sales page according to it. And you’ll also get ideas about the words you can put in your headlines, crossheads, and page copy.

Message mining sources

The original review mining idea talked only about scraping information off of review sites like Amazon but there are a ton of other sources of information you can use. Here’s a non-exhaustive list.

Review sites

Obviously, this is where you can always start. If you’re marketing a physical product, you can always use reviews from Amazon. If you’re marketing a services company, you can look at a specialized website. I’ve used Clutch for a bunch of developers and digital marketing companies. You can use Yelp for local businesses, Tripadvisor and Booking.com for travel clients, etc.

Keep in mind you don’t need to review direct competitors only. When you get free of this limitation, you can actually find a lot of insights on review sites. 

Here’s an example: I was working with a client who was selling online data science courses. We looked at their direct competitors, but I also did message mining for introductory data science books on Amazon. It works because this is a different type of solution to the same user problem.

Facebook and LinkedIn groups

Look at social media groups your target customer is likely to be in. If you’re marketing an e-commerce email software, a professional group for store owners can be your goldmine.

Pay attention to the questions people ask – these are often rich in insights about the pain points people experience and their particular use case. On Facebook, you can use the search bar to look for relevant key phrases and find older relevant posts.

The same logic applies to other online communities like forums – if your people gather there.

Twitter hashtags

The data on Twitter is much more fragmented and it rarely returns good content. But if your target customer is a techie who discusses everything in 280 characters, then it might be the way to go.

You can use Twitter search for regular keywords but I’ve found hashtag searches return more focused results.

Quora questions

Quora is a place where a lot of people ask a lot of questions – and for that reason, it can be a great source of information for your message mining.

You can easily search for topics related to your project and get opinions from live users.

One thing to pay attention to is who wrote the answer to a question. Quora is a popular place for marketers. That’s why I’ve often seen topics (especially ones related to software recommendations) where the main answers come form owners of competing products trying to push their solution. Make sure you screen these marketing messages out – or at least add them to your “competitor messaging” list of quotes.

Reddit subreddits

If you’re not familiar with Reddit think of it as the largest database of niche forums out there. Different topics live in different subreddits. You can find a subreddit for almost any topic you can think of – including a subreddit dedicated to the conspiracy theory that Jar Jar Bings from Star Wars is actually evil.

You can take a look at the topics discussed in your subreddit of choice and you can also search for more specific post topics within the subreddit. For example, if you’re working on a healthy meal subscription service, you can look at what people following the keto diet find difficult when it comes to meal prep.

Competitor spying

I don’t necessarily take competitor messaging as the end-all and be-all but it’s still good to pay attention to it. I’d usually add that information to a separate file or sheet because I don’t want to inadvertently copy what others are saying. That is a surefire way to create a bland and forgettable message.

You can strip the messaging from the following sources:

• your competitor’s website;
• their social media channels;
• their emails – register an account or subscribe for a free plan and you’ll get their onboarding sequence to review;
• their ads – go to their Facebook page and find the “page transparency” option. It will show you all the ads your competitor is currently running.

Now that you know what sources you can leverage, let’s look at the step by step of message mining.

Organize and analyze your audience’s messages

The process I’ll outline here is pretty simple and you can customize it depending on your project goals and personal preferences. One thing that I wouldn’t change, though, is working in a spreadsheet.

A lot of copywriters seem to be afraid of spreadsheets. I guess it’s our natural affinity towards writing long text that makes a word document look more inviting. But here you’ll be recording short keyphrases and you want to be able to slice and filter and compare. And to do that you need a spreadsheet.

Step 1: Get all the data into the same place

The first thing you need to do is do the actual mining. Roll your sleeves up and go through each resource.

“What am I actually looking for? How do I know what’s a message worth mining?” you might wonder. You’ll quickly get an intuitive understanding of the types of messages that are worth noting down. I would often start off by including even trivial quotes. As I go through more and more information I’ll gradually become pickier. Then I’ll only write down quotes that are particularly interesting or unexpected.

The next question is often “How do I know if I have enough? When do I stop?” I use something I like to call “the microwave popcorn rule”. You know how when microwaving popcorn, you stop the microwave if you don’t hear a pop for more than 3 seconds? It’s the same here. If you run through a bunch of online content and you don’t find anything new to add to your file, you don’t need to continue with message mining. I’ll generally try to have at least 100 lines of quotes in my spreadsheet by then.

Step 2: Split quotes into categories

Once you’ve covered enough ground, it’s time to go into categorizing your messages. You can also do this step as you record your quotes – the context will be fresher in your mind and it will take less time.

What categories will you use depends on the type of research you’re doing. I generally stick to five:

• user scenario – this includes nites about the specific situation a customer is in. For example “as a mother of three, I needed a course that I can follow at my own pace”, or “since we have an in-house designer, we needed a way to manage versioning of ad creatives in the social software…” This information is rarely used in your actual copy, but it can help you better understand your personas.
• user goals and jobs – this is the end goal the person is trying to achieve by using a specific solution. Usually, you can guess you’re looking at a job message if it uses a verb and you can rephrase it by starting with “I want to”.
• benefits – these are all the positive attributes of the solution the person is looking for or that they got once they started using a solution. They are more often formulated as things the solution had.
• pains and problems – these are problems the person encountered while using an existing solution or pains they had that made them look for a specific solution. This can be split into two separate categories. But I’ve seen that people rarely write about pains or problems and there’s no need to overcomplicate things.
• anxieties – this is the category you’ll probably get the least information for, but one that can influence your messaging a lot. It covers fears or doubts people have before they start using a product. Since these emotions can prevent people from trying out your brand altogether, it’s important to know about them and actively counteract them.

If you want to work with a simpler format, you can use the “Before state” and “Dream state” categories that Joanna Wiebe uses. This lets you split your data just into positive messages about things the user wants to achieve (the dream state”) and more negative messages about the problems, anxieties, and pains the user was subjected to at the beginning of their customer journey (the before state). This makes the current step simpler and will let you only work on creating topical groups at your next step.

Step 3: Split messages into topic groups

You probably saw while you were recording your messages that people were often saying the same thing with different words. Now it’s time to group all of these similar messages into topic groups.

Add a new column right next to the quotes themselves. You will then go down the list and assign a topic to each quote. In the beginning, you might be adding a lot of different topics but down the line, some patterns will emerge.

Once you’re done with the whole list, you can do a second pass and combine some topics that at first seemed different but are actually the same.

For example, I was doing message mining for a web development agency. I had a topic around time savings (getting a project done on or ahead of schedule) and cost accuracy (having a project that doesn’t end over budget). Down the line, I saw that the two categories often come up in the same sentence. It made sense. When it comes to development, the project cost is calculated based on a time estimation. So I ended up grouping the two under a parent topic of estimate accuracy and project predictability.

For a typical message mining project, I’d end up with something like 15-20 different topics, some with 10+ entries, some with 4-5.

Step 4: Quantitative review (but not really)

Before I describe this step in detail, let me reinforce the following: message mining is qualitative research. This means that we’re interested in the language used, not so much how often different topics come up.

Still, it’s enough data to see what topics are coming up more often than others. This doesn’t mean that a topic that was seen 8 times in message mining is reliably more important to users than one that was seen 7 times. But more often than not, it’s truly more important than a topic with 2 mentions.

To get your data straight, you can use message mining as your exploratory research phase. You can later run a survey with a larger set of respondents where you validate the importance of each topic.

What can you use your message mining findings for?

Any research approach is only as good as the actionable findings you apply to your business. So I really want to give you as many ideas as possible. Here are just a few:

General messaging

You can create a file that you always use when describing your brand as a whole and its key attributes and benefits. I usually do this at the start of copy work for a client and it saves us a ton of time later.

You might think that this will make your copy dull and repetitive but you need to keep in mind that you’ll be using only bits and pieces from it. Besides, your users don’t go through all your content in one sitting, do they? So they are unlikely to notice. On the contrary, this approach is more streamlined and persuasive.

Landing page copy

I almost always rely on message mining to set my page copy. And yes, I use words taken out from the messaging file exactly as users wrote them.

The more voice-of-customer data you can use in your copy, the more seen, heard and understood your average reader will be. They might feel like you’re reading their mind – but not in a creepy way.

Obviously, you won’t just string together a few phrases you read and call it a day. It’s up to you to create a structure based on the research but also working towards your goals. But here’s an example of a landing page we launched recently and how much of it is actually inspired by VOC data.

Advertising copy

Ads are another great place where you can use your message mining findings. Here, it’ll often be short phrases that you add to your copy since ad structure is a bit more rigid. But still, the main benefit you focus on can be derived from VOC data.

For example, a client of mine doing online software courses found out that what people want the most is to save time. We tested time-saving as a key message against improving one’s knowledge and becoming more hireable. And sure enough, the phrase customers “suggested” had a much better conversion.

Great artists steal

I’ve never thought of myself as a very creative writer. Ever since I found the message mining technique, I understood I don’t need to be one – I just need to rely on empathy and listen to what people are saying.

Message mining feels like cheating. After all, your boss or client pays you to write up copy that converts. But if your form of cheating is based on actively listening to your target market – then sure, cheat away!