We are always asked to explore user needs. How do we do that? Especially since there is not much data at hand, at most there is only one user's purchase record, and it feels like there is nothing to dig out. Today I will give you a systematic answer. When it comes to exploring user needs, there are many popular unanswered questions, which we will clarify today.

1. Wrong practices in user demand mining

The story sounds good, but it is completely wrong... From a business perspective, if a hardware store owner does not think about how to sell metal tools, but instead studies how to make connections, then the store will not be far from bankruptcy. Judging from the data, if he wants to find a girlfriend, he probably wouldn’t even bother to tell his aunts and uncles, let alone tell strangers easily (besides, he is a steel bar seller).

This is a common mistake: people mistakenly believe that in order to explore user needs, they must dig up gossip and anecdotes that others don’t know about in order to be considered in-depth, and that they must satisfy very deep needs in order to be considered real needs.

In fact, only a very few industries can understand users so deeply and satisfy them infinitely. For example, in the financial industry, private services for extremely high-end clients may be able to achieve this (it is not uncommon for a branch manager to personally drive the son of a major client to school).

However, most companies have limited business scope and face massive numbers of users. Therefore, we cannot divorce ourselves from business realities and do too detailed and in-depth exploration. Whether in terms of business or data, it is impossible to do so, and there is no need to do so.

Therefore, the essence of user demand mining is to screen key distinguishing dimensions from limited data and increase the probability of user response. What we need to do is not to figure out the needs of every user at every level. Instead, it increases the probability of user response and identifies core user groups through differentiation. The user response rate to our business will be higher than doing it blindly. Every percentage point higher is the data analyst's contribution to the company.

2. Five steps to discover user needs

▌ Step 1: Identify core users

Let’s take the hardware store owner as an example. When energy is limited, the key is to capture the big customers first, and classification is necessary. The user categories of a hardware store might be:

First class: property maintenance department, decoration team, construction site (B2B customers)

Second category: Customers who want to renovate, change water and electricity, or do repairs (large B2C customers)

The third category: Individual customers who occasionally buy a light bulb, socket, or nail (small B2C customers)

The problem is that when a guy walks in, the hardware store owner has no idea which category he belongs to. If left unchecked, you could lose a big business. But if everyone comes up and asks a bunch of questions, it will probably scare the guests away. Here begins the first step of exploring user needs. The question being explored is very simple: “What do you want to buy?”

▌ Step 2: Classify the business

The guy replied: "I want to buy nails" - what do you think of? This answer sounds simple, but it reveals a lot of information. Because each type of business may have a fixed product mix and consumption characteristics. For example, for a hardware store:

● Engineering business: large quantities of steel bars and various materials (not purchased in small quantities)

● Water modification: water pipes, wrenches, waterproof tape

● Electrical renovation: wires, switches, sockets

● Wall repair: cement, brush, paint

● Object repair: nails, hammers, drills

This is called: strong business relevance. Even without doing correlation analysis, these products are naturally bundled. And depending on the size of the business, there is a fixed consumption amount. It is very important to do a good job of prior business classification. When we are unable to collect a large amount of user information, we can use the few purchase records and business relevance to infer user needs.

For example, when the boss hears that the guy needs nails, he can quickly infer that he is not a Class B user and his business is related to maintenance. However, the boss still doesn’t know whether the guy is a Class C big customer or an individual customer, and a second step is needed to find out. The question is also very simple: "What do you buy the nails for?"

▌ Step 3: Grasp key information

The guy replied: "I want to buy some nails to nail a picture on the wall" - when you heard this, did you immediately think of what you wanted to say! Yes, we can see that after completing user segmentation and business classification, it is very easy to conduct demand mining.

Based on the previous classification, when readers hear about nailing a painting, they can immediately react: this is a single customer and the value is not high. Nails, hammers and drills are highly related and there are cross-selling opportunities. Here, with the help of two simple questions, we have captured the key information.

Of course, in actual business, traditional companies rely on sales, shopping guides, and salesmen to capture key information, while Internet companies rely on embedding, push/feedback, questionnaires, browsing frequency, etc. to capture key information.

▌ Step 4: Push products/activities

Now that we have a hypothesis, we can try to verify it by launching a product/activity as a test. At this time, the hardware store owner would not spend a lot of effort asking the young man if he wanted to fall in love. Instead, he would say, "If you need to nail a painting, a 1-inch small nail would look better than a 3-inch large nail. It's easier to nail and less conspicuous." In this way, he can lock in the young man's needs and have a higher chance of success than those bosses who ignore him.

At the same time, you can also make a cross-recommendation: "Do you have a hammer? You can buy a small drill, which is more convenient than a hammer and can also be used to repair other things." If the recommendation is successful, you can successfully increase the average customer price from 1 yuan to 200 yuan, which is also a small profit.

▌ Step 5: Verify the push effect

With push, there are two possibilities: success and failure, so the effect needs to be verified. Demand mining is essentially a probability problem. We need to verify our push through data, and then verify whether the mining dimensions and mining directions we choose are correct. For hardware store owners, there are two dimensions to verify:

1. 1-inch nails are recommended for nailing walls (Assumption: Based on user needs, it is easier to close a deal)

2. For men who want to nail a wall, we recommend a pneumatic drill (assumption: men like machinery and have a chance of success)

This is actually a small AB test. If there is a data to record, the boss will see that these two assumptions may be true or false.

For example, after making 200 sets, we find that users don’t care about aesthetics at all and just buy whatever is cheap. Then our future strategy will be to just give the cheapest things to individual customers when they come.

Of course, it is also possible that this strategy is found to be feasible and 3 drilling rigs can be sold cross-sellingly out of 10 orders. Then we will follow this strategy from now on. This is the end of our demand exploration. We found a differentiating direction and verified an opportunity point to increase transactions, and discovered the demand for drilling rigs from users buying nails. Doing this is much more reliable than wondering every day whether the boy has a girlfriend, and whether he likes loli or mature women.

Just a funny example though. (In fact, hardware store owners don’t have this patience, and hardware stores have no data to record). But it vividly shows the workflow of mining user needs:

This methodology can be extended to various industries, especially when there are fewer data records. Please note that there are industry differences in whether to distinguish users or businesses first.

Generally, the business types of traditional enterprises are relatively fixed, and they tend to differentiate their businesses first. Internet companies’ businesses are relatively flexible and can even create new scenarios out of nothing. They tend to differentiate users first and may even target different scenarios for one user.

But no matter what you do, distinguishing between users and businesses is the first step and the most important one. Classification can clarify the direction of subsequent excavation, determine the excavation depth, and provide standards for verifying whether the excavation is useful. Therefore, this step will be discussed separately below. Many students have no idea how to explore user needs because of the lack of classification. Many students get stuck in Abtest and lack overall judgment because of the lack of classification.

3. Notes on user/business differentiation

When it comes to user segmentation, many articles talk about RFM, which is very wrong. Not all businesses require high-frequency consumption, and not all businesses accumulate high amounts. It is even possible that a business has both one-time consumption and high-frequency consumption. From the perspective of frequency and amount, common businesses can be summarized as follows:

The business of traditional enterprises is relatively focused and business classification is relatively easy.

For example, houses can be divided into property purchase and investment. Home purchase is further divided into first time purchase, secondary improvement and retirement purchase. Secondary improvements include area improvement, environmental improvement, supporting facilities improvement, resource improvement, etc.

Home improvement, automobile, loan and other businesses all have similar classification methods (too much text, so I won’t expand on it for now), and each corresponding user need will be very focused. Therefore, traditional enterprises’ user demand mining does not rely so much on “big data”. It is more like a hardware store owner, who does a good job of classifying the business and collecting key information from front-end sales, shopping guides, and salesmen.

Internet companies need to pay special attention to the following: one platform may integrate multiple businesses at the same time. These businesses may seem similar, but their corresponding user needs and related businesses may be completely different.

As shown in the red circle in the above picture, for business travelers, a ticket booking platform may be a high-frequency and high-amount event that occurs frequently. At this time, RFM can be used to further segment it.

But for honeymoon tours, it may be a very low-frequency demand, and the related businesses they are looking for are hotels, car rentals, and leisure places after returning (after going abroad for more than ten or twenty days, it is really tiring and they need to make up for the vacation). Similarly, e-commerce platforms sell snacks, mobile phones, recharge cards, televisions, etc. When exploring demand, they must also distinguish between the common ones, rather than just throwing everything together.

4. Notes on Push/Verification

Students who work with data often do a lot of AB testing, but they do it very passively. Often the business has the plan, but the data is just manipulated mechanically. Poor ability to propose hypotheses and verify them on one's own. The key here is: making assumptions. Many students have no idea about transaction data, and there are too few comments, demands, and browsing data in the database. Here is a simple example. For example, if we see a shopping list, we can make bold assumptions:

So you see, you don’t need a lot of data to come up with hypotheses. Of course, not all hypotheses need to be put into AB testing. We can first distinguish them based on the data.

For example, when discovering a hypothesis point from a user, first look at: whether the user has strong characteristics. For example, if we assume that he is driven by discounts, then he participates in discount orders > n times, the discount intensity > 50%, and the activity participation rate > X%. In short, he must really show a special interest in discounts.

Then, see whether there are enough users with similar characteristics. If the number of users is too small, then even if it is an opportunity point, it may not be used by the business. If the above two points are met, you can consider making suggestions and let the business make a plan and run Abtest.

5. How deep should we dig into the needs?

Seeing the above, some students may ask: Since there are so many directions to dig, where should we start? A: Start with the most pressing issues in current business development. Business needs:

● Improve conversion rate: Explore products that users buy for the first time

● Increase the average order value: tap into users’ cross-category needs

● Increase transaction amount: attract heavy users

● Increase repurchase rate: tap into secondary purchase demand

● ……

It is easier to find answers when there is a clear goal to guide you. Of course, it is also possible that after a lot of digging, you find that there is nothing to gain and you cannot find any opportunities in the data. But at least it can also prove the opposite: spending money on fancy marketing is useless, and it can also guide operations to do some cost-saving work, which is also a credit.

The above is the basic idea of ​​exploring user needs. As you can see, it integrates user segmentation, hypothesis testing, ABTest and other specific tasks. It is a highly comprehensive matter. At the same time, you can also see that it is not achieved overnight, but requires a lot of basic work and a lot of attempts to reach a conclusion.

Exploring user needs is not like being like a fortune teller at a roadside stall who knows everything by just throwing a coin. To sift the dross and retain the essence, to eliminate the false and retain the true, to iterate repeatedly and get closer to the truth, this is where the value of a data analyst lies.

Author: Down-to-earth Academy

Source: Down-to-earth School (gh_ff21afe83da7)