Related Magazines Related Associations Related Books
hot key words: Drying equipment flash dryer spray dryer sludge dryer fluidized fluid bed belt dryer paddle dryer disc dryer vibrating screen centrifugal pump disperser
Your present location: Home >> News Channel >> Success Story >> Case: New product introduction, how much is the first order?

Case: New product introduction, how much is the first order?

Source: Shanghai Xianyun Automation Technology Co., Ltd. Release time: 2019/12/11 15:10:45 Hits: 431

The case enterprises have a strong fan group. Traditionally, the fan economy has been adopted. Although the brand economy is transitioning, the fan economy is still an important part of revenue. In order to maximize fan income, case companies continue to launch new products, basically a new product is launched every week.

The company follows a mid-to-high-end, differentiated line, with many varieties and small batches. The first batch of launches generally ranges from several hundred to several thousand. A variety of small challenges are that forecasting is difficult to make, either surplus or shortage, which is fully reflected in case companies.

No, there is a new product launched in April, sold out three or four thousand in a few days, and quickly returned to the order. When the second batch arrives, it is already June, the "hot" brought by the explosion It is almost dissipated, and it has to be warmed up again. If there are explosive models, there will naturally be slow sales. For example, there is a product. In order to reduce the cost and quantity, the first batch of thousands of items was purchased. As a result, only a few hundred were sold, and huge resources were not used for sales promotion. Very high.

Demand uncertainty is high, and the first batch of orders has a high risk, which puts great pressure on demand forecasting. Prediction risk is high, and no one is willing to take it. In the end, it can only be the boss's decision.

The boss makes the prediction, and naturally the reason for the boss to make the prediction: he is the * experienced, * authoritative person, and * can bear the consequences of the failure of the prediction. However, as the business doubles continuously, more and more new products, more and more things, the boss is getting busier, getting farther away from consumers and front-line operations, and less and less time can be spent on demand forecasting. Many times, you can only take a picture of your head to make predictions, and the disadvantages are becoming more and more obvious. "I hope that the big boss understands the difference between scientific methods and patting the head, especially one or two people patting the head," said an employee of the case company in a questionnaire survey.

The boss certainly understands the difference between scientific decision making and claps of head. No boss in the world would say that scientific decisions are bad, let me shoot my head. But the question is, when we lack effective mechanisms, can't effectively integrate cross-functional wisdom and information, and a group of people can't make effective decisions, what can the boss do without relying on his own head? Therefore, the problem here is not to change the behavior of the boss, but to improve the ability of collective decision-making, find better methodologies to organize data, integrate judgments, and make more accurate predictions.

This methodology is the Delphi * decision method. When the data is very limited, there are many unknown factors, and the uncertainty of the decision is very high, the Delphi method is a very good choice. * As early as the end of World War II, the United States Air Force used this method to gather people in related fields to judge the trend of new technologies and guide the development of new weapons. In the 1950s, RAND further optimized this methodology, such as predicting the possible consequences of using nuclear weapons [1]. There are many variants of this methodology, but the key is similar: *, anonymous, multiple rounds of feedback and corrections, until consensus is reached after *, or a predetermined threshold is reached, such as how many rounds are repeated, as shown in Figure 1.

*: This is an early decision, and the data is very limited, so it depends on judgment. Who has the judgement? *. However, the more * it becomes, the narrower the contact surface is-if you want to master a field, you can only choose to dig a little deeper. The development of new products and technologies involves a wide range of issues, and any one of them cannot be fully covered, so please judge from various aspects to increase the accuracy of prediction.

Anonymity: This is to avoid the influence of authority, title, position, personality, reputation, etc., to avoid strong functions affecting weak functions, and strong people affecting weak people. You know, the boss sits there, no matter how democratic the company is, everyone will consciously or unconsciously follow the boss's mind. Strong functions were present, and their arms were thick and their fists were big. Eighty percent of them were also in charge. Not to mention the celebrity halo. In the prediction of new products, anonymity is to omit titles, names, and sometimes even functions when summarizing forecast data. This also allows everyone to make their own judgments more confidently and independently.

Back to back: This is to further reduce the game between functions and functions, and between people. Everyone sat together to discuss. The process of expressing their opinions was actually a game. Each person represented his own function. What he said and what judgment he made depended to a large extent on what other functions and other people took. Position, and habitually hiding, such as unwilling to speak first-speak like negotiations, whoever shows the card first, who is passive in the game. Everyone makes judgments back-to-back, and collects and organizes them by special persons, which helps reduce mutual influences and games, and makes it more likely that everyone will make objective judgments as *.

0422-1.png Figure 1: Schematic diagram of Delphi * decision method

Source: Delphi Method, Dr. Phil Davidson, University of Phoenix, http://research.phoenix.edu/content/research-methodology-group/delphi-method

Another important feature of Delphi's decision-making method is the multiple rounds of circulation: each * anonymous, back-to-back judgments are collected, organized and summarized by a special person, and provided to all members to become the basis for the next round of judgment. On the basis of this, people who adjust their decision-making will generally be more consistent. In terms of prediction, the standard deviation of the predicted value is smaller and the dispersion is smaller; so again and again, * a certain degree of consensus is finally reached, such as the average value or The median is used as the final prediction [2].

* The essence of the judging method is not difficult to understand: "Three stinky tanners, who beat Zhuge Liang", everyone understands this truth; the real challenge lies in the following aspects. To the desired effect, and finally returned to the old way of the boss making plans, the boss patted his head.

First, the wrong * was selected. In terms of demand forecasting, * is defined as an in-depth understanding of the product in some way. For example, designers are familiar with whether this product is targeted at the general public or niche, product managers are familiar with the relationship between this product and other products (alternative or complementary, independent), e-commerce managers are familiar with promotional plans and consumer preferences, and purchasing managers are familiar with cost components. And * MOQ-most of this information is at the specific product level, which can significantly affect the demand for the product. But in practice, people often determine * at the professional level, and that * team becomes the chief technology officer, marketing director, product director, etc. These people are * yes (at least nominally correct) in their respective fields, but they are often not product-specific *, have limited knowledge of specific products, and are not the best candidates to make product forecasts; they have high positions and heavy responsibilities He is immersed in various affairs and often does not have enough energy to invest in specific new product predictions. The judgment of specific products is often lower than the average level, which instead reduces the overall judgment ability and reduces the efficiency of decision-making [3].

Second, make * judgment the same as * slap your head. * Judgement still has to follow the decision-making process of "starting with data and ending with judgment", but the data is relatively small and it is more unstructured. For product-level *, they focus on their own field, often lacking overall level of information. For example, is the relationship between this product and existing products competitive or complementary? What is the sales volume of existing related products? What are the previous demand forecasts and actual sales of similar products? What is the company's sales target for this new product? We cannot simply think of sales targets as demand forecasts, but the former is indeed a big consideration for the latter. This information needs to be organized by organizers and provided to each field * to minimize their learning curve, reduce the number of cycles, and reach consensus as soon as possible. Otherwise, Delphi's judgment method is nothing more than turning one person's head into multiple people's heads without changing the nature of the head and systematically improving the quality of decision-making.


Figure 2: Several reasons for Delphi's failure

Third, there is no feedback mechanism, there are no lessons learned, and there is no way to continuously improve the quality of decision-making. * Judgment method is easy to be regarded as a one-shot deal, but it is not: we have been introducing new products. One-shot deal is often done, it has become a regular behavior, and continuous improvement is required to improve the accuracy of new product prediction. The key here is to form a closed-loop feedback mechanism. New products are on the market. With sales, we have to compare demand forecasts with each specific * forecast, and see where they are right and where they are wrong; what is wrong with a certain *? The other * has always been low, why? This is an important task for organizers: they need to collect these data and build a database to truly form collective wisdom and experience and improve the accuracy of future new product predictions. But in reality, many companies make * judgment method a one-shot deal, without feedback and summary, it is easy to form a lesson, no experience, large randomness, low accuracy, and then return to the boss to shoot the head or strong functions. It's up to the state.

For case companies, we identified a specific product to be piloted to introduce Delphi * judgment. This product is in the late stage of development, and the first order quantity needs to be determined. This product is also the key development object of case enterprises, which can get the attention of the cross-functional teams.

Finding a good product, we further identified the new product forecasting team around this product, including the following positions:

  • Product manager. The company adopts integrated product development (IPD), with the product manager as the project manager, with full responsibility for the product, and plays a key role in the launch of new products, including new product demand forecasts.
  • Designer. This position is responsible for the design of the product, and is familiar with the positioning of the product, such as the basic model or the popular model, and the impact of specific designs on demand, such as the choice of colors and accessories.
  • Store manager. This position is responsible for the sales of major online stores, is familiar with consumer demand patterns, has the ability to make horizontal comparisons between different products, and is familiar with the new promotion plan.
  • Data engineer. Actual plan manager. This position is responsible for dataization, familiar with the demand history of various products, and can help demand forecast based on historical data.
  • R & D leader. The person in charge of research and development is particularly familiar with the product and the user. She is very active in community groups such as WeChat group and QQ. She is familiar with user needs and knows a lot about users.
  • Head of sales. Although directly participating in the development of new products is less, but experienced and familiar with the feedback of the customer service team, can help make horizontal comparison of multiple products.
  • Supply chain leader. Familiar with the cost of the product, the minimum order quantity, the ladder quotation of the supplier, etc. I am also familiar with the commonality and delivery of key raw materials, and can help demand forecast from the supply side.
  • General manager. Founders are irreplaceable, especially in small and beautiful but fast-growing case companies. The general manager is deeply involved in product development, corporate operations, pricing decisions, etc., and has extensive experience, and has always played a key role in forecasting the demand for new products.

After determining the product and the * team, the organizer summoned the * team together, explained the Delphi's * judgment method, displayed product samples, and started the * team's new product prediction process "starting with the data and ending with the judgment.

First of all, what exactly do we want the team to predict? Seeing this comic in a WeChat group, the doctors went on strike in the streets, but the raised signs were crookedly written with prescription-like words, and no one could see what their appeal was. You know, this is teasing doctors who prescribe prescriptions like Tianshu [4]. In the * decision of new product introduction, what exactly do we want these * to predict? The case companies said that the products were new and the first orders were forecast to be rampant. This is unclear. There are two problems:


Figure 3: Unclear appeal, another kind of garbage in, garbage out

First, the forecast has two dimensions, quantity and time. The first order has only quantity and no time. How long is the first batch of products expected to be sold out? We keep our time open, and * members have to make their own assumptions; different assumptions about time, * members' predictions must be different, lack of consistency, no comparability, and become garbage in and garbage out.

Second, the "first order quantity" asks about supply, that is, how many orders are placed with the supplier; and * most of the people in the team are familiar with demand, but cannot make good judgments about supply: the first batch How much goods are produced is also affected by production capacity, procurement cycle, scale efficiency, etc. For example, if the production capacity is limited, the order quantity may be reduced; the longer the supply cycle, the larger the order quantity; May be bigger.

After discussing with the case enterprise, we decided to ask the team two questions: (1) How much do you think might be sold within 30 days of the new date? (2) In addition, how many raw materials (long-cycle materials) do you think can be prepared, so that we can quickly reverse the order if it needs to be replenished? We especially remind that we hope that these raw materials can be consumed within 3 to 6 months in order to control the risk of sluggishness of raw materials.

Question 1 actually asks about the 30-day forecast of the new demand. It has time and quantity, and the limits are very clear. In the last stage, the sales, design, product management and other functions in the team are deeply involved, have certain experience in the past new products, and be able to make certain predictions for the next new product.

Question 2 actually asks for the second and third months of the forecast. The entire supply chain cycle of the case company is roughly 3 months: 1 month for raw material procurement, 1 month for processing into semi-finished products, and 1 month for processing into finished products. Due to the high uncertainty of Shangxin, case companies usually adopt long-term unified procurement of raw materials to obtain certain scale benefits, but only process part of them into finished products to control the inventory risk of finished products. From the beginning of the new one, the sales volume on the first day is very valuable, and it is decided whether to quickly process the remaining materials into finished products (this will be discussed later).

We understand that the first problem is relatively more intuitive and should be able to get good judgments; the second problem is relatively more difficult to make judgments. * The team needs to better understand the entire supply cycle in order to improve the prediction and management of long-cycle materials ( It was found in specific cases that * we judged the second question not very well).

Identified the specific questions to ask. What information is known next and can be provided to the * team to shorten the learning curve and improve the quality of decisions? You know, * decisions are not about clap your head; they make judgments based on past experience. And the past experience, in fact, a lot of it has been condensed in the data, such as sales, we can aggregate such data, unified to the * team. This is especially helpful for functions that don't deal with data often, such as design and product management. Otherwise, they will just clap their heads.

In the case, we decided to provide two types of information: (1) information about similar products, such as sales during different periods; (2) specific information about the product, such as product positioning, minimum MOQ of raw materials, and supplier Ladder quote.

Similar product information: Last year, there were 6 similar products. When were they listed on the market, how many were produced in the first batch, how many were sold in 30 days, how much were the first month sold out, and how many were sold in three and six months. This gives the * team more horizontal comparison information (see Figure 4). The organizer originally provided similar products from the previous year, but there were two problems: (1) the sales data was incomplete for a long time, and (2) there were too many products in two years, which easily caused information overload and adversely affected * Member effective judgment. The organizer also wants to provide four-quadrant classification of each product (the level of output and the level of sales), as well as the return rate. If there is also a problem of information overload, they will be removed.


Figure 4: The historical sales volume of the product can be referred to (case confidential corporate information is obscured)

Information about this product: Is the product positioning high, medium or basic (by volume)? Is the relationship with existing products complementary, cannibalizing, or independent? These will affect the prediction of the product. In addition, the forecast also needs to purchase relevant data, such as the minimum MOQ of the main material, the supplier's stepped quotation, whether there are special processes, such as dyeing, surface treatment, etc., and the corresponding additional costs (with * MOQ) related).

The organizer prepared the key background information, edited it into a page of A4 paper size, and convened a meeting to explain the methodology of Delphi's judgment method, display product samples, and distribute basic background information to everyone. Forecasting process.

In the first round, each * returned to his office, analyzed the existing data, collected more information, made independent, back-to-back judgments, scanned the QR code, and passed the questionnaire star (www.wjx.com) Fill in the following information online (Figure 5):

(1) New product forecast: What is the sales forecast for the last 30 days? How much raw material inventory should be prepared; (2) Reasons based on; (3) Suggestions for further improvement of the methodology, such as what useful information has not been provided and which appropriate people have not been invited to the * team, etc.

After the questionnaire *, we ask each * to fill in their name and other information. This is mainly to urge the people to complete their tasks carefully, and also help the organizer to track the results of each judgment, with a view to improving cyclically. Here, the organizer clearly stated that the results will be fed back to the * team in an anonymous way, so that the * people will have no worries and make * good judgments.


Figure 5: Questionnaire design for the first round of judgment

The organizer summarizes the results of the first round, such as the predictions made by each of the eight cross-functional *, what are the reasons, and additional information that can be provided (based on the first round * feedback), and distributed uniformly To the * team. Hold a simple team meeting to make sure everyone understands the results of the first round. Note that the discussion is not for everyone to judge who is right and who is wrong. Otherwise, the strong function may affect the weak function, and the strong person may affect the weak person, thus affecting the objectivity of the second round of prediction.

The methodology of the second round is the same as that of the first round. Each * member decides independently and back-to-back, whether to modify his forecast result in the previous round, and state the reason.

The organizer summarizes the results of the second round. If the results of the second round are still quite different, enter the third round. It is hoped that * three more rounds, * the team can reach a considerable consensus on the prediction of the new product, and the organizer * finally takes the average or median method to determine the * final new product forecast.

For the new 30-day sales forecast, after two rounds of forecasting, the results of 7 ** (one of them * vacation, only after the first round is eliminated) are significantly convergent, expressed in standard deviation and dispersion, that is, These two values are significantly reduced-the dispersion is reduced from xxx to xxx in the first round, and the standard deviation is reduced from xxx to xxx, as shown in Figure 6.

0422-6.png Figure 6: After two rounds of predictions, the dispersion of * predicted values has dropped significantly, and the predicted values are more similar

However, for the prediction of additional materials, after two rounds of judgment, the dispersion of the prediction is still very large.

As shown in Figure 7, although the picture on the left looks more "convergent", this is misleading: the first round of predictions has many values from 0 to 3000, and the second round is mainly focused on 1000 and 3000- -3 digits * predict 1000, and 3 digits predict 3000. This should not be a coincidence. It seems more like a considerable number of people are not at the bottom of their minds. They just "get together" and see who has good reason in the first round of prediction. It depends on that person. The cost and the differentiated design of this product, the target of this product should be basic models and high sales volume; the one who predicted the first round of 1000 said that this design is in contrast to other "contrast colors", "not pleasing in summer "It is expected to only be launched in the spring and autumn seasons," and there are many similar products on the line from March to May-this model is obviously low sales and lacks stamina. These two reasons seemed to be very good, and everyone followed suit and became two factions. However, the gap between 1000 and 3000 is too large, so we have not reached an agreement on this issue.

Of course, this also reflects that the case companies have not reached a consensus on the positioning of the product.

As a case designer, I think we asked the wrong question: * The main experience of the product managers, designers, R & D leaders, etc. among the members is before and after the new one. As for the new one, how much should be filled? Experience, of course, cannot make high-quality judgments. My suggestion is not to allow the * group to predict the subsequent material preparation, but directly to the supply chain to estimate the demand for the next two months based on the 30-day sales forecast ---- we will talk about later, the new period The correlation between sales volume and subsequent normal sales volume is very high.


Figure 7: After two rounds of predictions, the predictions for additional stocks are still very different

This case is not over. As an internship case, after the actual sales data of the product comes out, we need to further summarize and analyze it to see how accurate the forecast is and summarize the experience to improve it. For example, rewards * accurate *, the group shares successful experiences, summarizes lessons learned, etc., and allows more employees to participate in learning. For example, product managers should be deeply involved, and everyone should be familiar with this methodology. This is considered a closed loop completion.

In addition, as part of the methodology, companies must decide (1) what products are suitable for the Delphi method? (2) Which function is responsible for maintaining this methodology and the processes behind it?

The Delphi method requires cross-functional participation, multiple rounds before and after, plus preparation of basic information and multiple rounds of data analysis. It requires more resources and belongs to "heavy weapons", so it cannot be abused. Companies must define appropriate products, such as greater uncertainty, more new features, and fewer similar products.

In the same way, the Delphi method needs the responsibility function to maintain. Overall, this is a decision-making process that is part of the requirements plan and can be maintained by the planning function. Of course, in many companies, especially small and beautiful companies, the demand planning for new products is often the responsibility of the product manager, and this process can also be responsible for product management. However, the problem is that product management is relatively decentralized. For example, there are multiple product managers, and the degree of overlap between some products is not high. It is difficult to find the best interface. The best practices are difficult to solidify and spread. The planning functions, although naturally centralized, often have limited impact and are difficult to effectively promote cross-functional collaboration. It may be tried this way: As part of the new product integrated development (IPD), Delphi method is responsible for the overall product management, but the planning function is responsible for organization and execution. This is the same reason that the product manager is responsible for the entire product development, and the design is responsible for the specific design work.

[1] Avella, JR (2016). Delphi panels: Research design, procedures, advantages, and challenges. International

Journal of Doctoral Studies, 11, 305-321. Retrieved from http://www.informingscience.org/Publications/3561

[2] The median is used to avoid the influence of extreme values on the average. For example, Gates and I have an average wealth of $ 49 billion, but you know it's all Gates money. This is also to reduce the game, such as a function deliberately inflated. Of course, when the dispersion of the data is small, the average value should be closer to the truth than the median, which is more meaningful mathematically.

[3] I asked a company why the return period is long? The answer was mainly slow decision-making and long material procurement cycle. Long cycle materials are easy to understand. Decision-making is slow, of course, among those decision-makers: the higher the position, the less product-level information they are required to make product-level decisions, such as demand forecasting. Of course they are not assured, so procrastination prevails As a result, it is too late to make predictions; the predictions are wrong, and there is no way to correct them as soon as possible.

[4] I saw a related doctor strike on The Telegraphs in the UK, and another photo of this angry doctor holding a sign. It was clearly written on it. This picture was taken by PS. Original: Doctors' strike: the NHS exists to serve patients, not keep doctors happy, by JAMES KIRKUP, www.telegraph.co.uk.

  • ① For all works marked "Source: Chemical Machinery Network" on this website, the copyright belongs to Chemical Machinery Network. Please reprint the website of Chemical Machinery (http://z-han.com) for reprinting. Violators of this website will be held responsible for relevant legal responsibilities.
  • ② All works on this website that do not indicate "Source: Chemical Machinery Network" are reproduced from other carriers. The purpose of reprinting is to convey more information, which does not mean that this website agrees with its views or confirms the authenticity of its content. When other media, websites or individuals use this work from this website, they shall bear their own legal responsibilities such as copyright.
  • ③ All information published on this website does not declare or guarantee the correctness or reliability of its content; you hereby accept and acknowledge that the risk of trusting any information shall be borne by yourself.
  • ④ If you are involved in the content of the work, copyright or other issues, please contact this website within one week from the date of publication of the work, otherwise it is deemed to waive the relevant rights.
About Us - Contact Us - Announcement - Terms of Service - Site Help - Business Services - Legal Statement - Legal Advisors
National Customer Service Hotline: 400-667-8871
Copyright ? 2004-2013 Chemical Machinery Network z-han.com Jiangsu Dongwang Information Technology Co., Ltd. All rights reserved
  • Advertising business (image advertising, membership application): 400-667-8871
  • Equipment procurement (equipment procurement, shopping guide consultation): 400-667-8871
  • Information editor (member review, information release): 400-667-8871
  • Media cooperation (exhibitions, associations, magazines): 400-667-8871
  • Human resources: 400-667-8871
  • Fax: 0519-89897519
Su ICP prepares 05005804 value-added telecommunication business license Su B2-20110228 E-mail: