1. Look at the data to see the dimension
When analyzing a certain business or a certain module of the business, it can be analyzed from two perspectives.
First, look at some data from a broad perspective. For example, for a certain product (consumer product), it is necessary to analyze what kind of data it is in the general environment, such as market ranking and market share. It is also necessary to record the overall volatility of the market and the data of competing products. These can generally be obtained through third-party research institutions or industry reports.
FineReport Data Report Kanban
Then you need to analyze what data you care about internally for this product. For example, the monthly and weekly sales volume, the sales volume of various channels, the increase in sales and the increase in brand awareness brought about by advertising, promotional activities, offline activities, and joint activities.
After understanding the above data, you should have a macro understanding of all aspects of this product, and have a clear understanding of which data indicators need to be improved. After that, it is refined to a certain module dimension for further analysis. For example, e-commerce channels need to pay more attention, such as DAU, WAU, customer unit price, repurchase, user churn, etc., each module can establish funnel information. In actual analysis, we should also pay attention to data anomalies and do targeted analysis.
2. What is a good data indicator?
Not all data fields can be used as indicators. Before choosing the data indicators to monitor, you can ask yourself a few questions:
1. What is your core focus on this product? Such as sales growth rate, market share?
2. Can these indicators reflect the trend of this product? If these indicators improve, does it mean that the company is developing in a good direction?
3. Is this indicator controllable and can it be operated? If there are some indicators that your current technology can't count at all, it will not help.
The establishment of the four indicators is rigorous and there is no loophole. To describe a product, there must be a series of indicators, whether these monograms can fully explain the situation and whether they can fully support when verifying your hypothesis.
Let's talk about the data and analysis of indicators.
Sometimes the data itself is deceptive. For example, the sales this month is 600W, and the other two biggest competing products are 300W and 250W respectively. It seems that the sum is not as big as ours. But in fact, our sales last month was 800W, and the competing products were 200W and 180W. The numbers look nice, but we're way down in terms of comparisons and ratios.
Therefore, good data must exist in the form of ratios, and there must be relative data of contrasting nature.
单纯一个或几个数据情况是没有意义的，点连成线，线构成面去展示，用图表展示一段时间的整体趋势，才能客观评价产品的健康程度。通常我们用BI或者报表搭建数据看板（dashboard）监测重要的数据指标。
FineReport数据报表看板
FineBI-大数据BI分析工具
数据分析要有目的性，不要被数据迷惑。上面的数据就是只关注了销售额情况，但忽略了整体分析而导致的表面欺骗。
3、发现数据异常，该从哪几个维度去分析？
有时候总量的角度是无法看出问题的，比如销售额、UV下降了，我们需要进一步细化去分析。看销售额总量的时候明显是下降了，先确定大的市场有没有波动，竞争对手有没有动作，需要查看市场总额以及竞争对手每个品类的数据。然后分析自身，每个渠道的销售额情况监测，每个区域的销售额情况，每个时间段的销售情况，把活动时间比如五一的数据扣除拟合，将有问题的标记出来。如果是渠道问题优化渠道，如果是市场波动，需要全局考虑战略和市场对策。
所以如果是销售额的分析，需要从渠道、活动时间点、地域等情况去深入分析。比如，是不是因为十一的活动导致销量有个明显的上升趋势？是不是因为上周搞了促销活动，导致本周一线业务员有个消极缓冲的时段，整体销售额低迷。
另外，数据一场也不是什么坏事，如果再数据分析过程中发现某数据表现极好，比如某渠道的销售增长率很高，是不是可以思考为什么会这样，有什么好的经验借鉴，甚至是不是要考虑调整渠道的投放比率。
4、不同阶段的关键性指标应该是随着业务的变化而变化的
在做数据分析前，我们会确定分析的目标，每个阶段不同。以电商渠道为例，有时候是分析各类活动效果以进一步优化方案或者挑选最合适的方案，留下分析模型以便后续活动对比预测；有时候是研究广告投放，在预算内让营收最大化；有时候是增强用户粘性，提高用户活跃度。
所以不同阶段无论是关键性指标还是KPI都要做相应调整。比如产品投放初期，关注用户数、订单数，后续考虑用户活跃度，回购率，客单价等等。
本文首发于CSDN：http://blog.csdn.net/szd_happy/article/details/72457912