And within 90 days of the product launching, we were able to use this technique to be able to figure out
推出该产品90天后我们就能用这个方法来计算
what the one year value of an advertiser was, and we predicted it for the first year to 97% of what the number turned out to be.
一位广告商一年创造的价值我们预测出的数值是真实值的97%
So I think it's very important to look at your attention curve, and this is how we did it.
我认为观察用户留存曲线非常重要我们是这么操作的
If you see here.
大家看这里
This red line is number of users, who have been on your product for a certain number of days.
红色线表示使用产品一定时间的用户数
So, a bunch of people, that should say zero, a bunch of people will have been on your product,
有些人或是说没有人有些人一直使用你的产品
all of your users will have been on it at least one day.
所有的用户至少有一天使用了你的产品
But, if your product's been around a year, or whatever, you will have zero users who've been on it 366 days.
如果你的产品推出一年了使用了366天的用户数为零
Make sense? The curves sensible? Cool.
说得通吗?好的
So what you then do is look for all of your users, who have been on your product one day.
然后你这么来看用户数用户至少使用了一天你的产品
What percentage of them are monthly active.
月度活跃用户占多大百分比呢?
100% for the first 30 days obviously, because monthly active they all signed up on one day.
显然,第一个月是100%因为第一个月,只要是使用过一天的都算活跃的
But then you look at 31.
然后看第31天
Every single user, on their 31st day after registration, what percentage of them were monthly active, 32nd day, 33rd day, 34th day?
注册31天后月度活跃用户占多大百分比呢?看第32天看第33天,看第34天
And that allows you, with only something like 10,000 customers or whatever, to get a real idea of what this curve is gonna look like for your product.
这样你就可以--假设只有10000名用户就能得到你的产品的用户留存曲线
And you're gonna be able to tell does it asymptote.
你就可以判断其走向
And it'll get noisy out towards the right-hand side.
图的右边会有一些波动
Like I'm not using real data.
我用的不是真实数据
It'll get noisy out towards the right hand side, but you'll be able to get a handle on, does this curve flatten out or does it not.
图的右边会有一些波动不过你能看得出来线是不是平的
If it doesn't flatten out, don't go and do growth tactics, don't go and do virality, don't hire a growth hacker.
如果不是平的,就不要谈什么增长策略了不要谈什么病毒式增长,不要雇增长黑客