3 key points of growth related data work

Like product life cycle, growth related data work also have its life cycle. Generally speaking, in startup teams which I had joined before, there is always one epic guy?from big company, you know like, who have a book or sth else.?

Most weird records?can be handled in a discretionary?way by that guy. If you can't find someone else who is better than you after joining a new team, then it is you.?

For sure it?is not scaleable to?always rely on expert's knee-jerk reaction to sudden change of data. So in this article I will list out 3 basic poins of?growth related work from?a product?or?operation role in a bigger or more traditional?team.?

The reason why I am trying to write down this,?is due to after many new?members?joining the team during the?fast recover of Covid in China, we need to re-build the systematic way of regular data?work.

First - is to estimate

I know many people are like?to write down simpliest thing at first, but the very beginning of?growth related work is?the hardest part, which is to have a?forcast or estimate, and to review the gap regularly.?In the same time, to?align with all stakeholders including executive team.

This is not as quick as it sounds, like that, to imagine a?picture that a man points a curve in PPT, in real life,?we?need to breakdown the data set to very detail, like each region, each distribution channel and each platform.?

For example:

Lets say we have four main source of traffic, and in terms of historical record of the cost of new user, the retention rate (or churn rate), the difference between each region, we predict a monthly growth from beginning of one year. Every month?we can check the trend and to see, whether some part is over or below expectation.

If you already have some experts, these expectations are already?in their?mind, if you are running a team full of green hand, below are my two tips about how to enhance new members' sensitivity of data:

Know Your?Number.?if you?always say "I?need to check?offline", then you can be replaced by?that?guy you always?check with.

Those who?question your method of data collecting and?poke the way you use data, is helping you. No matter how they looks like?- I know many of them look horrible -?this kind of action means we are same kind of people, who trust?rationality.

Second, is to?recap?

Matching data to?real world issues is next thing to do, this is the main difference between product role and finance role,?the later part team do not very care about what exactly had happened in frontline of business.?

For example:

Lets say we have several calendars,?

1. One is user oriented campaign, to record the exposure and user touch point during a promo, which impacts traffic in several distribution channels;?
2. Other one is supplier oriented promo, which impact the price from the cost side, in a short period;?
3. Third one is like feature release calendar, which shows what kind of new feature is available to users from which day.
4. Fourth one, besides internal calendars, also external data is important, such as unknown seasonal impact (like you are operating a new market), some noticeable change from?your traffic source.

The main idea is like this:?

the prediction is based on understanding of historical regular pattern, if we find out gap,?it means some parameters need to be adjusted,?including?

1. The allocation of resource - both money and people;?
2. To consider?about the?priority we set before;?
3. Or just to be?noted, in case something out of our knowledge range is happening.

Third,?is to tell a story

In first part, we mentioned the regular data work need "to?align with all stakeholders including executive team",?if you have worked with those experienced guys, you may feel that, they just enter?last step?immediately, which is?to?suggest how to improve with a story.

If we?go through this short?article, the process is like:

1. To know the prediction, and?then to check the?gap, like which part?is over estimation?, which else?is lower than prediction.
2. To compare the "gap" with what had happened , and try to explain.
3. To combine with competitor data?to avoid overlooking
4. Then, to build up a story about,?how to improve.

For example:

What will you do if a daily tracking number shows weird record? This is one of my interview question to check the mindset. The answer could be like:

Before answering anything, try to double confirm the timing of this check: is it a regular report checking? or during a UAT testing? or randomly find something during a targeted topic report. The sorting of below actions may slightly change in terms of the background of this data work.

Then begin to answer.

Double check the event tracking feature, any release notify during this timing.

If tracking feature didn't change, and traffic increase a lot, check the campaign calendar and other paid/unpaid traffic in terms of which part traffic is increasing.

If traffic didn't change, or the impacted part of traffic didn't change, then step by step to see where user leave our booking funnel, this is also called traditional funnel analysis.?

And loop again, check event tracking, and check release notification.

If traffic, features and funnel are all looks nice, check?fraud related data like user register date, referral date, booking window, cancel trend and so on.

Above is all, try to estimate, and to record and understand what happen and to tell a story, sooner or later, most people can be good at data driven regular?work.?

The only thing which is hard to train is, you must love numbers and have passion to run?calculations in mind anytime, this is also one of my point of interviewing.

Thank you for reading.