Twitch: Monthly Business Review

How to write a monthly report and why

Monthly reports are a mechanism we use at Twitch to drive decision-making and continuous improvement. When done correctly, they are not busy-work: writing one should involve real thought and effort. And they’re not created for other people: you should be writing for clarity within your team, though they can serve the dual purpose of keeping people outside your team informed.

Monthly reports are for your team. While they can be used for the purpose of keeping people outside your team informed, they should be primarily written to help everyone on your team understand your progress and results together.

Monthly reports are not busy work. They require real thought and effort and time, and should generate learning and business decision.

Monthly reports are backwards looking. They summarize things that happened in the past month (examples revenue generated, roadmaps changed, projects launched, conversion rates that changed, etc) and specifically progress towards goals. They aren’t about your future expectations or theories (unless they changed during the month, in which case that’s a result itself…)

Monthly reports are about your customers, not your team. They’re about things a customer would care about, not how the sausage gets made. Use post-mortems or cause-of-error processes to figure out how to fix faults in process. Use performance reviews to drive improvement in people. Use architecture reviews to figure out how you’re going to create API boundaries. Use the product development process to make sure stakeholders are informed of upcoming releases. Everything happens because “we” did it in a monthly report, because our customers don’t care which team internally was responsible.

Monthly reports are a mechanism. Mechanisms are processes which have three attributes: they produce an artifact (in this case, the monthly report itself), they include an audit mechanism (we all sit down and look at the monthly report), and they are under constant change (one of the results from the review is to decide how to change the artifact production in the future, in addition to changes to the business). As a result, one key thing you should do at every monthly review is spend a small amount of discussion about how to change the monthly report next month. If you don't sit down together with your key stakeholders and key team members and review the report, then use the results of that review to make business changes and update what you put in the report in the future, you're missing out. Monthly report reviews are a GREAT way to keep other teams in the company aware and involved in what your team is doing without creating a large burden on them or additional burden on your team beyond the report. Try to avoid doing multiple reviews! One review of the data internally to decide what to write, write the report, a second review of the report with stakeholders and team members. Don't worry about it being perfect.

Monthly reports center around goals and metrics. Choosing good goals and good metrics is hard, and a full explanation is beyond the scope of this document. However, some notes on how to pick and think about them:

• Not all goals need to be metrics, and not all metrics need to be goals. But usually, the more we align them the better.
• For goals which aren’t metrics, use green/yellow/red reporting. Green means that you’re on track. Yellow means that the current trend is off-track, but you have a plan to return to green. Red means the current trend is off-track, and you don’t know how to get back on-track. Watch out for “watermelon” goals which are green on the outside but red on the inside…the real failure isn’t missing a goal (we aim high, so we’re ok missing sometimes. Specifically, we aim to miss about 30% of the time on average). The real failure is having a goal be green all year and switch to red at the last minute. If you’re off track, better to know up front.
• Try to choose only one or two “main” goals. At Twitch, these are usually things like “hours watched” or “revenue” or “gross profit”, although sometimes they are things like “daily active viewers” or “retention”. Generally, you’ll keep the same main goal for many years in a row. You’ll sometimes hear the main goal called an “output” goal.
• Considering the main goal, choose 3-5 “support” goals. A support goal represents a specific thing we intend to change in order to drive the main goal. For example, let’s say the main goal is revenue from advertising. Good support goals might include increasing the number of streamers regularly running mid roll breaks (if we think our problem is lack of reach), or increasing the number of mid roll breaks run by streamers already participating in the program (if we think our problem is insufficient density), or implementing and launching a 3rd party verification of our ad delivery (if we think our problem is lack of advertiser trust), or selling at least $10MM of advertising to consumer packaged goods companies (if we think our problem is that we need to break into that category). We could even set all of them, if we thought they were all problems. While you keep the same main goal for many years in a row, the support goals you pick may change from year to year as your strategy evolves. Advertising businesses almost always care about increasing revenue, but only sometimes care about increasing ad density.

The key writing pattern for a monthly report is what, why, so what. 

Example: Monthly Report Methodology

What happened, concretely?

1. “Our new event Rivals generated 5MM HW (+10% YoY, +28% vs goal)”, 
2. “We launched a new feature, adoption is at 10% in the first 2 weeks as expected”, 
3. “We signed a critical deal with Activision/Blizzard”, 
4. “We learned that the viewer action that most predicts retention is watching a video for at least five minutes”, 
5. “We cancelled the launch of Project Marrow and moved resources over to Project Starfruit.”

Why did that happen? (Examples corresponding to the ones above)

1. “We had 19% large streamers vs. our usual rate of 10% large streamers. Large streamers bring their audiences with them, driving more HW.”
2. (launches don’t need a why unless they’re delayed)
3. “We found a compromise where they get to keep 100% of the up-front revenue, but our streamers get the downstream value.”
4. “Our quantitative research into viewer habits has started producing results.”
5. “When we revisited our assumptions about Starfruit, we realized we had underestimated engineering cost on this critical project. While Marrow is very promising, ultimately Starfruit takes precedence.”

So what are we doing: (Examples corresponding to the ones above)

1. “We are have created a pipeline tracking metric for large streamer participation to make sure future events track towards having at least 15% large streamers. Currently we are tracking at 8 out of 10 future events above 15% and we have begun recruiting to fill the gap in the remaining two.”
2. “We’re on track and are continuing to work on our roadmap. The next iterative launch is in 8 weeks.”
3. “We will be looking to generalize this into a new model for future deals.”
4. “We are reconsidering our roadmap and running experiments to look at five-minute plays as the key metric going forward. The primary change is that we have de-prioritized projects which will drive people to watch longer (Project Grapefruit and Strawberry) and invented a new project which is aimed specifically at conversion to five-minute play (Project Mayhem)”
5. “We are revisiting our engineering time estimation process in order to prevent future last-minute changes like this.”

Some important style-guide notes on writing in general at Twitch that often apply here:

• Whenever you give a result, give the YoY change, and the vs. goal change if you have a goal. The format is “XMM hours watched (+X% YoY, +X% vs goal)”. Those two data points help contextualize almost any number. Any number that you care enough to report on multiple months in a row probably should have a goal. If you don’t know what you WANT the number to be, how can you know if your results were good or not?
• Whenever you give a comparative statement, be concrete quantify if at all possible. Don’t say your metric “grew quickly”, say it “grew 75% vs. prior month”. Don’t say your metric was “much higher than the competition”, say it’s “13% higher than our competitive benchmark”. Adverbs and adjectives create ambiguity which is not your friend in a monthly report.