Several Twitch creators seem to be overwhelmed by their overall metrics – total views, followers, watch time, etc. They can provide peace of mind, but do not help to determine what the next step should be as many of them are meaningless. The answer isn’t to collect additional metrics; rather, it should be to focus on individual decision metrics regarding the one lever at a time you want or need to use.
If you have ever experienced a rise in average viewers, but a drop in chat activity, or seen your follower count grow while losing regulars, then you are aware of the reason for this. The total number of viewers can change for reasons that are not necessarily repeatable. The ratio or percentage of change for an average viewer to total viewers gives you a true indication of how things are improving.
Some creators test scaling at the top of the funnel by utilizing a Twitch promotion service when their metrics are strong (notably clicks and retention) as a means to test whether their conversion rates will remain consistent with a larger audience size.
Build the lightweight dashboard
Twitch Analytics provides you with raw data used to calculate your ratios and track change week-over-week. Use Twitch’s own analytics overview link, located here, as a reminder of the analytics that are available natively on Twitch, but know that it’s how you use that data together that gets really cool!
You are developing several buckets: exposure and conversion. The exposure bucket indicates how many people had an opportunity to find out about my business. The conversion bucket shows how many people talked about or did something after discovering my business. Many people mix the exposure and conversion numbers to quantify growth. Too many people are mixing the exposure and conversion numbers to measure growth.
- Exposure: unique impressions (where available), unique viewers, live views, average viewers
- Conversions are all types of returning visitors, viewers who chat, follows, subscribers (if tracked), and clicks (if you are tracking from overlays or panels.)
- Parameters: hours streamed, total streams, category, format (DJ performance, artist’s practice / production show, artist just talking to fans).
The 4 ratios that matter
Each of these four ratios gives you a summary of what’s happening in your business, because they all correspond to an action you can take; when used together, they identify the classic trap where an improvement to one of the top-line metrics essentially hides bad underlying performance (usually for months).
1) If you have a packaging issue, you are likely to have a good number of clicks on unique impressions (packaging). Just because a user sees your channel or stream in their recommendations does not mean there is a problem with your content at this point. These areas include title clarity, category selection, publishing schedule, thumbnail style (if applicable), and your one line promise. Rewrite your titles as if you are selling a single moment, not simply “good time.” You may want to try: “Live Synth Jam – Creating a Hook from Scratch” vs. “Making Music.”
2) Returning & unique viewers (stickiness). This is the “did I get a second date” number. If this metric is low, you need to stop looking for new viewers and focus on reducing friction in the viewer experience (faster starts; clear segmentation; fewer dead zones; consistent on-ramp). Many music streams experience significant viewer drop-off within the first 3 minutes because the host spent most (or all) of that time setting up plugins or searching for files. Build a 2-minute cold open with something you can do on a bad day (e.g. short riff; poll; repeating warmup loop).
3) The number of chatters per 100 viewers is an indication of how many people are part of the community. After being raided, we see an increase in average number of viewers, but having chatters per 100 viewers gives us an indication if people are engaged or participating. If you want to see a higher chatters per 100 viewers increase, you should try to ask simpler questions. An example is if I ask how many keys will it take for this drop will happen, the results I receive will have a greater outperformance than if I say please provide feedback on this drop. Another way to drive more chatters per 100 viewers is to provide your viewers with something they can use as a point in time to correlate with, such as milestones and name the moments of the stream. Viewers respond to a stream when it has a rhythm to it, not when it feels like static background noise.
4) Follows/Hour (Conversion)This is a benchmark you should use to evaluate your conversion rates when streaming for extended periods of time. If your follows/hour drop significantly during marathon streams, you have not increased your conversions, you have simply built up the amount of exposure you have received on the site. Recommendation: create one clear follows event per hour by either adding a pinned copy of the previous payoff to your chat, or provide a non-random, six-minute follow reminder.
The other 12 metrics (and what they unlock)
The use of ratios as a decision making tool is important, but it is also necessary to have metrics to support the ratios, thereby avoiding any overreaction to the ratios. Here are 12 weekly metrics that you should track, and a single activity that you should trigger as a result of each metric. If you do not have a trigger activity for a metric, then you should not have that metric on your dashboard.
- Hours streamed: If this swings wildly, your results will too. Lock a sustainable baseline.
- Stream count: More sessions usually beats one long broadcast for learning loops.
- Unique viewers: If this stalls, you need more distribution outside Twitch (clips, collabs, raids).
- Average viewers: If this rises but chatters do not, your experience is passive.
- Peak viewers: Treat spikes as clues. What happened right before the peak?
- Watch time (total): Useful, but only paired with hours streamed so you can see efficiency.
- Average watch time: If it falls, your pacing or segmenting is drifting.
- Returning viewers: If this is flat, build recurring bits (weekly beat battles, fixed warmups).
- When your new follower numbers are increasing but your followers to per hour ratio is decreasing it means you gained more exposure but lower conversion.
- Chat messages: If messages rise but chatters do not, a few regulars are carrying the room.
- Unique chatters: If this grows, you are onboarding new community members well.
- Raids received: Nice, but track what percentage return next week (otherwise it is fireworks).
For a quick review of how to locate a lot of these items in Twitch, the practical video breakdown of everything can be found at the link referenced above. Tools come and go, but the logic is timeless. Get back to the spreadsheet and feel free to watch the walkthrough video provided.
What “good” looks like by size
Because genres and categories have different behaviors, benchmarks can be fluid. However, you can utilize size bands as a guidepost to help you avoid comparing a 5-view channel to a 500-view channel.
- 0-10 average viewers: prioritize packaging and stickiness. One solid returning viewer is a win.
- 10-50 average viewers: push community pull. You want more chatters, not just more lurkers.
- 50-200 average viewers: tighten conversion. Follows per hour should not crater as exposure grows.
- Track viewer stats and build a workflow: be consistent, have team members in place (moderators, use segmentation/plan, use pipeline to create clips).
I had a Manager who had a tendency of celebrating higher arrival rates from raids, but didn’t pay any attention to how many people returned to the show after having participated in a raid. The moment we started counting returned and unique participants, the “growth” channel very quickly changed. Therefore, we developed a reusable segment on Tuesday’s rather than continuing to do raids, for advertising reasons. While less exciting it has been a far more successful measure of growth.
Run the weekly review (one experiment)
You want to make sure you spend 20-30 minutes each week reviewing your accomplishments and insights, but you should only finish with one (not ten or fifty). Ten experiments on your calendar will just create anxiety, not productivity or progress.
- Pull the week totals into your sheet (same day each week).
- Compute the 4 ratios, then circle the weakest one.
- Select one activity associated with this ratio and write a ‘Test’ statement: “If I do action A, then ratio B will change by amount C”.
- Run the change for 2-3 streams, then check direction, not perfection.
By working with PromosoundGroup, you can use this dashboard for growth initiatives to help you to separate “we have an increase in traffic” from “we have increased our conversion rates.” If you do that you are saving financial resources as well as employee morale.
You shouldn’t overlook simple operational victories, such as having a steady schedule, segmenting similar activities, and having a good reason to return! The ratio of successful to unsuccessful experiences will give you data that is far more reliable than simply looking at how many customers you have in your account at the end of the month.
Once this Dataset Consistently Opens Up A Story – Continue To Do So & Record The Changes – Conduct Trend Analysis Only Having One Experiment Taking Place At Any Given Time And Letting The Total = Results. If You Are A Pro or A Tourist The Part That Will Win You Long Term On Twitch Is How You Measure As An Operator vs As A Tourist.
