How to keep track of your growth experiments with Google Sheets
This post describes a way of tracking growth experiments using Google Sheets.
You can find all of the templates in this Google Folder. The growth experiments belong to a method that is strongly based on Brian Balfour’s “How to build a Growth Machine” article. The original on www.coelevate.com no longer exists but this one KissMetrics article seems to contain most of the important points. In my opinion, Balfour’s writing and thinking is some of the smartest in the business.
Let’s quickly recap what is Balfour’s Growth Machine Method.
The single most important thing:
Growth is about process, not tactics
Rather than what some, if not most of the articles about growth & marketing want to make you believe, growth is not caused by hacks. It’s executing these hacks in a systematic way that creates growth.
Growth is a compound effect of small incremental steps on each part of the funnel through continuous activity that should be repeatable, scalable and (somewhat) predictable.
That’s more like it. So, how do we do this? With a process:
I’ll now go very quickly through each step, and describe how you can use my templates for some of these.
On a regular basis (once a month is a good cadence) get everybody who needs to be involved in the room, and have them throw all possible ideas on the table.
This is also a great opportunity for that colleague who keeps sending you links to growth hack articles on Slack and then asks “Why aren’t we doing this?”.
I usually do this by giving everybody a bunch of post-its and have them write their ideas in keywords on there, for X minutes. Then after X minutes, everybody presents their ideas very shortly and puts them on the whiteboard. After that, put all of the double ideas together, and (if needed) perhaps weed out the ones that are really not feasible.
* how to run an efficient brainstorm session is a whole other beast.
2. Prioritize (sheet 1)
Once you get a proper list of ideas, you pour them into the Experiment Backlog (Google Sheet Template with some awful examples) and start to quantify them. Each idea takes one row and you have to fill in all of the parameters for the Hypothesis and the Resources part by giving your best possible assessment.
- Metric: what’s the key metric that you will be tracking?
- Prediction: what’s the predicted improvement for that metric?
- $ Value/month: How much MRR will this add? This might seem tricky for some ideas, but it simply requires that you calculate for each step of the funnel (calculating back from the LTV) what is the value of a customer in that step. How much is one visit worth? How much one trial?
- Probability of success (Low/Medium/High) Have you done something like this before? Does anyone have any experience with this? Is it really out there? Or do you have a lot of examples?
Here you fill in for each department how much resources it will cost in terms of Low/Medium/High. This will probably differ per company in two ways:
- the actual configuration (you might have more departments/teams)
- the weight that you give to these. In the template, it assumes that development is scarce, and so it is weighted negatively if the resource requirement is high. (Although low development resources seems a very usual case)
After you have filled all of these parameters in, you select all of the ideas and Sort Data by a number of parameters. Ideally, you sorting order is something like this:
- Sort $ Value / month from high to low
- Sort Probability of success from high to low
- Sort Resources from low to high
This should generate the lowest hanging fruit with the biggest impact. In any case, the outcome is a prioritized list of growth ideas that you should then turn into experiments.
3. Test, Implement & Analyze (sheet 2 & 3)
Write the experiment doc first
Take a growth idea from your backlog and turn it into a growth experiment, using the Experiment Doc (Google Template with terrible fictitious example of a cold email experiment of a company selling SEO services). I’ve found it extremely useful to do this before anything else. Filling this in will make you think critically about your assumptions, your hypothesis, your exact experiment but also about your metrics: it has happened to me more often than I would like to admit that I ended up with an experiment that I couldn’t statistically prove to be true, because the tracking or analytics were unreliable, not available, or unable to separate the A/B individuals.
Important: Once your Experiment doc is ready, and you are going to work on it, give the growth experiment a number (start the first one with 001 and increment) and put it in the Experiment Pipeline (Google Template here with one lousy title). Put a link to the Experiment Doc in the Pipeline sheet, for transparency and easy access.
Pro-tip: My current Experiment Pipeline template also has Status and Category. If you’re a little bit handy with Pivot tables, you can use these two columns to create an easy overview of all the experiments in the pipeline and see which step of the funnel has the most experiments running etc.
Run the test, but follow up
Whatever it is, run it. And run it for long enough. Be patient, young Skywalker. In the mean time, follow up, and regularly put the latest data in the Experiment doc, and write a few lines in the Follow up box, ideally with date-entries!
Analyze the s*** out it: Learn everything you can from each experiment
In some cases, this is easy, in other cases you’ll have to do a lot of work to get the actual proper results of your test. Cohort separation, channel separation, A/B group separation, importing and comparing from all different kinds of sources etc.
Once you come to a conclusion of some kind, ask yourself these important questions:
What happened and why?
- Why did it succeed/fail?
- Why were we close/way off of the hypothesis?
- Are there any actionable items/next steps? → To the Backlog!
And of course, don’t forget to update the Experiment Pipeline.
Then, once the experiment is over and you have reached a final conclusion, think of how you can systemize this. If the experiment was successful, can you scale it to squeeze as much out of it as possible? Are there any other learnings that can be applied to other things you do as well?
For me personally, that’s where it ends. I feel that the Experiment Pipeline provides enough information for my needs. However, if you’re in a bigger team, it might be wise to collect all of this useful information in a doc, for current & future colleagues. This info becomes useful in the next brainstorm sessions.
This blog post was the result of a discussion in the more than excellent SaaS Growth Hacking fb group, on how to keep track of your growth experiments.
Also, I got the basis for these Google Sheets while working under Seva Moshanin at Yousician. I’ve simply expanded them a bit further to fit my own way of working.
Originally published at www.sjurd.com.