Feedback loop is a loop where the output of the system is fed back into the system and the system is calibrated.
In more simple terms, it means collecting feedback and using it to improve the system.
What Is Easy to Get Wrong
Looking at Data Instead of Talking to Users
Insights are qualitative in nature. And most insights come from qualitative data.
The more data you have, the more tempting it is to look at quantitative data (statistics) and dream up reasons for user behavior and insights.
This problem gets worse with time and success - the bigger the organization, the more specialized (blind to the bigger picture) people get and the more data there is to look at.
Real insights, however, mostly come from talking to users. (There are also some that come from understanding changes in technology and industry.)
Avoid the data trap by setting up processes that force everyone in the organization to frequently talk to customers.
Not Calibrating The System
As John Gall noted:
“Just calling it “feedback” doesn’t mean it has actually fed back. [...] It hasn’t fed back until the system changes courses.”
In simpler terms - just because some data gets collected doesn’t mean that changes get made based on that data.
Not Dogfooding The Product
“I’m not making it for them. I’m making it for me. And it turns out that when you make something truly for yourself, you’re doing the best thing that you possibly can for the audience.”
Dogfooding is the shortest feedback loop possible. It can be done either for the main use case or some more niche use cases.
If some of the roles in the company can’t use the product at all for their own use cases, consider:
- switching the roles temporarily;
- doing support days (for example, everyone in the company has to work in support for at least X days per quarter).
Not doing any of the above will most likely result in employees having no emotional connection to the problem (and therefore also lower motivation to solve it).
Types of Feedback Loops
Based On System Improvement
Automated Feedback Loops
Sometimes (especially with consumer businesses or when dealing with a lot of data) it’s possible to automatically improve the system as it gets used.
For example, each time you rate a restaurant on Wolt, that rating will show for new orders. Depending on the rating users will either order more or less food from that restaurant (i.e. they know what to expect). No human intervention by the startup is required.
A similar thing happens with automatic pricing in Bolt. If the demand is low, the prices will drop automatically, improving the service for customers.
Manual Feedback Loops
Manual feedback loops are a requirement for every startup. It doesn’t matter if automatic feedback loops exist or not. You need insight in order to keep improving the product.
Manual feedback can take two forms: quantitative and qualitative data. The first mostly deals with looking at analytics. The latter is also called a customer feedback loop sometimes and is the manual collection of data. JBTD is a method for having a manual feedback loop, for example.
Based On Sentiment
Depending on how you implement the feedback loops (and what your goal is), the loops might produce biased input.
Positive Sentiment Bias
You’ll often see very unlikely product reviews and app review scores in app stores and product directories. It’s mainly because those loops also feed into growth loops. It’s how some potential customers find your product.
The people who give a review are often selected based on specific criteria. For example, the app first asks you how satisfied you are with the app first inside the app. If you are very satisfied, only then will the app send you to an external public review site.
It’s useful as a growth loop but the feedback is highly biased.
Negative Sentiment Bias
Negative sentiment is often collected when users either ditch the app, cancel their subscription, or when the salespeople don’t manage to close the deal.
This is a manual feedback loop that’s very useful for improving the product.
Balanced Sentiment
Balanced sentiment tries to get an accurate picture of how satisfied the users are with the app. It can feed back both to product management and marketing. It can be used over time to track the progress.
Quite often, it’s done with NPS (Net Promoter Score) but other methods work as well.
Note that it matters a lot how exactly you define “users” or “customers”. Ask the wrong users and you'll get not only a sentiment bias but a product that tries to solve too many problems of too different people.
Even how you send out the survey (email, SMS, in-app message, phone call etc.) matters in terms of which type of users will respond more likely.
Innovation
Innovation is a process that results in either improved or totally new products, and/or business models.
From the business point of view, it makes sense to divide innovation into two categories:
- Sustainable innovation - improvements to the existing products;
- Disruptive innovation - improvements that change at least one of the Four Fits.
Existing businesses (especially market leaders) are very good at sustainable innovation but less so at disruptive innovation - it’s very hard to change an existing business model.
The speed of innovation is partly determined by the speed of feedback loops.
A good case in point is governments. The feedback loops there work very slowly or are missing altogether (in the case of the Soviet Union).
Your startup should be set up in a way where you collect feedback from everyone who deals with customers (sales, customer success, customer support etc.). And then feed that info back into product management and marketing.
