How Facebook uses data to guide design

Adam Mosseri, the Facebook product design manager, recently made a great speech: data informed, not a driven (not a driver)

The presentation is based on how Facebook uses data to guide product design, as well as Facebook's product design team architecture, a number of design cases, and how Facebook makes product design decisions. In this paper, the contents are summarized as follows (translation is not translated).

1. Small and fine product design team

"It ' s important to understand, at Facebook, we believe in particularly Sgt teams."
(one thing that needs to be clearly understood is that, on Facebook, we specifically recognize small teams)

Facebook's product design team usually has only 6 or 7 people because they think it's the most efficient and the team has a high decision-making power, a flat architecture (flat decision-making businessesflat-out), and some decisions can be passed directly to the CEO.

The structure of a product design team is usually this:

is a designer, responsible for interaction, vision, product strategy (products strategy), and even a little front-end to a user researcher. Two years ago, Facebook had only a small number of researchers and was only responsible for big projects, and now, as the emphasis on qualitative research has increased, almost every team has one. One to four engineers. A project manager (PM), the project manager is usually the same as the mini CEO in the team, and needs to be responsible for the product. Not only to ensure the timely completion of projects, resource coordination, but also to ensure product quality.

On top of the product design team, there will be a manager in charge of overall management. The speaker himself held the position, and he was responsible for nine product design teams. In addition, there is a dedicated 20-person data team, half the engineers, and half the data analysis experts.

2. How Facebook uses data

"Data helps us understand how users use the product, abound then in turn helps us understand"
(data helps us understand how users can use our products to help us better optimize our products)

Facebook collects 4TB of user behavior data each day, which can guide the design for continuous optimization. For example, through waterfall analysis, tracking the conversion/loss rate of interactive steps, a large number of A/B testing, observing behavior usage patterns, optimizing interface interaction and Operation flow.

Example 1: Reconstruction of photo uploading device

When the new photo upload is online, 85% of users upload only one photo at a time. In response to this pain point, the interface was optimized, after the user selected a photo surfaced after a upload tips, although a bit of "disturbing", but the results of each upload only one photo of the number of users to 40%, the average number of uploads per upload from 3 to 11.

Similarly, through the data they found that the user in Sunday to reach the usual 150%, this pattern of behavior recognition also brings a lot of inspiration for the design.

Case 2:what ' s on your mind displacement

After the launch of question this product, the original what ' s on your mind need to adjust the position. So they tested 8 design versions with A/B testing, but each version had no statistically significant impact on the core metrics, so eventually they opted for the most concise scenario.

Example 3: Modification of the cancellation page

In addition to waterfall analysis, the data will be used for retrospective analysis and optimization of small pages. For example, a designer has launched a cancellation page makeover, and he's thinking about how to undo it when the user is about to leave. Eventually he transformed the page through an emotional design (Figure 1), successfully reducing the cancellation rate by 7%, a great number for Facebook users (Facebook had 70 million users).

3. Vigilance over the use of data

"... at Facebook, in Product–we called product" Product management and product design "–that there ' a healthy skepticism of maximally ov Erly Data Driven "

The speaker mentions three reasons for data skepticism:

A set of data metrics is difficult to fully represent the values you value

There are many factors affecting product decision making, in addition to quantitative data, there are qualitative data. Facebook's user researchers do a lot of qualitative research, such as usability testing, eye-testing, and more. There are also strategic factors, user needs, competitive products, business interest factors.

Excessive reliance on data can lead to "micro-optimization" or "local maximization"

Excessive sensitivity to data changes may lead to so-called "micro-optimization" (micro-optimization), that is, excessive focus, constant pursuit of a certain indicator of ascension, while ignoring other factors. Especially when Facebook is growing in size and the teams are more focused on their own products, they can lead to only their own indicators, making the overall conflict, without a holistic view.

For example, the design of the application menu after several twists and starts, the position from the left column to the top, and then to the bottom, due to the traffic does not see a big increase, but also considered the use of Blue big button scheme. Although the flow rate increased 5 times times, but because the designers themselves find it difficult to accept, the scheme has not been released. In any case, it was not until a year later that they found that, although the target of increasing the flow of guidance to the application was optimized, there was a bottleneck of stagnation and local maximization (maximization). Eventually they set up a menu project team to overturn the previous restart and design the current version--and that's actually half the way back to the original menu.

Another example is the photo upload. The photo team spent several months improving the number of uploaded success rates, and although the numbers have improved, the success rate has been almost stable and there is no new change. If you keep circling around this indicator, there will be no qualitative change in the experience.

True innovation usually leads to poor data, but it's not necessarily a bad thing.
"... real innovation invariably involves disruption. And disruption is usually–involves a dip in metrics "

One of the core values of Facebook is breaking or even disrupting. This tends to make some indicators worse, but it is also the meaning of innovation. They have had a lot of disregard for public opinion on the innovation of data, such as the new home page, is not based on data. Of course, some of their innovations are successful (News feeds and mini feeds), and some failures (such as Beacon, import your outbound behavior into FB functions). Failure, as long as the understanding of failure, recognition of failure, can continue to move forward. It is impossible to refuse change for fear of poor data. But change can sometimes cause them to get annoyed by disrupting users, and insisting on change does not mean not listening. As the speaker mentions his favorite Facebook group is "I can't help but hate the new Facebook page." Need to learn how to better communicate the purpose and meaning of innovation to users.

4. The biggest adventure is the fear of adventure.

' For us, the greatest disorientated are really taking no disorientated at all ... at the ' end of ' we make decisions based on common sense, on Intere STS, on strategic interests. and we use data. We acknowledge it ' s important, but it ' s really ethically a Sgt piece of the pie. ”
(For us, the biggest risk is not to risk ...) Most of the time, we ultimately rely on common sense, insight, strategic significance to make decisions. We use data, data is important, but it's just a small piece of pie.

SOURCE Address: Http://piglili.blogbus.co ... 77069181.html