How to Use Data if You Are a Designer
UX Planet — Medium | Natalia Babaeva
Looking for the data that can actually drive your design
Imagine you’re working on a new product page of a non-fiction book store — it’s an actual case I had some years ago. The current design is too outdated, so we have to create a new one from scratch.
We have data about how the current design is working. Before making the new design we want to see those numbers to make the real level-up.
It’s not an easy thing to do. Usually if you ask a designer which element of the page is there because of a certain number in Google Analytics, he’ll probably have hard time to tell.
Let’s have a look at the 4 attempts of a designer and an analyst coming together to get the data-driven design.
Attempt One. Basic Numbers
The analyst collects some basic numbers about the page — the ones he normally gets for the business and marketing.
Traffic 80K, conversion rate 2%, biggest age segment 22–35 y.o., 15% more women than men.
Designer says “Got it!’’. But is it going to help him really make a better design? Don’t think so. We all know that it doesn’t work like this.
— Because the traffic is 80K, the buttons have to be green. If the traffic were 100K, the buttons would’ve had to be red.
— Conversion rate is 2%. To raise it to 4% we need to make the page twice as long.
— Because there’s more women we should add some pink details.
Absurd, isn’t it?
The designer and the analyst see the basic numbers are not helpful for the designer. But they are not giving up.
Attempt Two. Analysing the Performance of the Page Blocks
The analyst thinks: “The designer is basically doing what? Building web pages with blocks.” The analyst collects numbers about how each of the blocks is performing. The designer in looking for the blocks he can kill or improve.
[pic with blocks and numbers about each one]
The analyst says:
40% of the audience only see the first screen of the page. Blocks 3 and 5 get the least clicks. Blocks 2 and 8 get the most clicks.
At first the designer gets really excited. Now he knows he should relocate the blocks.
— Because users don’t scroll below the first screen, we should stuff as much as possible in it.
— We should kill the worst performing blocks.
— The most popular block 8 should go right after block 2.
The designer is happy to have numbers to drive his new design. But he looks at what he gets and… it’s terrible.
— The first screen is way overloaded.
— The killed blocks were working for the decision making decision even though users didn’t click on them. Now they’re gone.
— When block 8 follows block 2, there’s no logic in the order we tell about the product.
The analysis of the block performance made the design even worth.
The designer and the analyst decide to take third attempt. The basic numbers and the block performance analysis aren’t helpful. But what is?
Attempt Three. The Analyst Tests the Designer’s Hypotheses
The analyst says: “Listen! I have too much data. I can’t guess which numbers you exactly need. So bring me your hypotheses about the current design, I’ll test them.”
The analyst wants the designer to act like a detective:
(1) collect evidence,
(2) bring hipotheses,
(3) test them and find out the right ones.
“The first and second parts are your responsibilities. The third one is mine.” — says the analyst.
The designer is shocked. He believed the analyst would put the numbers in front of him. And he, the designer, would just have to look at it and come up with some insights.
The truth is that all of the data from the GA cannot fit onto any table
Now the designer is the first one to think. His questions will define which data the analyst is going to bring. He starts to collect evidence and build hypotheses.
The designer has dozens of ways to do start with. He can watch users, learn the typical support calls, meditate — whatever. Finally the designer comes up with the list. It turns out that it’s impossible to test 90% of the hypotheses. Both are disappointed.
Some of the data is just not being collected. In some cases you need to launch an A/B test (but first design and develop the B-version which is gonna take some time). In some cases the traffic is not enough to trust the numbers.
The problem is that the designer has no idea which hypotheses are testable in retrospective and which aren’t.
The analyst blames the designer of giving him wrong hypotheses. The designer starts to doubt that the analyst is a pro.
The third attempt didn’t work out as well. But the two guys are not giving up on it.
Attempt Four. The Detective Couple
Finally the analyst and the designer decide to sit next to each other and start playing a detective couple.
The pic from the Sherlock Series
The designer and the analyst come from different worlds — the world of creativity, intuition and EQ and the world of data, spreadsheets and precision. When they get together, they can produce a synergetic explosion.
Their worlds are connected through customer routs. The designer can draw them, and the analyst can translate them into numbers.
Customer routs are the typical ways that customers take on your website. The conversions, cohorts and other metrics are just number tags attached to the different steps of the way.
Here users come to your page. Than they push this button, or they don’t. So how many of them do? (This is “conversion rate”).
Last summer there were much less pictures on the page. Did users push the button less or more than? (That’s “conversion in dynamics”).
What about those clients that got to our email list by getting a white paper? Are they buying? (This is “cohort analythis”.)
And so on.
So, How Did They Actually Do It?
The designer came analyst ask him or her these two questions first.
- Where from?
- Where to?
Where from do users come to your website and what do they leave for do they leave?
This will help you guess what are those people about. And what brought them to your page. It works much better at giving you insights then age and gender.
«I’ll tell your fortune by the user routs», 1927, borrowed from.
What are we looking for in “where from” and “where to”?
- The most popular routs. There’s huge potential in analysing them. If you find how to raise conversion in a popular rout by just 2%, it will give you more outcome than if you raise the conversion of a rare rout by 10%.
2. The routs with the biggest conversion rates. Those users are fine already. We should find out why because we don’t want to brake it. If we are lucky we can find a way to multiply the success.
3. Insights. Insight — is a non-obvious true about your user that can help you get to your goals. If it’s an insight you can say it with “Surprisingly…” or “It turns out that…”
So, our designer is sitting next to the analyst and says.
— Where from do users come to our website?
— 60% come from search. (Looks like the most popular rout).
The designer draws this kind of picture.
Most traffic come from search
— We have huge audiences in social media and our email list is the biggest on the market. But it turnes out that most people are still coming from search. (We got an insight).
— What exactly do they google?
— Names of the books + “free download”. For example “Chinese study free download”. Or just the problem that is the same as our book name, like “will power”. (Remember, we only sell non-fiction?)
— It turns out that most search are more or less “irrelevant”. The users weren’t looking for a book to buy. They wanted to get it for free or to get an immediate solution to their problem. (Another insight).
The designer is updating the picture.
— Do they buy?
— Sometimes. 1% of them convert into buyers. While the average conversion rate on the site is 2%. Because they are many they stand for a third part of the sales. (That’s exactly what usually happens with the most popular rout users. They don’t convert well but still are very important.)
— What else do they do on our website?
— Nothing really. 80% come, spend a few seconds on the page and just leave.
The designer is updating the picture again.
— So the most popular rout on our whole website looks like this — “Came from search — left forever”, right? (The complete most popular rout.)
— Looks like this.
— I’ve got an idea! If they are so many and we can’t sell to them right away, what if we make them land to our email list. We have great email marketing it’ll warm them up so they’ll buy later.
Now the designer has the desirable rout.
Follow the yellow color
— But how can we land them to our mailing list?
— We can give them some free and useful thing in exchange for their email. A short version of a book or a couple of great chapters. It has to appear on the first screen of every book page because this type of users don’t scroll any further.
This is the point where the designer brings something new to the design — a point where you are offered a giveaway.
That’s an actual screenshort in Russian. This new block in the black frame became the main source of emails for our list. It let us grow the list from 20,000 to 100,000 in less than a year.
This detail wouldn’t be there without the data. The designer and the analyst were working as a team complementing each other. Designer asked questions to prove hypotheses on the routs. The analyst was adding those number tags, showing where we need to focus.
- The idea of a designer have access to data only makes sense if numbers turn into curtain decisions.
- Basic numbers are most often of no use for the designer.
- Analysing the blocks’ performance can be deceptive.
- The designer has no idea which hypothesis are testable in retrospective and which are not, so there’s no point of making him find ones alone.
- The best way is to make the designer and the analyst come together to find hypotheses, find answers and interpret them.
- Start with user routs. The two good questions to start with are “Where from?” and “Where to?”.
- Focus on the most popular routs and on the most well converted. Look for insights along the way.
- Think how you wish to change the routs you see. Here’s the point where new elements of design come up.
It was hard to read up to this point. Good job, guys!