Summary: UX salaries down by 12% in a year | To retain top talents, assign them to experiment with AI in your business | Measuring the memorability of famous logos | Nigeria Number One in UX in Africa | Personalized image style from Midjourney | Growing VC investment in AI
UX Roundup for July 22, 2024. Happy Northern Hemisphere summer from UX Tigers! (Leonardo)
UX Salaries Drop by 12% in a Year
User Interviews has released the results from their latest UX salary survey. Lots of interesting details in the full report — here I’ll just hit a few highlights. One difference between my write-up and the full report is that the report merges salary data collected in 2023 and 2024, whereas I’m only looking at the 2024 responses. (I’m grateful to User Interviews for making the full anonymized dataset available for download.)
The most interesting finding is that UX salaries dropped by 12% over the last year, after adjusting for inflation. (Specifically, I’m looking at the number for a UX professional in the United States with 5 years’ experience. 5 years is about the mark when somebody stops being junior, even though most UX people would benefit from a further 5 years’ of experience: 10 years is really the point when somebody can be said to be senior UX staff.)
Here's the chart of UX salaries in the United States from 1998 to now, adjusted for inflation. (The numbers have been converted to the inflation level in June 2024, according to the U.S. Bureau of Labor Statistics.)
My chart plots the salary drop in 2024, but in reality, UX salaries were mainly lowered in 2023. The chart shows the years when the salary statistics were collected. When asking people to report on their salaries in April 2024, the responses will mainly reflect 2023 hiring (for starting salaries) or salary adjustments (for experienced staff). Similarly, data collected in early 2023 will mainly reflect what happened in 2022, the last year of the bubble.
We see that experienced staff’s salaries are simply back on trend. Yes, salaries dropped by 12% last year, but that was simply letting the air out of an overinflated balloon.
For starting salaries, however, this latest data is the first time since data collection started in 1998 that new-graduate salaries have dropped below the long-term trend. We all know that the UX job market is terrible for people graduating this summer and that it was terrible last year as well. (As mentioned, this chart's dot plotted for 2024 mainly reflects new-graduate hiring in mid-2023.)
It remains to be seen whether we have a permanent reset to lower starting salaries, or whether this latest number is a blip on the curve similar to the one we saw (in the upward direction) for new hires at the tail end of the dot-com bubble in 2001.
Cracks in the piggy bank for UX newbies. Hopefully, they’ll see strong compensation growth as they gain experience and become more useful to companies. (Midjourney)
Median salaries were as follows, converted into US Dollars (USD) at the exchange rates in April 2024. (These are the midpoint of all UX professionals, no matter their seniority.) For each country, I have also indicated the salary relative to Poland:
India: US $28,185 = 0.8x Poland
Poland: US $35,559 = 1x Poland (by definition)
Germany: US $76,027 = 2.1x Poland
USA: US $135,000 = 3.8x Poland
Thus, if you’re a German company, you can hire two Poles for the price of one German, whereas if you’re an American company, you can hire four Poles for the price of one American. I always say that this is not what you should actually do. Instead, take advantage of hiring much better foreign talent than you could afford to hire domestically. So a German company should pay the same for a super-Pole as for an average German and an American company should hire two super-Poles for the price they would have paid an average American. Seemingly overpaying like this will land those companies vastly superior staff in the top 1% of the UX talent pool (and probably even more elitist), because most companies watch their pennies and don’t want to pay local staff above the local salary level. Such super-experts perform at a vastly higher level than their extra cost.
4 Polish UX designers or 1 American UX designer cost about the same. But I recommend spending that money on 2 of Poland's absolute top UX professionals. (Ideogram)
A final interesting observation from the new salary survey is comparing salaries of senior UX individual contributors and UX managers. For companies with fewer than 1,000 employees, the senior contributors make 14% more than the managers. On the other hand, in big companies with more than 1,000 employees, the senior contributors make 3% less than the managers. This difference between big and small companies would be interesting to study further. A likely explanation is that in big companies, the actual quality of UX research and design (where senior UX staff excel) is less important than the ability to manipulate the bureaucracy (where UX managers excel).
I predict that AI will cause a pancaking of the UX field, with fewer management levels and more emphasis on the skills of individual contributors. This will likely mean that the salary difference in favor of senior talent relative to management will grow for small companies, and that the salary advantage currently held by managers in large companies will vanish or reverse.
To Retain Staff, Let Them Learn AI
You don’t have to let them eat cake, but you do have to let them learn AI if you want to keep them. Your 10% most talented staff are the people you really want to keep since they probably generate more than half of the value in your company. These are also the people who know that they only have a future if they gain AI skills in their chosen profession now.
The Wall Street Journal had an interesting article about this point. (Subscription required.)
The WSJ quotes Nancy Xu, founder of Moonhub, a recruiting firm that hires talent for companies in artificial intelligence. She says that many talented tech staff might be at an otherwise desirable company where they don’t get to work on AI, “but they’re leaving these companies because they want to go to an AI company.”
The 2024 Work Trend Index report from Microsoft and LinkedIn shows that job posts that mention AI perform 17% better than job posts without this attractive keyword. This is not a great statistic (because who knows if the extra job applicants are worth hiring), but it does show that employees value the ability to learn AI on the job. The only good way to learn AI is to use it for real work tasks.
Your company may be stuck with legacy procedures for the next two to five years. It takes time to invent the best way to restructure the organization and your product offerings to fully embed AI, and doing so is probably done better with next-generation AI than with the current tools.
However, the only way to find out what AI approaches work for your business is to experiment. Get hands-on experience with AI and try a broad range of different approaches. You should put your best staff on this assignment, or you’ll lose them. In my opinion, the bottom half of the staff should also learn AI, because doing so will upskill them and narrow the skill gaps, making them better performers. But if you don’t want to assign everybody to study AI, keep the less-valued staff on legacy work, and it won’t be a big loss to the business when they quit after realizing that they have been sidetracked to a loser job, working on legacy tasks.
Who should experiment with AI now, even if there are a few years to go until you make your entire company AI-driven? I would say everybody, but if that’s not in the cards, you must allow the top 10% of your employees to use AI, or they’ll quit and move to a company that protects their future careers by giving them AI experience. (Dall-E)
Measuring the Memorability of Famous Logos
The company Signs (which produces custom signs and banners) conducted a study of the memorability of 10 famous corporate logos. They asked 150 participants to draw each of the logos from memory and analyzed the results.
I particularly like the conceptual analysis of the results, showing how many people got various elements of the logos wrong. For example, the shoe label Adidas shows its name in all-lower-case, but 21% of respondents drew a logo with the standard capitalization of a proper name. This is almost twice the number of people who produced a near-perfect drawing of the Adidas logo. 11% of respondents drew four or more stripes in the logo, when the correct number is 3.
The winner among the 10 brands tested was IKEA, where 30% produced a near-perfect drawing and 56% produced a good drawing. The runner-up was Target, with 52% good drawings. These were the only two logos that more than half of the participants could reproduce fairly well from memory.
The worst-performing logo was Starbucks, with only 17% good drawings — less than a third of IKEA’s.
The general conclusion will probably not come as a big surprise to readers of this newsletter: the simpler designs were the most memorable.
The report from this study is entertaining and easy reading with many graphics. I particularly like the reproductions of all 150 drawings for each logo, sorted from lowest to highest adherence to the original. The specific results are probably only important for those 10 brands, but the methodology will generalize to testing UI elements, such as icons.
Participants were asked to draw 10 famous brand logos from memory. Only 2 out of the 10 logos scored better than 50% for “a good enough drawing” when the researchers analyzed the results. Sorry to Walmart for my cartoon character misremembering its logo and even the color of its starburst, but those were the kinds of mistakes observed in the study. (Ideogram)
(Hat tip to Giuseppe Montella for alerting me to this study.)
Nigeria Number One in UX in Africa
My latest subscriber data show that Nigeria now has 30% more subscribers to my newsletter than South Africa, which has been Africa’s number-one UX country for as long as I can remember. I haven’t been tracking this closely, but the two lines must have crossed several months ago for the difference to be so big now. (Of course, Nigeria has 3.8x the population of South Africa, so South Africa still has a much higher concentration of UX, as opposed to the absolute numbers.)
See also my article on UX in Africa.
If you want to hire African talent, it doesn’t really matter how many other UX people there are in a country. Just hire whoever will fit the best with your team. But when having remote staff, there’s still something to be said for the person living in a physical location with other local UX specialists who can support some in-person events. You might still tip the balance in favor of hires from countries with stronger local UX communities.
Africa has strong UX growth. (Midjourney)
Midjourney’s Personalized Image Style
I have not written about this before, since I have not found the feature particularly useful in practice. But in theory, it’s interesting, so let me tell you about it. (In theory, theory and practice should be the same, but in practice, they’re different.)
Midjourney has long been the best image-generating AI in terms of visual style. Usually, you would specify a visual style by adding descriptive keywords to the prompt, such as “in the style of 1960s superhero comic books.”
More recently, Midjourney has also supported numerical style references, so you can add, for example, “--serf 1032711502” for a psychedelic style. The benefit is precision; the downside is that it’s impossible to remember such long numbers. (Tip: maintain a prompt library of all sref numbers you’ve found useful and annotate each one with a few words describing the style it creates.)
Midjourney’s personalized styles are activated by adding “--p” to the prompt. This feature only works if you have “liked” more than 200 images on the platform, so that it has some inkling of your stylistic preferences. Here are some examples, using the prompt “a duck” followed by a style indicator:
Top row: Left image “in the style of 1960s superhero comic books” and right image using the psychedelic sref. Bottom row: using --p to optimize images for my personal style preferences. To the left, the prompt was “cartoon of a duck --p”; to the right, the prompt was simply “a duck --p”. Of these 4 images, the lower left is my preferred duck, so Midjourney did get personalization right in this case. (Midjourney).
In this example, I actually do prefer the images in the bottom row, which were created according to Midjourney’s understanding of my personal taste. Thus, the personalization feature did what it says on the tin.
However, for the purpose of making images for this newsletter, I have yet to get a result I preferred using --p. I am not sure why, except that I have been following a deliberate policy of pushing for divergence and variation in my images, so I don’t want a single style. (To quickly scan 600 of my newsletter images to see what I mean, check my Instagram.)
My style is no style. This is partly to symbolize my freedom from corporate conformity. But it’s also because I aim to make images that supplement the text and each individual topic.
From the UX theory perspective, it’s interesting that Midjourney can construct a decent understanding of something as vague as stylistic preferences. This is a good example of AI’s ability to comb through masses of weak signals to discover a latent structure. (Something we can use to make sense of huge volumes of qualitative user feedback.)
Growing VC Investment in AI
CB Insights has published an overview of trends in venture capital investment. The most interesting data in that report is that investment in AI has been growing steadily over the last three years, as shown in this chart:
We went from 17% of VC investments allocated to AI firms in 2022 to 28% in Q2 2024.
The growth was 5 percentage points from 2022 to 2023 and 6 percentage points from 2023 to mid-2024. The year isn’t over! We’re seeing strong growth in AI investments and possibly even accelerating growth.
Such growth is necessary to fulfill AI's super acceleration in the current decade. This vision will lead to superhuman performance by the end of the decade, but achieving this may require as much as $1T to be invested in AI by then because of AI’s voracious appetite for compute.
(Hat tip to Allie Miller for alerting me to this report. If you want to keep up with developments in AI, she’s one of the top people to follow. She has 1.4 M LinkedIn followers for a reason.)
AI is overflowing with investment money: up by almost 2/3 in two years. (Ideogram)
600 Instagram Posts
Celebrating 600 posts to the UX Tigers Instagram channel. More importantly, this means that I have published more than 600 illustrations in this newsletter since I started it a little more than a year ago.
