Luis Abreu Newsletter

Comment

“Amazon continues to take industry after industry not because those companies lack engineering talent but executive talent.” – Simon Wardley (Advisor to Governments and Global corporations.)

⏩ Fast-forward two years ⏩

This is still the same newsletter you subscribed to: the best content I come across, summarised to save you time.

I came to realise 99% of the industry takes 5+ years to adopt better ideas (e.g. outcome-driven roadmaps, Lean, and even User Research – we're about 30+ years late on this one). So the more these ideas are spread, maybe the faster we can grow as an industry.

Like Simon says above, success isn't about features, designers, or engineering talent, but leadership mindset.

Happenings

Apple Event: Upgrades, Upgrades, Upgrades

Carolina covers the pricing strategy of Apple’s new products and services. It’s all about getting a better price point range, should be noted that cheap subscriptions don’t have the same negative connotation as cheap hardware.

Also: smartphones have peaked, it's all about incremental camera/battery/software upgrades until the next big thing in a year or two.

Also also: that UWB chip in the latest iPhones? Far more important than 5G. It’s a robust technology that’s been around for 20 years, millimetre-location precision of other devices, 10x faster than Bluetooth, etc. Read more if you’re curious. Apple has been doing multi-year plays since ever, usually unnoticed. Eg TouchID (2013) and related hardware predates Apple Pay (2014) and Card (2019); idem for Auto-Layout (2012) and larger iPhones (2014), iPad multitasking (2017).

AI and ML

Personalised Product Images

Instead of a single product preview (apply lipstick to a user photo/video), this demo previews 9 separate products and places them in an AR view.

The AR bit is mostly a gimmick, but personalised product images help the user understand product fit and value, lowering purchase risk – a big detractor with online purchases, that retailers usually fight with easy returns, pay later schemes, and try before you buy.

This is 2 years old but no one’s tried it yet as far as I know.

Saving time with Machine Learning

Crude but interesting experiment where we save time creating a 3D model by using real-world architecture, but transform its appearance to match a specific style (AKA Style Transfer).

In the demo they create a cartoon village and one other style, but the possibilities are greater. I can see a similar experiment with a base dress shape and using ML to apply 20 different materials and styles to it, picking the production version at the end, maybe based on Instagram post sentiment analysis.

Design

Increase app reviews by 12x

Some people feel uncomfortable iOS controls the number of times a review prompt is displayed, this is a reminder fewer prompts, at the right time work best, users ignore frequent desperate calls for reviews.

Key takeaways

Prompt at the end of a happy journey
Review is done in-app via Apple’s prompt
Reviews are real, not rewarded/incentivised

Results

Don’t disable buttons.

I’ve made this mistake before but this tweet changed my mind: Why * Inline form validation doesn’t work for all field types (e.g. multiple selection fields) ˜ related article * Disabled buttons don’t clarify next steps * Interactive disabled buttons are unintuitive What to do * Use hint text, inline validation, and leave buttons enabled

How to use colors in UI Design

60–30–10 Rule: 60% is your dominant hue, 30% is secondary color and 10% is for accent color;
Color Psychology: Colors carry meaning;
Grayscale-first: Forces you to focus on layout instead of color hues;
No pure black or gray: Absolutes feel unnatural, there is no pure black or gray in nature, always add a hue and some brightness;
Color scheme generator: coolors.co.

Process

Putting your users first with problem-centric roadmaps

Ever realize you’re designing a feature that doesn’t make sense like it did when you put it on your roadmap 6 months ago? If your product roadmap is feature-based, then this probably happens a lot.

The reason I generally dislike planning artefacts where features sit at the high-level is that they don’t allow you to adapt to unexpected complexity/events/value changes. Janna argues you should instead have:

Company value: problem and USP
Objectives: company value metrics: Pirate Objectives - AAARRR, ARPU, etc
Themes: problem to be solved, NOT feature/release/promise
Time Horizons: Now/Next/Later

Research

74% Of Consumers Prefer Surveys With 5 Questions Or Less

Source: "Trends in Nonresponse Rates, 1952–1979" by CHARLOTTE G. STEEH: https://academic.oup.com/poq/article-abstract/45/1/40/1934469

Better Insights from Satisfaction Surveys

Careful with equating quantitative data with objectivity, price is objective but satisfaction isn’t.
Averages scores (e.g. NPS) aren’t actionable, hide potential improvements
Behavioural data is essential to understand what satisfies people
Segment feedback by usage frequency and satisfaction (happy infrequent users require a different approach from unhappy frequent users)
Replace NPS with Product-Market Fit
- Would you be disappointed if X was not available? De-prioritise "disappointed and don’t use".
- Who benefits from X?
- Best thing about X?
- How to improve X for you?
Interviews/qualitative data can produce richer insights from fewer users (related: research calculator from a few issues ago)

The Ladder of Evidence: Get More Value From Your Customer Interviews and Product Experiments - Product Talk

Here are the key takeaways from the ladder of evidence:

Avoid speculation — don’t ask people what would they would do, it’s easy to do but unreliable. Past behavior is better at determining future behavior.
Ask for specific stories about the past to generate insights. — You’ll get a lot of unstructured data about people’s context
Look for evidence of action to evaluate solutions. — Through quantitative data, competitor analysis, prototype testing.

Security & Privacy

Stop the Open Data Bus, We Want to Get Off

A research paper demonstrating how there’s no such thing as anonymised user data. Researchers were able to identify politicians and specific individuals based on just a few data points.

The only way to get anonymous data is not to collect personal data at all, use differential privacy (add noise, sampling, choose hours instead of minutes), and do on-device processing.

Detailed personal data is a bad predictor of conversion anyway so you're just collecting legal liability. My age, name, location don’t strongly correlate to my cat ownership (cat toys), upcoming stag do (fancy dressing), or food preference (roast). No wonder adtech conversion rates never go beyond 7% and often sit in the 3.5% – as good as firing a shotgun in an open field.

An exception would be spending millions to bombard users with your propaganda in the last few days before a vote, but we don’t talk about that.

Stats

Global Time Spent in Shopping Apps Grew 45% Since 2016 (~30min/week)

The report does a lot of hand-waving to make numbers look bigger than they are, classic marketing and not helpful if you need reliable data.

Fortunately you're in luck, I looked at their concrete 2016 numbers (10hrs/year/user) and applied a 45% increase in 2018, assuming continued growth in 2019/2020.

FYI: Photography apps have seen the biggest increase since 2016 (210%), followed by video players and editors (125%), no wonder companies like Lighttricks (photography and video editor apps) have a 1bn valuation.

Strategy

Are metrics undermining your business?

Unlike most articles ending in a question mark, the answer to this one is very likely: Yes.

Andrew Chen, partner at a16z, argues data can only help you make small improvements to high-traffic journeys.
Data-driven doesn’t help overcoming the Innovator’s Dillema
Innovation requires pairing data with research and making bigger bets/changes to your product
Data does not predict long-term retention, satisfaction, infrequent important features (e.g. account deletion)
Beware of vanity metrics: impressive but not actionable (e.g. registered users, sessions, downloads)

And Finally…

Tweet by Outstar on Twitter

Don’t watch this at night before bed.

Where does `Upper Case` and `Lower Case`come from?

Do you know?