In 2014, Facebook changed its mantra for developers from “Move Fast and Break Things” to “Move Fast with Stable Infra”. Only moving fast doesn’t bring the business benefit home. At a certain tipping point the engineering organization introduces more “broken things” than they’re able to fix in the long run – speed of the organization stalls or declines.
The new mantra relies on two cornerstone principles Release Quickly and Often and Release with Confidence. Releasing quickly and often leads to small and quick iteration cycles. These foster quicker learning and fuel innovation. Releasing with confidence keeps damage away from your brand, your revenue streams, your customer base and your employees. Confidence refers to accurate software, reliable and well-performing releases, scalable software and released software without privacy or security concerns.
A good example of a well performing product organization is Google Search. They run a total of 15 product teams. They work on over 2.000 product ideas – no product optimizations – in Discovery with less than 500 built in Delivery. The ratio is 4:1 – 4 ideas in Discovery, 1 get build. They aim for 10.000 ideas in Discovery with less than 500 being built. See “Rigorous testing” from Google for further information. Google mentions in this article they did 595.429 Search quality tests, 44.155 Side-by-side experiments, 15.096 LIve traffic experiments resulting in 3.234 Launches – in 2018 alone!
The industry benchmark in well operated digital product development organizations is to kill 75% of ideas in Discovery. If not a minimum of 50% of the ideas are killed then it’s Design, not Discovery and a clear signal of a malfunctioning organization.
Issues of Conventional Agile
“Agile is all about building and delivering software, but it says almost nothing about how to come up with a valuable product backlog.”
Agile software development helps a lot, but doesn’t say anything about building a valuable backlog. It is completely blind on the value side, answering the what should be built. Marty discusses a solution to this issue – dual-track agile – on his blog.
Issues of Conventional Agile:
Thanks to agile, teams move faster. But agile doesn’t provide teams with a valuable goal.
Moving fast without a goal leads to feature chasing. The goal of agile is to produce more output, not necessarily more business results or more value. This is especially tricky with mature products where removing features might provide more value to the business and customer (use A/B testing to understand the value). Also in mobile products more features most often fail – they distract more since there is too less inventory – and finally don’t have an impact.
Engineers are most time only coders and don’t contribute further more to the overall value generation process.
In agile it’s very hard to predict any dates and make commitments since it’s based on user stories only.
UX Design suffers because there is simply no place for UX in agile. The sprint iteration doesn’t reserve time slots for UX to work as part of the iteration. If engineering is ready to go UX should be ready as well but usually start right now to work as well.
The architecture of the system suffers as well. Fast moving teams produce more technical debt. Without really knowing good ways to solve an issue, suboptimal long-term solutions are implemented and increase the need to repair architectural decisions.
The rest of the company – marketing, sales, customer support … are brought in too late into the process.
Salesforce rolled out agile by the book about 15 years ago. As a consequence the whole UX & Design team threatened to quit because conventional agile didn’t include UX & Design at all. So, after some discussion Salesforce included UX & Design in their agile working model and pivoted from the Conventional Agile approach.
Continuous Discovery and Delivery – Dual Track Agile
“Discovery, by definition, means you don’t know the answer when you start.”
Ed Catmull, Co-founder of Pixar
“Dual-Track Agile” emphasizes the parallelism of Product Discovery and Product Delivery. They happen all the time in parallel. The Discovery track is all about fast learning, building a validated product backlog. Delivery is all about creating software that could be released with confidence, shippable software. The Discovery track is led by Product, Delivery by Engineering.
Product Discovery provides missionaries with purpose. In Discovery, the key risks are addressed and meaningful answers to these questions give guidance to the team. The given answers are all validated, there is enough evidence proving it will work.
Will they buy it? (Value)
Can they use it? (Usability)
Can we build it? (Feasibility)
Can our stakeholders support it?
Bonus question: Should we build it?
When all critical questions have good and validated answers, describe what to build in Delivery. Therefore, use e.g. JIRA stories with prototype and UX design (prototype-as-spec).
It is, however, not needed to validate all risks all the time in discovery before putting anything into the backlog. The trio of people (PM, PD, TL) does the risk assessment (takes 5 minutes talking). Pick and choose the techniques to validate if needed. Don’t be religious about the process.
Example: At ebay, one base assumption is: create a safe place for transactions. This assumption is the root cause for the overall reputation system within ebay.
Discovery and prioritized backlogs
“Don’t prioritize your backlog – simply try thousands of ideas!“
Follow only those ideas in favor of the overall business objectives. Don’t spend too much time thinking about them, just try them. One iteration in Discovery means: One new idea or another approach to an older idea.
Put the ideas that work into the product backlog. If the product backlog is empty switch to “feed-the-beast” mode and allow the teams to work on e.g. technical debt or other meaningful tasks. If the product backlog is too full – means more than 2 weeks of work – you have a rotten backlog. Here, ideas will no longer be valid or well memorized if too much time has gone before going into Delivery.
“The most important thing is to now what you can’t know.”
1) We know
we can’t count on our customers to tell us what to build
Avoid focus groups. Customers don’t tell you what to build because they simply don’t know what is possible with latest technologies in place. Furthermore, they don’t know what they actually want until they see it – hence use prototypes during your customer feedback sessions. Customer-facing product teams go 3 hours per week to see the customers. “Customer inspired – customer enabled”.
The iPhone wouldn’t happen with focus groups. Skype hired Nokia, Motorola and Apple people and they compared the different approaches to innovative products. It took 3 ½ years to build the iPhone. Nokia and Motorola were doing focus groups when Apple started the iPhone. Palm has just released the Treo with Touch Screen and people didn’t like it. The focus group gave clear indication to skip the touch screen. So, at the time, Motorola and Nokia decided not to build phones based on touch screens. Apple hat a vision – the iPhone vision – “Build a touch screen powered phone.”
2) The most
important thing is to establish value
Value for customers, creating value for customers. Typical product roadmaps line up features. The underlying assumption of feature roadmaps is that there’s value in the feature and business viability is granted. But that is typically not true.
recognize that engineering is hard, but the user experience is often even more
difficult, and more critical to success
This is especially true at B2C companies – 50 engineers and 2 visual designers don’t work out. The experience and value perceptions created at the customer is vital – UX is more critical to success than technology.
recognize that functionality, design, and technology are inherently intertwined
expect that most of our ideas won’t work, and those that do will require
iterations never make it beyond us – the team; they’re simply stopped
internally in early stages.
iterations should one follow before skipping the idea?
Depends on the importance of task
Ask yourself: “Are you still learning with every iteration?”
Change the approach to the problem (e.g. churn)
Timebox: 2 days for one approach
Avoid the “fall-in-love”-pitfall with design: use less than 2 days for design before result is shown to users.
Example: Ceramic class separated into two groups with separate goals. The first group should build one high quality pot. The other group is instructed to build as many pots as possible – it’s only the weight that counts. The second group won the quality challenge because of practice! The first group wanted to produce the “perfect” pot but failed with just 1 mediocre pot.
6) We must validate our ideas with actual customers
Go out of the building, talk to real customers. It’s valuable to start with your team mates, colleagues, people working inside your company – but finally, you need to get the opinion of your real customers.
validate our ideas as quickly and cheaply as possible
even in hardware: 15 iterations per week. Google Glass team build an engagement
model prototype and improvised a futuristic user interface with simple
technologies. They quickly learned about a key problem: shoulder sourness and
skipped the key assumption for success.
8) We use
both quantitative and qualitative techniques
Why is it happening?
Example: Etsy switched to endless scrolling from pagination. The A/B-Test showed people buying less. The qualitative analysis showed: there were simply too many cool things to buy and people couldn’t decide – paradox of choice
9) We must
validate both technical feasibility and business viability in product
discovery, not after
engineers ways before planning. They need to be involved and see what to build
before sprint planning.
all about shared learning
Shared learning happens all the time in a co-located team. Discussions over coffee, exchange of opinions at the work-desk, discussions are all around. Furthermore, allow engineers 30 minutes playtime with the prototypes each day to let them express their concerns and start having good ideas on how this could be build.
With a distributed team in place, schedule a 30 minutes meeting every day to allow the exchange of ideas, to give engineers air time with the prototype and help them understand the ideas behind.
Big companies run innovation labs for product discovery. But good ideas never materialize due to the separation of Discovery and Delivery.
This blog post is part of a series. It summarizes my personal notes of the workshop held by Marty Cagan “How to Create Tech Products Customers Love” from 5th to 6th of June in 2019 in San Francisco.
The Opportunity Assessment is enough 90% of the time.
What business objective are you focused on? typically one of the OKR objectives
How will you know if you have succeeded? typically one of the OKR key results
What problem are you solving for our customer? do you really know it’s a problem?
Who are you solving that problem for? typically a target market or persona from the strategy
Internal Press Release
The Internal Press Release is not intended to go public – it’s for internal use only. The anticipated audience is new product’s customers – actually it’s the team, management and stakeholders. It’s typically 1.5 pages maximum and written in “oprah-speak”, not “geek-speak”. Sometimes, 3-4 pages of FAQ for anticipated questions are added. The structure of the internal press release looks like this: heading, summary, problem, benefits, quote from you, customer quote, closing / call to action
Amazon uses Internal Press Releases for big ideas / efforts (e.g. site redesign or moving into new country).
An alternative to the Internal Press Release is the Happy Customer Letter describing the benefits for a customer written by the customer and the CEO-letter describing benefits for the company.
User Story Mapping helps to visualize and deconstruct the problem or solution space. It provides an holistic view and gives context. Through the collaborative process it encourages shared understanding, identifies holes in thinking and improves planning and estimates. Furthermore it heavily influences prototypes and actually helps to scope releases for the product.
The basic idea of the Customer Discovery Program is to discover and develop a set of reference customers in parallel with discovering and delivering the product. At the stage where the reference customer signs up for the program there is no product ready to be delivered. The Customer Discovery activity makes only sense for real big efforts, absolutely not for features. Serious enterprise customers are very likely to sign-up because they’re burned by the practices of Oracle, SAP and the likes – sell and run.
The reference customer bought the product without any side deals, runs the product in production and loves it enough to tell the world about it. In Customer Discovery, we’re looking for “Earlyvangelists”. An Earlyvangelist is best characterized by these criterion:
With the Customer Discovery Program it’s simple to tell if product / market-fit is reached: achieved if 6 reference customer for a single market segment (e.g. industry, geography, …) are found. The product / market-fit product is the smallest possible product that meets the needs of this group. If you find only 4 customers or less this means the product market fit is invalidated and a pivot is needed. Work with 5-6 companies – and not more than 8. Talk to as many as possible – e.g. 50. Select only the most attractive for the Discovery Program, the other customers go into the beta program. Agree with the selected companies to be a discovery partner and ensure the right level of access to people and input. Agree with them to become a public reference if they like the delivered product.
In Enterprise business the goal is to find a single product solution that fits all discovery partners. Again, it’s important that all partners are from one single market segment. In Consumer services you should identify 8-20 Earlyvangelists and agree with them on regular phone calls to synchronize.
Examples for customer discovery: OpenTable (SMB), Symantec (Enterprise), Bazaarvoie (B2B2C), xoopit (Platform Services), Apple (internal tools), BarkBox (Consumer Service)
Customer interviews are needed to understand your customer, to get rid of assumptions and start working with facts. Marty summarizes the value of interactions with customers in his post “Don’t talk to customers?“. Below are the key questions to answer:
Public API’s (let others innovate on your product) – be aware of bad usage: Cambridge Analytica + Facebook
Hack Days (directed and undirected)
Data Spelunking (Hackathon on data)
The prototype should minimize the time by factor 10 to provide something to look at. See more on prototypes at Marty’s blog “Flavors of Prototypes“.
User prototypes are created very fast and lightweight by nature. The prototypes are used for value testing with a consumer or customer to quickly gather feedback on both, usability and value. Low fidelity user prototypes are used for team internal iterations. Use high fidelity prototypes to show-case internally to executive people. The prototype is usually created by the Product Designer with support from the Product Manager. Ideally, when finished, the prototype could be used as a specification for the Delivery process – “prototype-as-spec”.
The feasibility prototype validates the solution approach. Usually, the prototype is build by engineering to gain further insights on the implementation and test technical (e.g. scalability, performance) risks. The prototype might not be more than a code fragment or a successful validation of a 3rd party software or API integration. It may also happen that product people are not involved in the prototype at all.
Live Data Prototypes
The purpose of the live data prototype is to collect further evidence pro or contra a product decision. This prototype is more expensive to build than the user prototype, but still far less than the actual product. The prototype is not the real product, it’s usually 5-10% of the real product. It includes quantitative a/b-testing but also qualitative testing and is based on real data. A small amount of real traffic could land on the prototype to collect data. Engineering is typically needed to create the live data prototype within 2 days up to 2 weeks.
A lot of people get excited when they see the live data prototype and tend to confuse the prototype with the real product. But there’s still a significant difference between production ready software and the live data prototype. The real product needs:
All required use cases
Instrumentation / analytics
Scale and performance
Internationalization / localization
A good example of a live data prototype is Amazon’s “Frequently bought together” feature. The idea of building the features was rejected by SVP Marketing. So, strong evidence was needed because it was simply too expensive and risky to productize the feature without further evidence. So, the team decided to build a live data prototype with a small amount of real traffic in a specific product category. The prototype was a/b-tested and the collected data provided a significant uplift in business KPI’s. This is a great example of a high-integrity business case.
Hybrid prototypes mix those elements needed to tackle the specific risks at hand. They blend various techniques and are mainly limited by your own creativity.
A good example of an hybrid prototype is Zappos. Zappos solved the problem of female shoppers to buy fashion shoes online. They defined and understood their personas and their key problems with shopping online: 1) returning goods 2) no timely delivery 3) not knowing the size 4) bad product images. Zappos prototyped a potential solution to the persona’s problems by mixing a variety of prototypes: user prototype (appealing front-end), live data prototype (product catalogue and images) and the “Wizard of Oz” (buying the shoes at the shop over the street and delivering to customer). Most important was: the users shouldn’t recognize the prototype character of the solution. Zappos controlled the amount of traffic via AdWords and made sure they could handle the manual part. So, with this mixture of prototypes – that for sure doesn’t scale – they could validate demand, value and usability.
Testing Product Ideas
“Prototype as if you know you’re right, but test as if you know you’re wrong.”
Marty writes in his blog about “Prototype Testing” more detailed on the various ways to testing.
5. Testing Usability
Usability testing includes interacting with customers, getting their feedback. For the test session, have the prototype ready – up and running and focus on the prepared questions to understand if users have issues using your product. The session may be conducted at your office, the customer’s office or at a mutual convenient location (e.g. Starbucks) or – if not possible – remote via video conference.
Recruiting users in B2B context is done via the customer discovery program, in B2C via AdWords. AdWords allow acquisition of users based on keywords and / or geo-targeting. It’s the most cost-effective solution and easy to stop and restart again. Payment is entirely based on performance.
6. Testing Value
Testing value focuses on three aspects: testing demand (is this really a problem?), testing efficacy (how well does the product solve the problem?) and testing response (how excited are the testers?).
Testing demand answers the question if people are willing to use the product, if they understand the value and see the product solving a real problem they have. Marty talks more about Desirability Testing on his blog about “Product Validation“. Some techniques for demand testing are described below
Fake Door Test
A fake door test fakes a product feature for the customer. If the customer acts with the fake product a thank-you-message is displayed and sometimes contact information is collected. Furthermore, nothing happens. The goal of the test is to collect data, to measure the click-thru ratio. More information on the fake door test can be found here: http://learningloop.io/plays/fake-door-testing.
Landing Page Test
A landing page test pitches product features, products, product lines or other promises to the customer combined with an explicit call to action. Push traffic on the landing page via e.g. AdWords or other comparable methods. Now, measure the conversion – how many people do actually interact with the landing page and are interested enough to follow the call to action? With the click, nothing happens with the customer other than a friendly thank you message and sometimes the question for contact details. More information on the landing page test can be found here: http://learningloop.io/plays/spoof-landing-pages
The explainer video shows a high fidelity prototype at work. It’s basically a video of a product demo. It is then distributed and measured like the landing page test above. The goal, again, is to measure demand for the demoed product. More on the explainer video: http://learningloop.io/plays/video-demo
A great way to test a product idea without jeopardizing your brand is to test demand on kickstarter.com. Just place the product idea at the crowdfunding platform as a “nobody”. If the idea creates enough buzz it’s worthwhile a further investment, if not it can be dropped silently without creating any noise. Read more on the idea from Mark Dwight “How to Kickstart Your Market – Why even established companies can use crowdfunding.“
Qualitative Value Testing
“Find everything that’s wrong with the product and fix it; Seek negative feedback.”
Qualitative testing explains why it’s working or not, it gives insight why something is happening or isn’t. It doesn’t try to prove anything. You won’t get the answer from any one user test; every single test provides another piece of the puzzle. It’s important to test with real users and customers to judge the value.
Qualitative value testing is done with prototypes or the real product. It provides insights from usability and value perspective. On top it usually provides unexpected insights from the customer. It’s typically done fast and cheap. To really understand how much customers value the product, various questions or tests can be conducted:
Payment – will they pay for it? credit card information, pre-order form, letter of intent (in B2B)
Reputation – will they recommend it? NPS, introduction to peers or the boss, public reference
Time – will they meet again? Will they invest their time? agreement on follow-up meeting, non-trivial trial
Behavior – will they switch from their current solution?
Quantitative Value Testing
“Features are not inherently valuable. The value for our customers is only realized when a feature fulfills a need. It’s only realized for our business when we see the results of our work move the needle. That’s why we need to concentrate on the outcomes over the outputs.”
Quantitative value testing can provide evidence or even proof that something truly works – or isn’t. It generally can not explain why it’s so. It’s done to get a clearer picture on the impact on your revenue, your brand, your customers and also your employees.
Quantitative testing can be done with the existing product – or a live-data prototype – in an A/B testing setup on a certain amount of your traffic. Alternatively, it could be done with a limited amount of your customer through invitation. In a B2B scenario, you’d use your existing customer relation via the Customer Discovery Program to get exposure of the test to people.
A good example for a quantitative value test is Spotify’s “Discover Weekly” feature. Data collected in an A/B test was compelling enough to fully implement the feature. The launch-ready implementation included some big hurdles and a lot of effort on the data crunching side. So, it was well worth the effort to test – before – putting the feature in Delivery.
7. Testing Feasibility
Testing feasibility mainly addresses technical concerns – are we able to build this at all? To test feasibility it’s important to create prototypes that focus on the key areas of concern. Emphasize speed of learning over reuse of the written code. Your Tech people need to answer these questions:
Do we know how to build this?
Do we have the skills on the team?
Do we have enough time?
Do we have the right architecture or
Do we understand the dependencies?
Will the performance and scale meet
Do we have the infrastructure to
test and run this?
8. Testing Business Viability
Business viability – does this feature / product fit with our business? – needs to be addressed to be successful within the own organization. Your stakeholders (e.g. Senior Executives, Sales, Marketing, Finance, Legal, Security, Business Operations, …) need to be informed regularly, you need to earn their trust. Have discussions with them and make them feel you understand them – but remember: everyone has a voice, but not a vote! Try to engage individually with them, group meetings can cause a lot of damage and are harder to handle. When talking to your stakeholder, have your data ready – data always beats opinions. And read the signs during the meeting – differentiate between stop signs and yield signs.
Techniques you can use for a stakeholder meeting is typically a high fidelity user prototype and / or a product walkthrough.
This blog post is part of a series. It summarizes my personal notes of the workshop held by Marty Cagan “How to Create Tech Products Customers Love” from 5th to 6th of June in 2019 in San Francisco.
One transformation technique Marty recommends is the Discovery Sprint. He recommends to do a Discovery Sprint when a team struggles to learn how to do product discovery, when the team has something big and critically important to solve or if the team is just moving too slowly. Marty talks more about it on his blog https://svpg.com/discovery-sprints/ and refers to the Book “Sprint” by Jake Knapp et al.
Another is named Pilot Teams. The idea behind the Pilot Teams is to create success within a smaller protected environment and convince doubtful or fearful or lazy people to follow the change process. The principle is borrowed from the technology adoption curve (aka “Gartner Hype Cycle”) – some people are early adopters, others are less eager. Chris Jones from SVPG talks about this technique in “Pilot Teams“. With these pilot teams the idea of A/B testing – well known from product development – can be applied to organization development as well.
Outcome-based Roadmaps is yet another way to start the transformation process. Simply continue working with product roadmaps, however introduce two differences. First, annotate every roadmap item with its associated expected business result. Every time this item is discussed highlight the expected business result. Second, after the launch of an roadmap item report immediately the actual result vs. the expected result. So, during the next 3-12 month the opportunity assessment information will get its way into the roadmap. For prioritization try to move away from prioritizing ideas to problems.
Variable compensation models and motivation – an experience report
We’re quite a young and digital organization. Our shareholder – a big publishing house – demands variable portions of the salary as a motivational factor. If that’s set in stone, you better look how to implement it best. Since 2015, we implemented variants of a variable compensation model and I’d like to share some of our learnings on the various models.
Salaries in the knowledge workers’ world
Today’s salaries usually have at least two components: the fixed part (paid usually every month) and the variable part (paid usually quarterly, twice or once per year). The fixed part represents the compensation for the working hours and fulfillment of the work contract. The variable portion can vary according to the agreed targets. Depending on the degree of achieving the targets the multiplier for the variable portion may vary from 0 to 1.5 or even higher.
Motivation and knowledge worker
My absolute favorite to understand motivation – and the impact of money on knowledge worker is the video by RSA describing Dan Pink’s thinking behind his book “Drive”: https://www.youtube.com/watch?v=u6XAPnuFjJc (10m 47s definitely worth watching!)
Motivation = three major ingredients: Mastery, Autonomy, Purpose.
Mastery is an intrinsic driver of knowledge worker. They want to improve their skills, they want to get better and better. A great workplace environment takes this into account and leaves room for people to practice, practice, practice. Autonomy gives people space to solve the problems they’re working on. Nobody tells them how to do so, nobody’s looking at the output. Only the outcome counts. Purpose sets the work activities into a greater context. Everything has a meaning, people understand why they do what they do. They see the greater picture, the vision. Purpose, in my opinion, is the most important ingredient for motivation. Put all three together as a fundament for a work environment and you have very likely intrinsically motivated people.
Money, on the other hand, as a benefit for work kills motivation. It is nice to receive your variable salary once, twice or 4 times a year. It’s a nice way for your boss to say “Well done, thank you!”. BUT the money is not a long lasting instrument to create or increase motivation (look at 2:12 into the video – they present research results there …).
So, why do we still have variable salaries to push motivation and output?
I know quite some young companies, some startups, some larger organizations who operate 100% on fixed salaries. They understood the basic principle of motivation and compensation. On the other side there are still some old, traditional and rusty organizations with variable compensation plans. Some decades ago, they connected the compensation to personal goals and never questioned themselves. Or it’s so common, they can’t even think about getting away from this model.
2015 – The “yearly revenue and individual targets”-model
Management sets a revenue target and every employee and his/her manager agrees individually on targets. The thesis behind the model: money is a key motivator to achieve top performance. The variable part of the salary is 100% connected to a company revenue goal and one or many individual goals.
We observed lengthy and excessive negotiations on individual goals with our employees. Even worse, the just-agreed goals hold just 4-6 weeks until they need refinement. Furthermore, we observed individuals stating they couldn’t help each other because this action would directly conflict with them achieving their goals. In essence, the model leads to people optimizing their personal benefits and defocusses company goals.
2016 – The “yearly revenue and department targets”-model
Management sets a revenue goal and each department head sets a yearly department target. The department target (e.g. “Reduce page load time to less than 2 seconds”) holds for the whole year and is always present. It will influence the way people work together but is not always a focus topic. The department target occurs in the variable compensation plans for all department employees. Thesis behind this model is again: money is a key motivator to achieve top performance. The variable portion of the salary is again connected 100% to the company revenue goal and one or many department goals.
Applying this model we observed less conflicts between individual employees. It’s quite time consuming during the identification of the department goals. They need strong alignment amongst each other. We managed to achieve quite good alignment – however had some occasions where people ended in conflicting department goal discussions. In the end the whole staff focussed ways more on achieving the department goals – better than with the previous model. However, the organization didn’t “feel” aligned on joint goals, more trying to achieve the department goals. Producing shiny winners on the cost of the overall company targets.
2017 – The “yearly revenue and quarterly company targets”-model
Don’t name the model OKR. We successfully burned OKR in 2014. Tried to implement it without external, experienced help. It ended in a process-by-the-book implementation and a perception of a grass-root democratic, inefficient and cluttered way to set 4 management and 4 team targets per quarter.
In this model, the management sets a yearly revenue target and company targets for the next 3 month. The company targets include everybody in the company – no matter what function or department. The targets need to be S.M.A.R.T. (Specific, Measurable, Achievable, Relevant, Timebound) and led to quite some discussion during definition – both between department heads and employees. The targets include e.g. specific product features, specific sales products or marketing activities. Thinking behind this model: everybody sits in the same boat and again money is a key motivator to achieve top performance. The variable portion of the salary is connected 100% to the company revenue goal and quarterly company goals.
During the implementation of this model we found “same company, same targets” led to some motivation for the active influencer for the goals (e.g. product, technology, sales, marketing) and quite some frustration for those having no influence at all (e.g. finance, HR, administration). They had to rely on their colleagues delivering the best job possible. On the other hand, we were able to set relevant goals for the next 3 month and were able to steer the company in a clever and agile way through the rough sea of economical changes. We had the staff members being focussed on achieving few topics of high relevance for the company.
We also learned – the hard way – that goal achievement communication can lead to some confusion and irritation if not done 100% transparent. And setting quarterly goals leads to quite some overhead to define goals every 12 weeks. Agility comes with a price tag!
2018 – The “100% guaranteed and 150% possible”-model
2017 ended with some really bad environmental messages for our business model. We needed to change quite some things. Amongst them was the variable compensation model. Our ambition at the time was to bring maximum calm to the staff members and allow them to focus on the company’s focus goals. One measure was to put away the variable portion of the compensation model. We guaranteed 100% of the bonus and made 150% possible if we achieved a specific traffic target earlier than we expected it.
Thinking behind this model is (see Dank Pink above): money doesn’t have any influence on work performance, but it has on work morale. We decoupled the compensation from achieving targets – allowing people to work on the company’s focus topics. And as a bonus, there is this 150% stretch goal. Nice to achieve – and desirable – but it just sits there.
We observed almost no discussion on compensation and fairness / unfairness of goals. People focused on getting the job done. I’d refer to the state of people as “intrinsic motivated”. At year end we didn’t manage to catch the 150% goal which led again to some frustration amongst the team. Furthermore, some specific departments (e.g. sales) perceive “100% fixed” less as an adorable state. For them it’s less motivating since their working model always follows includes “catching numbers”.
2019 – The “revenue and EBIT goal”-model
Beginning 2019 we found ourselves in a more stable environment and switched back to a variable compensation model. This time, we decided to focus on setting company wide financial goals. Everybody is able to influence them and the effort setting them is limited. The goal is set end 2018 for the whole year.
The thinking behind this model: money doesn’t have any influence on work performance, but it has on work morale. At a first glance, the model doesn’t look like an advancement but it effectively decouples financial goals from specialist topics. It’s the same for everybody and done once a year. So far, so good. It turns out to be a bit problematic since the environmental conditions changed quite a bit and the target corridors defined end 2018 are no longer achievable. So an adjustment is needed!
Which model worked best so far?
The 2015, 2016 and 2017 models worked better and better each year. Still having significant flaws with the direct connection between motivation and money paid. But the got better.
2018 was the most successful model so far. Less friction, lots of motivation and high pace – outcome over output. But we also had some frustration in performance-oriented departments (e.g. sales).
2019 doesn’t feel like an advancement from 2018. But 2019 is not over – let’s see.
The holy grail? Well, I don’t think we found it – so, we need to move on and adapt.
Software projects failed a lot in the past. They failed to deliver the value for the business, were too late or ways out of budget. The selected process method was usually the scapegoat for the failure with agile methods being the answer to any question in software development. But as usually in live, it’s not black or white. The selection of the right software process method depends on the surrounding of the project. I gathered some industry input and combined it to reflect the current thinking regarding agile software development methods vs. traditional methods.
The adapted Stacey matrix
The original stacey matrix supports decision making processes suggesting appropriate management actions and defines four areas: simple, complicated, complex and chaotic. The suggested actions depend heavily on the context of the decision making.
The dimensions of the adapted matrix: HOW and WHAT
The x-axis of the adapted matrix deals with the HOW. If the team knows the technology well and has used it many times before, we’re on the left. Otherwise, if the technology is completely new to the team we’re on the right of the dimension. The y-axis positions the WHAT. On the bottom of the axis, the stakeholder of the project all agree on the goals and have the same understanding of the expected outcome. On top it’s the opposite, no agreed requirements and no alignment on expectations. The individual mix of the project points to a certain area with a process model suggestion in the adapted matrix.
Waterfall is a traditional project management method with sequential steps and no iterations. Massive upfront planning is done before any implementation work starts. If all goals and steps are clear, waterfall produces consistent results in a predictable and repeatable way. The clearly defined tasks lead to an optimized sequencing and optimal resource allocation. Waterfall optimizes resources and return on invest if cause and effects are clear to anybody in the project team.
… vs. Agile
Agile stands for SCRUM, Kanban and LEAN methods with flexibility, quick response and constantly changing environments in mind. They start quicker with smaller scope for the current increment with the scope being like a rolling window. Uncertainty of the projects’ goals needs quick adjustment and adaptation during the whole execution. Only close and frequent collaboration with all team members make agile projects successful. If causes and effects aren’t clear, agile works in small steps towards a value-generating and broadly accepted result.
Simple = easily knowable, the known knowns
Projects in the simple zone unveil very few surprises, decisions are fact- or evidence-based, advancement occurs in orderly, sequential steps and the WHAT is clear to anybody. Any size projects with clear activities and repeatable results fits in this category. It has been done multiple times before and best practices exist as benchmarks. The process is simple and could be handled in a check-list style.
Going forward in a simple, fully predictable project means reducing it to the maximum to make the single pieces easier to understand. Examples of simple: recipes, tasks on an assembly line, checklist based work.
Complicated = not simple but still knowable, the known unknowns
The complicated zone segments into socially and politically complicated and technically complicated. Complicated means less simple but still somewhat predictable.
In Social/political complicated environments people can not agree on the purpose of the project and the expectation on results is not clear. Requirements are conflicting amongst the diverse stakeholder which could be resolved with waterfall to get clearance on WHY before WHAT before HOW. On the other hand applying agile techniques could help convincing stakeholders to agree on already achieved results and smoothing the further requirements discussion. The project team needs to pay special attention on getting early agreement between stakeholders in place.
In technically complicated contexts it’s clear on WHY and WHAT to achieve. Still, the HOW is not clear. An agile iterative approach helps getting feedback from the project team on the achievements making adaptations possible.
Going forward in a complicated project means as well reducing it to the maximum to make the single pieces easier to understand. Technically complicated is e.g. using a specific technology for the first time. Political/social complicated is e.g. if the relation between cause and effect are not clear enough or conflicting opinions amongst various stakeholder exist.
Complex = not fully knowable but reasonably predictable, the unknown unknowns
The complexity zone stands for high risk and uncertainty and requires a high feedback frequency. Neither requirements nor the execution are clear. Holistic defined process methods don’t work any longer. The context asks for a more explorative approach with transparency, frequent inspection and adaptation. SCRUM as a process method in the toolbox of the agile mindset is the method of choice. It increases transparency with small iterations and frequent check-points allowing cheap adaptations. The team planning is the start point for each new iteration and allows immediate feedback from stakeholders to the teams to adapt the next iteration.
Complexity can not be reduced, some understanding can be achieved and complexity can not be planned, it simply grows. A good example of a complex project is software development in general. The requirements are rarely fully defined right at the beginning and it’s seldom clear which architectural solutions are superior to others.
Chaotic = neither knowable nor predictable, the unknowables
In chaotic zone requirements and execution path are both undefined and the risk is high. Kanban as the most flexible project management method is the tool of choice. With no structure like sprints and the only focus on work in progress (WIP) Kanban focuses on continuous delivering results to allow further modifications in direction and backlog items.
The goal is to move from chaotic towards complex by dividing the problems. The principle “Act, Sense and Respond” helps navigate towards the zone of complexity.
Innovation is usually part of agile product development methods. Sometimes, however, agile methods just replace other methods. SCRUM replaces Waterfall, KANBAN formalizes previously unordered work. Obviously, the innovation dilemma remains still open. Where comes the creativity from? The ideas? Where to test those hypotheses which are not part of the daily routine?
The hackathon as innovation tool
Some organizations run hackathons once or multiple times a year. We did and are doing this as well. We organized already 6 hackathons in the past. Once yearly. Did we see the innovation boost? Well, yes – and no.
How do we organize a hackathon?
A hackathon at gutefrage.net is a timeboxed activity (usually 2,5 days) and surrounded by a lot of social activities. We cook, we bake, we experience Virtual Reality, we do some board games, we play the football table and have a good time and fun. The whole company participates usually and is excited to validate hypotheses which are usually not part of the product development. There are no limits from a topic perspective. Teams organize themselves via a democratic voting exercise right at the beginning. People pitch their ideas and convince other people to become part of this specific project group.
What’s the typical outcome?
During the hackathons at gutefrage.net one out of five ideas launch during the hackathon. This one idea is production ready and creates value right from the launch. The other ideas typically proof aspects, create prototypes of various qualities, cover maximum the best-case implementation and still need an investment of 80% to be ready. The hackathon is a great team building event, it’s great for the morale, the culture. The hackathon drives people’s motivation and frustrates them if their project doesn’t make it into the finals.
What issues do we see?
The hackathon validates some hypotheses, some not and the question remains open what to do with all the started work? Will we follow some traces? Will we just abandon the work? Needless to say – after the 2,5 days hackathon there waits daily business in form of agile software development work. At gutefrage.net we promised to launch the winning idea and typically abandon the remaining work. We found that’s not the most efficient way to drive innovation.
The RocketLab – our way to innovate
As a learning organization we drew some conclusions from the hackathon experience. Mixed teams with participants from all relevant areas worked very well. The one project going live was a real push for team and company motivation. The others weren’t as good for morale. Thinking about this for a while we came up with a slightly different format – the RocketLab.
What’s different between RocketLab and the hackathon?
The RocketLab stands for outcomes and can potentially solve any topic: Product, Technology or others of cross-discipline relevance. At day one of the RocketLab there is one specific hypothesis the team focuses on. The team exclusively works for a defined period solely on solving this one issue. No distraction, just 100% focus. The team contains all disciplines to solve the issue on hand. They are all committed to create the best solution possible within the given time budget. It’s a team of maker, not talker, not theorists, no visionaries. The hackathon is broad and unspecific by nature, the RocketLab has a given goal to accomplish with the solution delegated 100% to the team.
How do we organize a RocketLab?
It’s typically either Product or Technology bringing up a specific hypothesis (or a technical complex problem to solve). A short discussion determines the amount of time we’re willing to spend on finding a solution – usually 3 to 5 days. The organizers invite people to participate in the RocketLab and the Lab kicks off. It’s never the whole company, only few people but interdisciplinary.
The initial task after kick-off is an intense planning session. The organizers introduce the hypothesis to a greater detail and the team sets goals – together with metrics. Right afterwards with a clear goal in mind and a good understanding of the metrics solution ideation starts. Ideally, the team ends this activity with a solid set of tasks for each team-member.
The RocketLab needs 100% dedicated team members – no excuses – and sits co-located in a special meeting room.
What’s the typical outcome?
The expected outcome of the RocketLab is a solution for the hypothesis from the beginning. The solution is live, up and running. If the team was not able to solve it 100% they have a clear understanding of the remaining efforts and a thorough plan. The plan is then executed in regular agile development work. One hypothesis, one implementation, one proof. The RocketLab is an efficient and effective tool to work concentratedly on a hypothesis – goal-oriented but very intense.
What issues do we see?
We did around 10 RocketLabs for very different topics. Very concrete, technical topics up to very abstract conceptual work. The results were sometimes simply spot on, other times needed further perfection during daily work. In essence, the RocketLab is a tool which borrows aspects from the hackathon but is more effective and efficient. It simply works for us and produced some very surprising solutions.
We still see some issues with the spill-over effect of the RocketLab – but that’s a minor problem. Only 20% of the Labs experienced the spill over.
Design Thinking puts one person group in focus – the user. All activities in Design Thinking circulate around making the user’s life better. Design Thinking is an activity where all stakeholder take part. Not only designers, product owners or developer are part of the group, marketing, community management, finance and even legal should be involved in this very early stage of the product development cycle. Why should they? It’s a matter of understanding and communication. All participants of the Design Thinking process are part of goal setting, reasoning and the detailed planning. They need to share the vision behind the product. Time is wisely spend in the beginning to smooth the following implementation steps. Involvement of participants? 100%. No e-mails, no meetings, just the user and their pain points.
Other names for Design Thinking are “Design Sprint” (Google) or “Iteration 0” (infoq) or “Design DOING”. Design Thinking is a great planning tool to let all people understand what is build when and with what purpose. The process consists of a set of methods typically executed directly before the typical lean and agile development cycles start. It defines what to build and to communicate this purpose amongst the participants in an efficient, effective and fun way.
What’s the result of Design Thinking?
The result of the Design Thinking activity is a verified prototype, common understanding of what to achieve and a plan on how to proceed building the Minimal Viable Product and which features to add after going live. On top of that comes the certainty to build a set of functions that user’s really want and need – not a feature an executive has fallen in love with. Further definition of Design Thinking: IDEO, interaction design foundation or wikipedia.
What’s the purpose of Design Thinking?
Customer centricity uncover real user needs
Boost communication and understanding bring business and technology together – let the whole team understand what to build and for what
Create business value create and test solutions for these user needs
Be better and faster build better products and bring them to market sooner
Be focused and measurable extract clear goals from real customer needs
Get prioritization right feature prioritization gets a lot easier (no “pet” features)
Phases of Design Thinking
Design Thinking is a fun exercise, fast paced with the goal to create a concrete problem definition and an implementation plan for the most promising solution.
Starting with an idea of a user problem in mind the user observation phase results in a better understanding of the real user pain points. The team identifies the root pain points of the users and document them in an experience map. Creation of multiple solution ideas means to apply ideation techniques with the clear goal in mind of not implementing the first-thought-about solution. The paper prototype challenges the solution idea from ideation with the lowest possible effort and tries to resolve the user’s pain points. The implementation planning starts right after the finalization of the paper prototype. The result of the Design Thinking process is an implementation plan.
Implementation Guideline – Design Thinking in 10 Steps
1 – Identify user pain points
Go out of the building, watch, observe and interview real users experiencing the problems to solve by the solution you’re about to build. Watch at least 8 individuals! User’s verbal feedback usually contains their thinking of solutions, not their problems. The understanding of the team prior to the user observation is significantly different from it after watching and talking to users.
2 – Show the customer journey on an experience map
The experience map shows the current product experience (if any) and the user pain points. Organize notes from the team’s observation phase on sticky notes on a wall. All members of the Design Thinking team take part in the observation phase and make their own notes. The experience map shows the flow from the beginning (left) to the end (right) of the user journey.
From top to bottom the experience map shows the separate steps (above in blue) in the user journey. Below the user journey steps yellow notes represent the observations from the team members. Themes / cluster – represented in green – create an umbrella for the various observations. Additionally, emotional states – e.g. with smileys – enrich the overall journey and make it more visual.
3 – Extract and prioritize pain points
At this stage the experience map contains contains a lot of observations from various people. The next step in the process is to extract the pain points to focus on during the remainder of the process. “Dot-voting” is the tool of choice here. Each participants gets 4 to 5 dots. The biggest pain points receive the dots. Dot-voting happens all at once – at the same time. Observations may receive none, one or many dots from each participant. Counting the dots identifies the most important pain points.
4 – Goal and metrics setting
Having the ordered list of pain points extracted, the next step is to convert them into goals and make the progress measurable. Setting the goals can result in either quantitative or qualitative metrics. If possible assign real numbers to the goals. Orientation criteria for setting the goals can be:
efficiency (e.g. time spent on task)
effectiveness (e.g. reduction of # of errors)
satisfaction (customers’ happiness with the solution)
5 – Create example people – personas
Instead of developing a product for anybody – and hence most likely nobody – define real people, personas. Be specific and narrow user segments down. The persona should include the name, attributes, goals, concerns, quotes and other emotional elements. Ideally extract only 2 to 3 personas either individually or in small groups. Share the results afterwards with the rest of the group.
6 – Ideation – create ideas to solve pain points
Now it’s time to think about multiple solutions to address the user’s paint points. Don’t jump to solutions – don’t build the first “obvious” solution. Most likely there are more clever approaches available. All participants take part in the solution creation phase. Individuals create ideas time-boxed (e.g. 15 minutes). Afterwards they share them with the broader audience. There are no stupid ideas – it’s important to listen carefully to all ideas – some of them will stipulate further thinking. After the individual session a group session to detail the most promising ideas follows. Involving all team members creates shared ownership for the solution.
7 – Reality check
During ideation the goal is to create a lot of ideas. Are they all feasible? Can they really solve the issue? How does this look in reality? Since the ideas are in early stages we need to check them with reality. Place the ideas on storyboards. The storyboards show the interactions of the users with the idea. They are a sequence of scribbles which fit easily on a sheet of paper and focus on the positive user path. Let all edge cases, negative paths, recovery steps aside. Just focus on the positive user experience path associated with the idea. Reality check happens as an individual task or in smaller groups.
8 – UI Paper Prototype
The reality check sketches the user flow on a very high level – only focusing on the positive path and keeping intermediate screens aside. With the Paper Prototype the user flow should be as realistic as possible. Now the team members need to think through intermediate screens and pop-ups, edge interactions with the user and error cases as well. They transform the high-level storyboard an user interface. Sticky notes represent the UI elements – so the elements can change easily.
9 – Usability testing
Now it’s time to get real feedback from real users on the ideas developed. Any feedback is welcome and influences the prototype iterations. Feedback is usually more open since the prototype signals “Hey, we’re still in early phases. Don’t hesitate to give honest feedback!”. For the overall product changes at this stage are ways cheaper than rework done in later phases. Identify around 5 test participants – ideally based on the persona descriptions used so far in the design process. The storyboard serves as the base for tasks for the participants. The UI needs to change as an answer to the actions of the participants. Do so by exchanging notes or other elements of the UI manually. During the Paper Prototype testing give the participants no hints. The overarching goal is to identify gaps or errors in the user flow and collect feedback to improve the prototype.
10 – Implementation planning
The last step to finish the Design Thinking process is to come up with an agreed-upon implementation and launch plan. The plan is incremental in nature, delivers business value as early as possible and involves all disciplines. Implementation planning usually results in a story map which reflects the desired state of future interactions. Jeff Patton explains story maps in great detail in his book “User Story Mapping”. Here are only some important aspects.
What’s the story map?
The story map is a one page explanation of the big picture and shows details of the planned product or feature. It includes a release strategy, describes iterations around the minimum viable solution and identifies areas for additional research. First of all list all stages of the user journey as they use the product. This is usually similar or equal to the blue sticky notes produced in the experience map. Below, in green, put the concepts of the UI which help to fulfill the step in the user journey. Yellow notes hold capabilities of the product extracted from the paper prototype. Take a note on the sticky notes of the expected outcome the capability delivers to the user. The yellow notes usually translate into implementation epics in agile development. Put pink notes on capabilities marking additional research need.
With the finished story map, you have a quite comprehensive view of what needs to be build for a perfect solution. To further mitigate risk of failure you want to start with the least minimum effort in development – identify the minimum viable product feature set. Now, it’s time to sort the yellow sticky notes below the UI areas according to their criticality to the product success. Sort them into these categories having the question in mind: “How necessary for users is this capability to fulfill their task?”:
critical: absolutely necessary for user to get tasks completed
commercially acceptable: adding these features will allow commercial success
later: add capability later
nice-to-have: capability which can be added in later phases if time allows
When finished with the prioritization, you have your iteration planning finished. The top lane includes all capabilities needed for the minimal viable product. Next lane adds functions to enable commercial success. Later lanes store further functions to add to the product to become a full solution.
Find some videos on lynda.com by Chris Nodder on the specifics of each step and the applied technique (need a Lynda.com subscription).
Design Thinking – Timeline
Typically Design Thinking activities fit into a 5 days week. You might want to organize them like this:
Tracking of product metrics occur on various levels. PULSE and HEART stand for a logical structure of metrics to measure several aspects of your web product performance. PULSE reflects a more low-level and direct approach to performance figures. HEART on the other hand focuses on the customer experience.
The PULSE framework focuses heavily on direct impact KPI’s to measure the performance of large-scale web products. They typically reflect technical or business aspects of the performance. PULSE stands for Page views, Uptime, Latency, Seven-day active users and Earnings.
Page views reflects the amount of users visiting your site. Uptime gives the percentage of time the server infrastructure is up, running and serving content. Latency gives a proper indication of the performance of your site infrastructure and your overall software development efforts on execution speed. Seven-day active users says a lot about retention – the ability of your site or product to motivate people to come back multiple times within 7 days. Even if seven-day active users looks like an user centric KPI it doesn’t tell anything about the level of satisfaction of your users. Earnings, finally, gives a good indication if the product works – or not.
These dimensions of a product are definitely worthwhile watching and should be observed thoroughly. But – are they good candidates to focus on user centered product development? Are they any good when it comes to value generation?
HEART – higher level and user focused
The HEART framework is less generic, it costs more work to identify the right metrics – but it helps a lot to focus on users and makes value generation the most important goal. HEART is more adjusted to the individual product, it’s less direct and needs a good understanding of the product to measure. HEART stands for Happiness, Engagement, Adoption, Retention and Task Succes.
Happiness, Engagement, Adoption, Retention and Task Success in a nutshell
Happiness is a very fluffy description of a very important state of mind of product users. If the product touches people, if it really helps it makes user more happy. A variety of KPI’s express the happiness of your users. The KPI’s are very product specific – in our case we took “Net Promoter Score”, “User Survey”, “#Bugs on the board” and “Upvote index” to measure this very qualitative dimension.
Engagement measures the level of engagement of users with your site. It’s not overall engagement with a site in general, it’s engagement with the core aspects of the site. Focus is on specific pages and sections that are critical for the value perception of the user. We track “PI per Visit”, “Engagement on QDP” (our most important page type), “User activity”.
Adoption focuses on the amount of new users discovering the product and actually decide to become active. We decided to go with “#Registrations” and “Daily Activation”.
Retention measures the amount of users coming back to the product and use the product over a period of time. We measure retention with “Stickiness 30d” and “Churn Rate”.
Task Success measures the amount of tasks completed by the users. Not any tasks – those tasks providing most value to our users. It’s important to understand how many users are really engaged with the product and perceive value from the most valuable functions of the site. We are looking at “Q&A Index” (ratio of answers and questions per user), “Time 2 Answer” and “HA Ratio” (ratio of helpful answers to all answers).
Engagement, Adoption and Retention metrics are typically measured over specific periods of time. For some products it might be worthwhile to focus on a 7-day-period others might need a 30-day-period.
We’re still fresh on HEART but I strongly believe it will change the way we develop our product in future.
Anil Cheriyan is Director/Deputy Commissioner, Technology Transformation Services for the U.S. Federal Government. Previously, he was managing partner of Phase IV Ventures, a consulting and advisory firm.
Two important aspects associated with the terminology “Cloud”. First it’s important to understand the implications of the various cloud strategies (ranging from private cloud over hybrid constructs towards public clouds). Get your strategy clear on which areas to host where. Criteria to look at are: business value provided, business continuity, resilience, security. Second aspect is the organization. Get your people involved. They need to participate in the strategy definition. They will execute them actually. No time for information hiding and bimodal IT infrastructures.
Getting towards a modular architecture introduces flexibility in decisions, eliminates bottle necks and allows a decentralized governance. Today’s architectures are still monoliths or more advanced SOA stacks or somewhere in between. A more modular architecture exposes API’s via micro services. This architecture allows distributed ownership models. Complex is actually the implementation of these architecture rewrites. A lot of business related activities and the re-architecture work is a hard effort to combine.
DevOps is finally all about the mindset of people and the break-up of silo-ed organizations. People need to learn and understand the importance of collaboration and trust. This sounds simple, turns out to be a heavy change project. Anil started pilot projects and introduced the true DevOps mindset and collaboration through success cases. It’s not about adopting rules and processes from the DevOps movement “by the books” – it’s about training your talent to work closer together.
Agile development in software development is quite wide spread and commonly used. The acceptance over waterfall models is – where appropriate – high. Issues occur if the agile software development processes get surrounded by traditional waterfall-oriented functions – control functions. The most challenging part is to get agility into release management, deployment and integration testing.
Most important aspect of design thinking is the customer centricity. Understanding the real problems of the user to be solved is at the core of the approach. Not hunting the 100% perfect solution with all nice and “useful” features. Going for the most valuable solution, ship it fast. This requires heavy re-thinking within the organization. It’s more about talent and collaboration models. Important is to get people together with a thorough understanding of the industry and processes to help solving the customer’s pain points.