2018 year end update
Mindset Update - 2018 year end
Happy new year! This update will cover the last month’s progress, as well as a recap of what we accomplished in 2018.
Jacob Flood, December 31, 2018
As the year draws to a close, we wanted to post one last update, speaking to some of the progress we’ve made in 2018. It’s been an eventful, and at times arduous year – we’ve made incredible strides on the product despite many unfortunate delays. Before we move into 2019, we want to reflect on the good and bad of this year, to learn from our mistakes and hit the ground running in 2019.
Happy holidays to everyone, and enjoy our year-end summary!
We’ve finished up a lot of the major changes on our way to tooling, and kicked off the process. In parallel to the tooling being cut, we’ll be going over several smaller details, to prep for making our DV batch in march. We’ve finalized the shape of the headphones, which you can see below.
The app is coming along well – we’re all using it internally. We will be sharing public beta versions in the early new year to test without the headphones.
Below we’ve gone into a lot of detail with our year-end review. Enjoy!
Before diving into our year-end review, we’re going to give a brief update on our progress last month, focusing on the hardware and software.
Our goal with the product development since last update has been to go through a pre-tooling sanity check on each of the parts, to ensure that with the changes we’ve made recently we are still achieving our specifications. This verification has resulted in a few changes:
Changes to the overall shape of the band to improve ergonomics and sound quality
Finalized Bluetooth stack, and tested in the final form factor.
Implementation of the cable harness passing outside of the headphones
Implementation of a solution to compensate for the spring force of the cable
Mechanical changes to improve reliability of the top band
Internal modifications to shape, to account for minor changes in PCB shapes & mounting points
Improvement of the user interface (buttons)
Improvement of the assembly method (related to upper band and wiring)
Since our last update, each of those changes have been validated and implemented by our manufacturer. We are currently in the process of getting the tooling (read: the moulds) ready for a first small test batch production – the batch we’ve been calling DV in previous updates – which will serve as the final verification before starting to scale mass production.
Below we’ll highlight two particular points: Bluetooth testing, and ergonomics changes.
While the most important part of the product remains the ability to measure brainwaves, we must spend considerable amount of time testing and validating one of the lesser-discussed features: streaming audio and data through Bluetooth.
While the software side of Bluetooth streaming is well established – what protocols to use, and how to initiate the necessary handshakes – the electronics side is less concrete. Where you place your Bluetooth chip in the device, and how you design the antenna can fundamentally change the way the product interacts in real-world use.
The main challenge becomes ensuring that the product works in the majority of environment. In order to validate the performance, naturally the best test is to use the headphones in many different environments.
Bluetooth devices interact by sending data packets wirelessly across any of 40 different semi-overlapping frequency channels. There are two important implications in this. Firstly, the more Bluetooth devices in a confined area, the more likely that two devices will partially overlap frequency spectrums – this kind of interference can corrupt packets, increasing the likelihood that a connection is dropped. Secondly, the more closed the space, the better the connection, as the radio waves can bounce off of walls and furniture.
The result is that the easiest environment for a Bluetooth is a closed room, with few other electronic devices. To contrast, an open environment with many devices is more difficult.
For our test, we do a first validation within a normal setting, like an office. If this test passes, then we progressively move to more difficult environment. For our worst-case scenario, we test the headphones streaming capabilities in outdoors areas near Tsim Sha Tsuiin Hong Kong, where many people are using devices simultaneously, and there are very few walls for signal to bounce out of. If there’s no major issues there, then we are very confident that you will have very few Bluetooth connectivity issues in your regular office environment.
Our first tests many months ago yielded very limited results: 1m distance, operable only when perpendicular to a device. After several iterations – including firmware edits, and fixing a loose antenna that was causing this poor performance – we’ve been able to reliably achieve a standard of 10m distance.
With our latest batch, we didn’t run into major reliability issues, which gives us confidence to move forward with the design. The final tests will be conducted on DV units, from which we’ll derive our Bluetooth certification.
Before we started the tooling process, we tested the fit of the headphones using both head jigs and real users. This allowed us to discover a few minor fit issues and improve the ergonomics of the product before our design freeze.
The EV2 prototype completed at the end of the summer was able to just barely pass our requirements for head fit. You can see from the picture below that the headphones technically fit on our jigs, but that the ear-cup angles were often very close to their maximum size. This results in an unbalanced force on the head, where the cushion pushes harder on the top of the ear than the bottom. Beyond the comfort problem, this would impact the contact quality of the cushion electrode, as well as the sound quality and ANC performance, which are both directly related to the quality of the seal that the headphones create. For these reasons, we decided to undergo a few changes on the shape before cutting the tools.
All of the subtle, yet critical changes can be seen in the comparison below, where the left one is our old prototype, and the right one the final one (cable is not shown):
Summary of main industrial design changes:
Flat cable ribbon is passing on the outside (not shown above)
The ear-cups are more angled
The electrodes are slightly closer together for better contact
The user interface button design was changed to better locate each button
Cushion electrodes are now button shaped, instead of a foam pad with conductive fabric
The clip where the arm enters the upper band is slightly larger to accommodate cabling
In addition to these changes, the underlying mechanical design had to evolve quite a lot as well. As new requirements such as the cabling were added, the supporting structure had to be adapted. In particular, the upper band was changed from silicon to plastic for increase quality control and robustness. This includes many changes were also performed to reduce the weight as much as possible.
However, we listened to your feedback in the previous update and decided to keep the aluminum arm bands. Although they are heavier, they will add a more premium allure to the product, and likely last longer.
The images below show mock-ups with the corrected shape, tested on the large and small head sizes. You can see that the ear-cups fit much better in both small head and large head shapes, as compared to the previous tests above. These changes have been validated on the test jigs (below) as well as with people in our Shenzhen office.
In addition to the ergonomics, we have tested the location of the button, the feel of the fabric, and the sound quality with real people. Each of these tests led to small, subtle changes that we’ve since validated with our factory’s engineers and incorporated into the design.
While the preliminary tooling is ongoing, we’ll be spending our time in parallel going over some last steps before the mass production scale-up:
Finalizing CMF (Color, Materials and Finish)
Resolving issues coming up with the molding process
User Experience Testing
Finalizing electronics, and getting international certifications
Final ANC testing and tuning with factory & Onkyo
Final audio testing and tuning with factory & Onkyo
Electrodes environmental testing and lifecycle analysis
Having achieved our goal of starting the tooling by end of year, we’re eager to enter 2019 with the same timeline as before: DV batch by march, and production after!
Following last month’s unveiling of our neuro-adaptive music platform, we’ve spent the last month working on UX testing this new feature. In practice, that’s meant that the team is now using the Mindset app while we work, in order to give our feedback, test our new features, and find any bugs that exist.
This month we focused on the gamification, and packaging of the app. This includes front-end features centered around the app flow (starting a deep work session, visualizing your data, and features like to-do lists) and back-end functionality (cross-platform compatibility, app install procedure, and server operability).
In order to test each feature quickly, we’ve created a stripped-down version of the app. In this version, the app UI (the visual feel that makes the app smooth) was removed, to allow us to quickly swap in and out different features to test with. The pictures below contrast the Kickstarter UI, and the current test-environment.
We’re experimenting with new ways to visualize the data over time – in the above image, we were experimenting with a bar graph instead of the traditional line graph.
In January, we’ll be finalizing the dashboard features, and will then reapply a UI skin to the app. At this point, we’ll reach out to a few beta-test backers to experiment with the app for feedback.
We have two requests we’d like feedback on this month:
Do you have any suggestions for the best way to visualize data on our dashboard? Line graph interfaces, bar graphs, calendars – let us know what you prefer!
We’re looking for some help with the authentication protocols for app (using OAuth to allow facebook/google login, while also allowing traditional email+pass accounts). If any of you have experience in this, please send an email to email@example.com!
2018 has been a very eventful, and often stressful year for all of us. As the product’s advanced, we’ve become more and more conscious of the trade-off between quality and time, as we push to get the product finished without compromising quality.
As a result, there have been recurring themes throughout our updates that have been the main drivers of our attention. Electrode builds, ergonomics, and ear cup fabrics, for example, have taken a lot of iterations to get right.
In parallel, there’s also another side to the company that we don’t often share: the marketing, fundraising, and operations necessary to keep us moving. These are significant parts of our experience, that don’t make it into the limelight.
So as 2018 draws to a close, we wanted to reflect a bit on the last year’s progress in each of our 3 domains: hardware, software, and marketing. For each category, we’ll talk about the progress we made, and reflect on any learnings we accumulated in the process.
The product, naturally, is our #1 priority. At the end of 2017, we had built our first engineering prototype, a test-unit that we brought with us to CES. The goal since then has been to refine this design sufficiently to be able to cut our tooling, and move into mass production.
Upon analyzing the EV1 unit, we found that we were still far from our goal. EV2 and EV3 since then have focused on solving the main issues we encountered, in order to get our product production-ready.
Barring the small changes, these improvements have fallen into 4 categories: ergonomics, earcup fabric, cushion electrodes, and top electrodes.
ERGONOMICS & SHAPE
The ergonomics and shape of the headphone have undergone several iterations as we tweak and improve different parts of the assembly. The goal is to ensure that the headphones fit the majority of people comfortably, while still staying true to the industrial design.
The original industrial design (seen above) was optimized for aesthetics more than for function. We used this design as a baseline, as we worked with our factory to improve the fit of the headphones.
For EV1, we used a mathematical model of the headphones to measure their fit on MRI data for people’s heads. Alongside our factory’s suggestions for mechanical requirements, this resulted in thicker earcups, pulling the electrodes further down, and shortening the arm bands.
Once EV1 was built, we went back to the drawing board to improve the design from our learnings. We introduced the new electrodes, changed the arm band design, and tweaked the curvature of the upper band to better match the lower electrode band.
Despite being functional, the post-EV1 design had several flaws. The largest issue was that because of the wide upper band, the headphones looked very bulky when worn. This widening also caused issues with the ear cup pivot, which couldn’t flex enough to reach all head shapes. For EV2, we used several reference headphones to redesign the shape, making it rounder and more shape-fitting. This also mediated the issues with the electrode band, which could be made concentric to the upper band.
Finally, as described above, we’ve made the last tweaks to ensure that the shape truly fit all the constraints. This final shape is dramatically more comfortable and better-fitting than the original, and ensures proper contact of the electrodes. We hope that these changes will ensure that everyone’s experience of wearing Mindset is smooth and comfortable, and matches your expectations of the product.
Before starting our design process, we had a team of industrial designers pick out the fabrics to be used on the headphones – this design was the one seen on the Kickstarter campaign. However, as you can see below, despite this fabric looking good in the design, it definitely did not age well
We then went to a small fabric market in Shenzhen to choose an alternative fabric that would be more comfortable, yet still looked like the original design. However, this turned out to be more difficult than expected, as we discovered that the properties of the fabric have a big impact on the sound quality. The fabric shown in the image below was comfortable, but resulted in poor sound quality.
In order to tackle the sound quality issues, we consulted with our manufacturer to directly source acoustic fabrics. To contrast, these acoustic fabrics matched the original design and had good sound properties, however they were quite uncomfortable and hot.
After some minor design changes to the way the fabric attaches to the ear-cups, the manufacturer suggested a new fabric that was comfortable, and met the sound requirements. Talking with our industrial designers, we weren’t happy with how the pattern of the fabric clashed with the rest of the industrial design.
The design change mentioned above, however, had solved the sound-quality problem, allowing us to use non-acoustic fabrics while maintaining the audio quality. As a result, we took one last trip to the largest fabric supplier region in Shenzhen, and found the exact fabric we wanted.
In a headphone, the ear cup fit is foundational to comfort: fabric that is too rough, too hot, or too tight can singlehandedly ruin the headphone. Because of the tight requirements we set on the ear cups, this iteration process took many months.
While there may be some slight alteration to the color and finish of the fabric as we move to mass production, we are confident this is the final fabric we’ll use. Switching colors and finish of fabric has no impact on the rest of the design or it’s functionality, which means we’re done with the long iteration cycles. After all that work, we’re really happy with how it turned out, and we think you will be too.
Through the evolution of the product, we’ve gone through several fundamental changes to the cushion electrodes. Our main goal from the start was to make the electrodes as comfortable as possible, while keeping a high signal quality. The challenge was to take what we know works, and adapting it for consumers: in other words, it must be a cheaper, more comfortable version of a typical medical grade EEG electrode.
Here’s a quick recap on the ear cushion electrode development so far:
We started we the intention of using conductive fabrics embedded in the ear-cups to measure the electrical activity near both ears. You can see above the different samples of conductive fabrics that were tested.
We first tested the conductive fabrics recording signals from the back rim of the ear. This turned out to work well, but only for some users. In particular, wearing glasses was a big problem in this design. It was also sometimes uncomfortable, as it necessitated pushing the ear outwards to get a snug fit.
Trying to resolve the previous issues, we moved the fabric electrode against the outer part of the ear. This extra fabric ended up being a bit too hot, and because of the large surface area it resulted in more significant antenna effect (picking up ambient electrical noise). We found that overall, touching any part of the ear rim became uncomfortable over time, and should be avoided.
Then, we designed a novel assembly method to create ear-pads: extremely low impedance foam pads that could be placed in different locations within the ear-cup to contact the ear lobes. This was quite comfortable, but made it difficult to control the contact quality for all users.
In addition to the design changes, we had yet to test the long-term reliability of conductive fabrics: corrosion and decay were known to be problems with this material, and hadn’t been solved yet. At the end of the day, we decided that it would be quite difficult to rely on conductive fabric in this new product.
Finally, because of the uncertainty surrounding the conductive fabrics, and the difficulty of running rigorous studies on the materials, we decided to move to a button electrode, located on the outer rim.
Taking what we’d learned in our ear cup tests, we found that the bone on the edge of the jaw is the best location for the ear cup electrodes. The images above show this design. Note that the exact position will likely change.
Having this type of electrode allows us to properly match the materials between all electrodes in the headphones and for various other reasons, improve the signal quality, and ultimately, the performance of the product.
The lesson learned throughout this process is to test through building. We spent several days early in 2018 debating the best design for the ear cup electrodes theoretically, rather than simply building several versions. Once we started putting together samples of each, progress was significantly quicker – we’ve since carried this mentality forward with other parts of the product.
To date, the top electrodes have been the most difficult part of the design by far. This can be seen in the sheer number of iterations we’ve gone through.
Way back, we ran our initial test to validate the feasibility of the product using clinical grade gold-cup electrodes, which were mounted on an old gaming headset.
We knew however that for a consumer device, it was very impractical to use electrodes that required conductive paste to operate. After testing several pin, felt, and silicone electrodes on the market, we decided it was best to build our own electrodes.
Our first design utilized a dry electrode using pogo-pins. We quickly ran into all sorts of electrical, mechanical and ergonomics issues with this electrode, and ruled out this design very quickly.
After being inspired by conductive silicone electrodes, we came up with a concept of electrode that could bend like silicone, while keeping the material properties of high-quality medical grade electrodes. You can see the concept electrode blow. This turned out to be only good on paper - it was very difficult to make and could not be scaled to a large production.
By the time we started the Kickstarter, we had already designed an electrode that could be used in production. We still had some issues on comfort, ease of assembly, and reliability, but these were to be improved greatly during the design for manufacturing process. This was our goal for 2018.
After Kickstarter, we quickly iterated over the form factor to nail down the final design. This included robustness improvements, and quality control, so that each electrode manufactured would yield an identical signal quality. Our biggest issue, however, still remained the problem of mechanically getting a good contact with the head, while staying comfortable. The electrode below was our first iteration that successfully achieved all of these criteria.
Another issue became clear earlier this year: we needed to find a way to get the electrodes to maintain an even pressure across the 3 sensors. We went through several design iterations of a flexible mechanism that would even out the pressure across the sensors.
This system was eventually scrapped as our software team ran an experiment that demonstrated that the middle electrode was redundant for our purposes. Moving forward with a 2-electrode design greatly simplified the process of maintaining a reliable contact, as fitting 2 electrodes on a head is dramatically easier than 3. The lesson learned is to test our assumptions early, before moving forward with complex design work.
The next step was to make sure that the electrodes could be made removable. The design blow involved a small piece that could be screwed in and out of the headphones for easy replacement. These sensors checked all of the boxes, but were simply too expensive to manufacture at scale.
However, in parallel to the modularity, we were able to majorly improve the comfort and signal quality of the overall design through better adjustments, and more careful quality control on the materials used. The image below shows the flexible mechanism that allows the electrodes to bend and shape to the users head.
Finally, the last issue that was tackled was to reduce the cost. The final electrode design below displays major cost improvements, which is one of the main reasons why we can provide an EEG device at costs like other headphones on the market. The remaining changes between now and shipping time will be purely cosmetic, and can be handled in parallel to the normal manufacturing process.
As the signal quality is of #1 concern, we’ve spent a considerable amount of time iterating through different electrode designs to get it just right. This final design, after many months of work, achieves all of our requirements for quality, cost, and comfort – we’re really happy with how it turned out.
In parallel to the hardware development, the software team has been working hard to build the app experience.
Our priorities for the software in 2018 were twofold: to improve our ability to measure concentration, and build an enticing app experience that improves productivity.
By the end of 2017, our algorithms were able to accurately measure cognitive engagement and mind wandering, two great indicators of concentration.
Moving forward, our goal in 2018 was to validate these same algorithms in less controlled settings: users wearing dry electrodes doing their work, instead of users wearing gold cups doing controlled tasks. The first few months of 2018 were spend designing and running experiments to collect data during which users wear our prototype headsets and perform their work at their computer. We would then analyze this data after-the-fact to see whether our concentration predictions matched the user’s self-reports.
This type of experimentation continued iteratively through most of 2018. We gradually introduced before/after surveys into the experiments, ran some meditation experiments to measure mind wandering, and used several different versions of the electrodes and electronics. Each time, we were able to improve our process for gathering data, and learn more about how people perceive concentration, and how we can measure it.
In parallel, it became increasingly important as the product team tested new electrodes for us to establish quality baselines for these tests. The software team spent several weeks putting together electrode evaluation tests: protocols that a user can undergo to evaluate the quality of the signal gathered from an electrode. These were used by the product team as success criteria for future developments.
Finally, later in the year, we addressed the training problem: since our algorithms need data to train on, it’s hard to get them to work well out-of-the-box. To overcome this, we worked with several neuroscience researchers (now our advisors) to create neuro-quantifiers: pre-trained models that recognize distinct neural features that correlate with certain brain states. We’re now using these quantifiers in parallel to our evolving machine learning models to give a more accurate image of a user’s concentration level, in particular during the first weeks of use.
Our priority from now until shipping is to improve these quantifiers, so the out-of-box experience is as smooth as possible. We’re excited to get the DV batch, so we can all start using the headphones alongside the app to create a great user experience.
At the beginning of 2018, we were still using the same app design from the kickstarter. Our goal for 2018 was to achieve 3 goals:
Create the infrastructure to support the app launch
Create an environment to iterate on user experience tests
Finalize how the audio feedback will be sent
We started the year by focusing on the infrastructure: the app framework, the cloud server, and the data streaming. With these 3 pieces in place (and constantly evolving) we were better able to run the data science experiments we needed.
During early 2018, as a result, the app team spent most of their time supporting the experiments run by the data team. This involved designing new experiments, ensuring that the data streams correctly, and setting up the data acquisition interfaces prior to each run. A lot of the app design got put on hold as data science priorities took the foreground earlier in the year.
As the product got delayed, the software team got additional opportunities to delay the final-touch app design (UI, packaging the app, operating system compatibility, user authentication) and focus instead on the bigger unknowns (data science, user experience, and music feedback).
By the summer, the infrastructure to run the data experiments had evolved to the point that we could focus on app development in parallel. This motivated a shift towards thinking about user experience testing, user interface design, and gamification. The rest of the year focused on these elements: building prototype versions of the app that we can use internally while we work. As we iterated, our infrastructure solidified, and the time each iteration took became shorter.
Finally, in parallel, we began working on the audio feedback. As we discussed last update, we quickly realized that simple audio cues are not ideal. Instead, we’ve been working with professional adaptive music designers to create neuro-adaptive tracks for focus. Last month we got our first demo track, set it up with our headphones, and were really impressed by the result.
Moving into 2019, all of the foundational pieces are in place. Our goal now is to continue our UX testing, both internally and externally, until the product is ready to ship. In parallel, we’ll be creating more neuro-adaptive music, to add variety to the playlists. We’re excited to see how much better the app gets as we continue iterating on it!
While the hardware and software teams have been busy building the product, the marketing team has also been hard at work. With the kickstarter and indiegogo success behind us, our goal for 2018 was threefold:
1. Solidify our value proposition
Moving beyond the kickstarter, it was important that we solidify the value proposition of our product. Is it more about concentration, or more about productivity? Should we talk about the tracking feature, or the audio feedback?
Through our facebook/google ads and our website, we’ve been testing out several different phrasings of the product to determine what connects best with our users. We also spoke to several of you 1 on 1, to get insight into why you bought the product, and what you were most eager about. We’ve been consolidating this information, and will be unveiling a new website and branding in early 2019 to coincide with the product launch.
This process has also taught us a lot about what our customers (you!) value in the product. This information has been passed over to the software team, who are incorporating it into our latest releases!
2. Isolate our beachhead market and target customer
In parallel to value prop testing, we’ve been testing sending our ads to different audiences, to see what customers are most attracted to our product.
Put simply, this involves setting up different “lookalike” audiences in facebook and google, monitoring the conversion rate of those ads, and using that information to motivate conscious choices about who your target customers are. Iteratively, this lets us put together a more thorough understanding of who is interested in buying our product, so that we can more effectively reach them in the future.
This information will also be motivating our rebranding in early 2019. We can’t share much yet, but we will forward over insights and followup questions early next year!
3. Continue producing sales to extend our runway
We mentioned in a previous update that our finances were very tight. As the product development continues, it’s increasingly important that we focus on sales, to ensure that our company stays afloat well passed shipping out our pre-ordered units.
We get asked a lot “why are you still sending ads if you haven’t shipped the product yet?” In addition to the insights above, the simple reason is that we need to make money in order to survive, and sales are the only way to do so. It’s crucial that we continue pre-selling the product, so that we can continue improving the product and iterating on new features well past shipping our first pre-orders. Every dollar of profit gets recycled back into improving the product for everyone.
If you want to help, please like our posts and ads if you see them on Facebook. This kind of positive encouragement helps tell Facebook to spread the word, and build up a wider audience for our product. The more people know about us, the better we can continue making the product for you and everyone else!
Our #1 priority for the business side in 2018 was to close our pre-seed fundraise. This financial support was crucial to ensuring that the company can survive beyond our first product, to continue improving the experience post-launch.
We’re really excited that in August, we closed our pre-seed round with SOSV and Real Ventures, two of the world’s most prominent seed-stage VCs. Since then we’ve been working with them to plan out for our 2019 launch, and future products. Their support has been incredibly valuable in helping us prepare for a big year ahead.
Our second priority, in parallel, was building the Mindset platform. The business team spent a lot of time in 2018 talking to developers, researchers, and companies about potential partnerships with Mindset. From apps developed using out SDK, to research projects in academia leveraging our hardware, to corporate incentive programs to increase focus in the workplace, people from all walks of life have been committing to building out the Mindset platform, to increase the scope of what’s possible with neurotech.
If you’re a developer with an idea that can be built using Mindset, we’d love to hear from you! Send an email to firstname.lastname@example.org a little blurb about what you want to build, and we’ll discuss how to make it possible.
Looking back, 2018 has been an incredible year – we’re really excited to see what 2019 will bring, with our product launch and beyond.
We want to thank you all for your continued support on this journey. We’ve had our share of ups and downs, and as we near the final stretch we’ve more and more eager to share Mindset with you all. We’re incredibly grateful for your feedback, and your kind words throughout.
We wish everyone a happy holiday season, and an incredible new year!
Lots of love,
- The Mindset Team