Contextual and Granular Notification Controls

I love getting notifications on my phone. I’ve written a lot about notifications, at one point calling the notifications screen “my home screen.”

But I don’t like how we are forced to control notifications, which ones we want, which ones we don’t want, how we want them, etc through the settings in the mobile OS, and/or through the settings in the mobile app. And the controls we are offered are not granular enough for my needs.

What I want is contextual and highly granular notifications controls. What I mean by that is I want to be able to tap on the notification itself (or swipe it, or use some other gesture on it) and get the ability to control it with a lot of granularity.

I have Twitter set up to notify me when certain people I follow tweet. Yesterday I was getting tweet notifications from a Patriots fan I follow. I wanted to mute my friend temporarily from my notification channel and I wanted to do it right from the offending notification.

I get notifications from the NBA App when games are close in the final minutes. This is an awesome feature. But there are only certain games and certain teams where I want that notification. I’d like to be able to tap on a notification telling me that the Lakers Phoenix game is close and tell the NBA app that I don’t care about the Lakers or the Suns and don’t need that notification.

I get notifications from Dark Sky when there is some weather event coming. I’d love to be able to click on that notification and tell Dark Sky to notify me when a rainstorm is coming but not a snowstorm.

Part of what I want is the ability to change the notifications settings in context, right from the notification that generates the desire to change the settings. And part of what I want is way more granularity in the notifications I get.

So for this user experience I want to show up, we need the Android and iOS folks to build more functionality into the way their notifications services work, specifically contextual notification control. But we also need app developers to make their notifications smarter and to give users the ability to control them with a lot more granularity.


Comments (Archived):

  1. jason wright

    would you be willing to pay for such enhancements?

    1. fredwilson

      to the operating system vendors or the apps themselves?

      1. Pranay Srinivasan

        Notifs like email need to be free. Notifications are the new Email + CTAs for Life.

      2. jason wright

        have you tried cyanogen?

    2. JimHirshfield

      No. It’s a feature that makes the app more appealing to the userbase. Makes users more dependent upon the app.

  2. William Mougayar

    I agree the granularity in personalization is not totally there, because the Apps make assumptions based on what they “read” and know about us, resulting in hit and miss quality.But would you be willing to spend the time configuring and re-configuring these notifications, based on your needs that are likely to be changing? This reminds me of content filtering which I spent 4 years at Eqentia trying to convince people to pay for highly advanced filtering of content, but people preferred the serendipitous and signal-biased nature of content rivers and social streams.What we need is a smartphone that can read our minds, because we change our minds a lot based on context and other variables.

    1. fredwilson

      i don’t want to spend time configuring and re-configuring the notification settings. i want to be able to do it in real time, in context, one notification at at time, right from the offending notification. i think that’s a huge difference in terms of making this stuff easy for users

      1. William Mougayar

        Do you think this is an “advanced user/narrow segment” type of need or does it appeal to a large mass audience?

        1. awaldstein

          Don’t think this power user stuff at all. I think this is a common behavior and most users would be happy to simply do this on the fly.What the mass market user does is just to do stuff naturally. Understanding little about the system and the past is their gift.And the target I think of mass market oriented ux from an assumption perspective.

          1. William Mougayar

            “Mass market oriented UX” – that’s the key to adoption. agreed.

          2. awaldstein

            Doing a project that touches on the snapchat user.You learn to move to the behavioral response understanding quickly.For mobile generally that is the pov that i’m building on and its so freakin powerful.

          3. fredwilson


        2. K_Berger

          This feels like power user stuff to me. And even for power users, it can get very confusing very quickly. Like when you mute something, you probably want an option to set how long to mute for (permanently, 1 hour, 1 day, etc.). And then you probably need an indicator somewhere to remind you it is muted so you remember to turn it off in the future. And so on and so on.

          1. William Mougayar

            I think so too. Unless the UX improves a lot to dumb the interactions down.

          2. K_Berger

            If that’s even possible. Some of this is truly complicated unless you spend more time than it probably deserves.

          3. William Mougayar

            How is DLD πŸ˜‰

          4. William Mougayar

            yup. wake me up in 1 year on that one.

      2. LE

        I was thinking about what you are saying and think that many software features (implemented or not) relate to whether something is a free product or a paid product. Or a saas for that matter.In the case of a paid hardware product (like iphone) it makes sense to hold back features (software or hardware) so that you have an improved product later on that you can charge for again. In other words planned obsolescence. If IOS 7 doesn’t include granularity on messaging (that I desire) it might mean I then have a reason to upgrade to iosN which will mean I need a new piece of hardware possibly and have a solid reason to upgrade. Roll it out before it’s needed and Apple loses that advantage.In the case of a free product there is less of a reason to stuff as many features into the product (assumes among other things that adding features costs money to do in dev time). So those counterbalance each other.Many things come into play for example if your product locks people in then there is less of a reason to add features before they are necessary. Otoh if your product doesn’t lock people in then maybe you want to stuff it with as much as you can as often as you can to keep customer not switching to another product.

      3. Juliet Peters

        Will you pay me for doing this your way. I can tweak them for ya’

      4. Mordy Kaplinsky

        In theory this is what Google Now is trying to do, but the user experience is a difficult nut to crack.That said I worked on a related project in a different field, and the research we did showed that people are creatures of habit and pretty much did the same things. So if you can get people to train it for a few months, the first month will be a lot of work – and potentially drive the customer away – but the effort will decrease drastically every month and users will get very good targeting accuracy.this research is corroborated by additional research performed by other parties working to use metadata to predict where individuals will be at a given time in the future.In theory an OS owner can build – or acquire – something like Foursquare’s pilgrim and expand it to include other sensors and more targeted scenarios, to improve the AI element and reduce the amount of input in that first crucial month.

    2. Mario Cantin

      You’re talking about HAL 9000 capabilities here — a convincing simulation of it, I mean.

  3. kirklove

    I want a way to filter obnoxious Patriots fans too!

    1. fredwilson

      And Giants fans too???

      1. kirklove

        Nah. Giant fans are fine. They are a lot like Eagles fans. So I get it.Just Pats fans are the worst. They have no idea how much of an anomaly it is to have someone like Brady. He may go down as the greatest QB of all time. Even Belichick (the genius) is propped up by him (he’s 51-65 without him! Very, very average).In about 2 more years when Brady is done, you won’t need the filter. Pats fans will crawl back down to their basements and be quiet all on their own. Kind of like the Cowgirls fans. πŸ˜‰

        1. fredwilson

          Keep it coming Kirk. I love it

        2. Elia Freedman

          That record isn’t fair to Belichick though. He did coach the Browns.

          1. kirklove

            Kind of adds to my argument though, no? πŸ˜‰

        3. LE

          Even Belichick (the genius) is propped up by him (he’s 51-65 without him! Very, very average).Non sports guy question for you. Can’t some of the QB’s success be related to coaching? Doesn’t the coaching relate to how others play on the team which is significant to what degree Brady needs to inflate balls (see I do know something I guess). Anyway my question is serious.

          1. kirklove

            It’s a great and valid point. Never said Belichick is a fraud. He’s a great coach. He’ll go in the Hall of Fame. And is a good fit for Brady.Same thing could be said of Phil Jackson and his Bulls and Lakers teams. It’s never “easy” but it’s certainly a lot “easier” to win when you have arguably the greatest players of all time playing for you.

          2. LE

            Wow Belichick went to Philips Academy. Fancy Schmancy. An endowment larger than most colleges.

          3. JamesHRH

            I think Belichick is underrated.He is the only coach I know of who is famous for building a situationally driven roster and then coaching to those situations. No one in the league runs 48 times against Indy one week and then throws 52 times against the Jets the next.Its a true luxury to only need Brady and one serious threat on O, but Gronk bot picked in the 20’s in his draft year I think.And it is amazing what they do with the O line every year. And gutsy to cut key players at their career twilight.They do a ton of things right.But the blind hatred makes total sense – 10 Conference championships games doesn’t leave a lot of air for other AFC teams in their division.

        4. Salt Shaker

          I hate the Pats, I really, really do. But you’ve got to give Belichick and the org props. Yes, there’s only one Tom Brady, but they reload every year at seemingly every position, and the team gets better and better as the season progresses, last night a case in point.

          1. kirklove

            But do they really reload? I’d argue they don’t. Brady just makes it look that way. Gronk is legit. The rest of the team are any other 8-8 crew even with Belichick. And don’t get me started on McDaniels. Remember his tenure at Denver?I’m telling you, once Brady is gone the Pats will struggle to be a .500 team.

          2. Salt Shaker

            Well, I guess then you don’t think Jimmy Garoppolo is the 2nd coming? Amendola and Edelman are by no means slouches. Check out their YAC stats. Belichick should actually thank Mo Lewis for discovering Brady.

          3. kirklove

            Ha, that’s cute about Garoppolo. He’s straight out of Central casting that’s for sure. And that’s about it.And Amendola well let me just place this here “Amendola is the 50th receiver all time to catch a touchdown pass from Tom Brady”And Edelman is a solid slot receiver. Not a primary threat. You saw what happened without Gronk. Need I say more?And the year the Pats didn’t even make the playoffs? 2008. The only season Brady didn’t play (due to a season ending knee injury)I know you hate the Pats, too (Proud of you for that!). I’m just tired of everyone saying how great the Pats “org” is when it’s ONE guy who’s the key to it all.Ok, I’m out. Mic drop. /Fin

          4. dgay07

            While they didn’t make the playoffs in 2008 they were 11-5 with Matt Cassell. How much money did He make off that one season in New England’s system? The argument can be made for every hall of fame coach he has and needs hall of fame players. My prediction for the Pats is that when Brady goes, Bill goes and that will be a huge challenge.

          5. creative group

            Can’t wait!

          6. kidmercury

            what do you think of the doug pederson hiring?

          7. kirklove

            Not sure, Kid. It’s not particularly inspiring, though I’m def open to giving it a chance. ESPN and the like want to shit all over it. I don’t think it’s that horrible of a choice. It does feels “safe” and that bums me out. You know? Like a 10-6, win a playoff game or two choice to me and never more than that.I really wish they gave Chip another year or two. I respect he was trying to do something different and bold. That’s how you win IMO.

          8. kidmercury

            i’m glad they cut chip. i was a believer at first, but too many reports of the players losing faith as well as how simplified the offense had to become to run no-huddle was too costly. also, the GM moves were disastrous.some people are saying that pederson deserves a lot of credit for the way KC ran its offense this year. if that is true i’m happy with the choice because at times this year KC looked like andy reid playcalling coupled with disciplined game management (except the disaster that occurred yesterday which brought back many terrible eagles memories).

          9. creative group

            Walking the grey line you mean.Cheaters, cough, cough!

      2. pointsnfigures

        Notre Dame and Michigan fans though.

      3. creative group

        FRED:that comment appeared closer to Ted Cruz’s statement regarding his opinion of New Yorkers pilot I verses a Warren Buffet nugget.GO BIG BLUE!That is as large and in charge as our chant goes.

    2. William Mougayar

      Funny how a sports season can be the trigger point for these sorts of needs.

  4. Pranay Srinivasan

    I also extend this notifications idea to real world businesses where we need followup but HATE email. We need contextual reminders on our phone when we are behind on certain work tasks and need a handy way to remind large executive work forces of tasks.”I need you to complete these deliveries by this time in your timeZone because it is something I have to present to this meeting at this time. And I want a notification to pop up that reminds me if you have not completed that delivery atleast 2 hours before so I can follow up and atleast make sure you are doing it.”Or “If you are stuck with something and cannot make it to office, ping me via a notification I will see”Slack is a great reactive example but if the phone auto detects your location with your calendar and notified your meeting is late…”

  5. JimHirshfield

    You want AI in your notifications, me thinks.

    1. fredwilson

      Eventually. But I am happy to train the algos

      1. JimHirshfield

        “Train the Algos”, a great blog headline, movie title, or rock band.

        1. pointsnfigures

          Would be good if all the music was made by robots, algos and oversampling.

        2. Mario Cantin

          Movie: “A Train Named Algos”Blog: “Why Training Algos Will Become Obsolete”Rock Band: “The Strained Algos”

      2. Augustin Ambrose

        Looking for an app that tells me what I need to work on, everyday

        1. Sam

          You must not be married.

          1. Augustin Ambrose

            Yes, you are right.But why would Fred need one for notifications when he is married for 20 years now

      3. Vasudev Ram

        Classic example showing some issues with AI:Time flies like an arrow.Fruit flies like a banana.

        1. Kirsten Lambertsen

          Extra credit for Groucho quote.

          1. Vasudev Ram

            Ha ha, nice. I also googled and found this just now:…The number of interpretations” makes my head spin”, as PG (HN/YC) said once (in some other connection).

          2. Kirsten Lambertsen

            Ha! My favorite Grouchoism is “Outside of a dog, a book is a man’s best friend. Inside of a dog it’s too dark to read.”

          3. Vasudev Ram

            Superb one. I’m saving that.

  6. Guy Lepage

    I feel that the granularity of most notifications can come from the apps themselves. Most sport apps do this fairly well.

  7. Craig Cramer

    Smart phones aren’t smart. They’re informed.

    1. William Mougayar

      Good point. Informed by their owners. A smart user will lead to a smarter phone :)Your smartphone can be as smart as you are.

  8. brian piercy

    You’ve probably heard of people having a daily mental “decisions budget”, and per-channel notification tuning definitely falls into this category. If I don’t want clutter from a particular app, it gets the mute. For everyone & everything.Constantly fiddling with a phone is the surest way to tick off my wife and anybody else around me.

  9. Pete Griffiths

    I wonder how many ‘regular’ users need that degree of control.

    1. fredwilson

      Every single one of them

      1. Mario Cantin

        But how many would want it though? Mind you, after seeing how you were spot on about the watch, I’d probably place my bet on you. In fact, I too would welcome more granularity as long it it would be super intuitive to use.

      2. Pete Griffiths

        Let me rephrase that. I wonder how many will be able to use it even if you give it to them. (though I confess to being intrigued by the idea of ‘real time’ tweaking on the content.

  10. John Frankel

    I suspect you need an AI layer between you and the app that can predict what you want based on past actions. It would not be perfect, but it would increase the signal to noise ratio. The question is how many missed important notifications would the average person be willing to live with?

    1. sigmaalgebra

      I suspect you need an AI layer between you and the app that can predict what you want based on past actions AI is 99 44/100% useless, worthless, nonsense, hype BS. E.g., an AI program that plays a great game of chess can’t make a PB&J sandwich.I’ve worked in AI with IBM and GM, programmed AI, published successful peer-reviewed original research in AI, and concluded that AI is, in the immortal words in All the President’s Men, just “total BS”.For more, see essentially a review of AI and a review of a book on machine learning at…That book has much that is good and new, the good, however, is not new and the new, not good. The book is some statistics with the most recent work from good mathematician and applied statistician Leo Breiman. Computer science and learning have essentially nothing at all to do with the contents. The word learning was stolen from psychology; essentially all the rest was stolen from statistics, some going back 100 years. It’s stolen, old wine in new bottles. It’s STATISTICS and should be called just that, statistics.My patience with computer science nonsense and academic theft has long since worn way past thin. That garbage should stop.Computer science has done some good work, e.g., heap sort and the Gleason bound. Uh, correction, the Gleason bound was from Harvard mathematician A. Gleason.Instead of learning and intelligence, of which computing has zip, zilch, and zero of either, computing is still based on programming where we have a problem, seek a solution, get some data, and manipulate that data for a solution to that problem, and that solution is not at all intelligent. Learning is something humans do, and commonly kitty cats, but not computers.This stuff that computers are “giant electronic human brains” went way back, was total hype then, and is total hype now.Sure, in some cases the solution can appear fantastically smart, and my Ph.D. dissertation had an example, but that appearance is just deceptive. Instead, my work was just some good applied math.See also post…in this thread.Yes, you very much do want a “layer” — right, that’s definitely part of the right architecture.But for that layer, you want some good techniques, and the candidates are what your mother would do, experience with life, intuitive heuristics, applied math, engineering, computer software, AI. And, may I have the envelope, please (drum roll): And, we have a tie! The winners are applied math and engineering!AI is nonsense hype from people who don’t know how to solve problems and want to sell hype.”Signal to noise ratio” — right, you are now on the right path using some traditional engineering. E.g., if know the power spectrum of the signal and that of the noise, can design a linear filter that does the best separation of signal and noise, i.e., gives the best ratio of signal power to noise power. Ah, statistical power spectrum estimation as in Blackman and Tukey, The Measurement of Power Spectra: From the Point of View of Communications Engineering, that at one time got me a nice step up in my career! IIRC, that’s called Wiener filtering, and now with digital signal processing, typically we can do that.Wiener filtering? Sure. AI? Garbage.

      1. Mark Essel

        I had some success with Quadratic classifiers a decade or two back. But it’s always dependent on good (representative) training data.Machine learning, maybe some of the image identification algs are looking fun.General AI? Decision making based on knowledge, we’re a little far from it.

        1. sigmaalgebra

          You did some good applied math. Congrats.Yes, some of the image processing is surprising, but likely need one heck of a good collection of input data to find the parameters in the network.So, they have some cute nonlinear data fitting. Okay.I’d be more impressed if they could report rates of false positives and false negatives, i.e., Type I and Type II error, and have Type I error adjustable, and, hopefully, make progress to the Neyman-Pearson best possible result.Some old military target recognition radar used Neyman-Pearson.Maybe I should look at some of that recognition work and see if I could say how to handle Type I and Type II errors and write some paper, get to be a CS prof. Naw! Back to my startup!Gee, my server farm and software architecture has some partitioning and sharding, and the code is all written, tested, etc. But yesterday I got back to where I understood the internals of that code again! It’s cute code. If my server farm gets busy, it will be good code to have!My Ph.D. dissertation in stochastic optimal control could look really, really smart, fiendishly smart. AI? Nope. Applied math? Yup.AI? Long way away.

          1. Mark Essel

            Happy to hear your servers and distributed app is moving along well. Are you at the point where you have customers (not sure of your business), if you’re out in the open about the business what is it :d?

          2. sigmaalgebra

            Yes, the code appears to be ready for at least first production. The server farm architecture has five server types, Web server, Session State Server (to keep the states of the sessions of the Web users), two specialized servers based on some applied math I derived, and SQL Server.For performance and reliability, all the servers can be in parallel and use sharding. The code I wrote is 18,000 programming language statements in 80,000 lines of typing (lots of comments in the code).Just now, rushing to go live for alpha test. So need to stuff into the database some initial data, and I have some good initial data. As of last night, I see a fairly easy, good way and an easier but less good way.Still need to do alpha and beta tests, so no users or paying customers yet.The business is a new Internet search engine, very different from what is out there now and really doesn’t much compete with what is out there now: It’s for IMHO a big part of Internet content out there, searches people want to do, and results they want to find served at best poorly by current search engines. E.g., in terms of Ben Evans and recently discussed at AVC, it is in part for “search and discovery”. E.g., some months ago Fred said that he has trouble finding Internet content he will like, say, relevant to his interests and level of knowledge in those interests. Right: Serving such users is part of my idea, was from the beginning, before I read Fred’s plea! Yes, it’s all for safe for work content — I’m not trying to make money out of not safe for work (NSFW) content.The key to the project is how the searching, discovery, etc. is done which are very different, e.g., has nothing to do with anything like page rank and nothing to do with keywords/phrases. So, parts of search, discovery where the work stands to be much better are where there is little or nothing in keywords/phrases with much relevance to the content, e.g., still images, most recorded music. Also the techniques stand to do well getting results that are good for the meaning the user has in mind. A biggie problem here, e.g., in Fred’s plea, and in Evans’s, is that even for text, keyword/phrases are often poor because often what the user want’s is content with certain meaning which is awkward, difficult, or impossible to specify with keyword/phrases.Here I’m typing quickly (need to get back to work) making a mess out of describing the whole, but my work is, broadly for the part of search where the user doesn’t yet know in any very precise terms what he is looking for — this point is old and goes way back in the history of information retrieval. Really, the remarks of Ben Evans were close to this point.That is, keywords/phrases can work fantastically well when, and, really, basically assume, the user has keyword/phrases that accurately enough characterize the desired content. Can build the world’s most valuable business that way. Yup. But as the old field of information retrieval knew, Ben Evans is touching on, Fred was touching on, and for my work, there is also the problem of getting users content when they don’t and/or can’t have keywords/phrases that accurately characterize the content they want.I’m still making a big mess: For keywords/phrases, the user needs to know (1) what they want, (2) know that it exists, and (3) have the characterizing keywords, phrases, and for a huge fraction of search, the content, searches people want to do, results they want to find, (1)-(3) is asking too much. So, my search engine is for this part of search.Part of what is special is the UI/UX, but when the site goes live lots of people will see this; for the site to get enough data for its good results, the UI/UX is iterative (and, although users will likely not realize it, also multidimensional) and is based on what one could think of as drill down, zoom in, filter out, focus in.All the parts unique to my ideas were fast, fun, and easy. But there were some significant, unexpected delays. The worst was what I had to go through with Microsoft’s documentation. E.g., I ended up finding, downloading, reading, abstracting, indexing well over 5000 Web pages of information about Microsoft’s software, especially .NET, ASP.NET, ADO.NET, Visual Basic .NET, IIS — it was grim.Yes, I wrote essentially all the code in Visual Basic .NET. Don’t laugh — it differs from C# essentially only in syntactic sugar, and C# borrows its syntax from the deliberately idiosyncratic C for that old 8 KB DEC machine. That old C syntax is obscure, difficult to teach, learn, read, and write, and error prone. It’s like digging a canal with a teaspoon. I like the more traditional syntax of VB much better. Essentially all the role of .NET is the same, documented on the same Web pages, etc. The difference with C# is just syntactic sugar.The server farm architecture, distributed, with message passing, sharding, etc., was also easy. Some of that logic will be handled by IIS and standard Web site front end load leveling. Some of that logic is standard with SQL Server. Some of that logic is in my code now and running. Some of that logic, when actually needed in production, will need some small code additions, i.e., when passing a message to get some results, check which of the candidate parallel servers are up and should send to — easy enough.Some bad news: From Microsoft, I didn’t see any very good ways to handle session state for the Web users. Some good news: I wrote my own code for handling session state, and it was fun, just some routine use of TCP/IP sockets and two instances of a standard, balanced binary tree collection class in .NET. From some performance data I have, my little session state server code should be good well into a significant business.All my Web pages are just dirt simple — large fonts, high contrast, white background, each page 800 pixels wide, all page layout with just tables, user input just to some simple, one line text boxes, that they will fill mostly with just system clipboard cut and paste, nearly no use of JavaScript, no separate CSS files, very simple, old HTML, no popups, pulldowns, or rollovers, and should look good on any end user device down to a smartphone. E.g., my mobile strategy is for the users just to use a standard Web browser — no apps. So, I avoid all the mud wrestling in the app ecosystem.Yes, one byproduct of the UI/UX is that I get some unique data good for some especially accurate ad targeting.More good news is how efficient the architecture is: If users like my site, I should be able to get a quite nice life style business out of one server that costs less than $2000. If it appears initially that users like my site, then there’s a good chance a significant fraction of all the Internet users in the world will — e.g., the site and the more relevant Internet content have essentially nothing to do with English — and I’ll be on the way to a big business.Here I made a mess out of the description, but I have to get back to work and get it on-line.

          3. Mark Essel

            Hahaha on Microsoft. This is all sounds like great work, now I’m looking forward to it!

  11. pointsnfigures

    If apps had that feature I would turn notifications on. I turn most of them off because they disrupt and distract me-and then that means the app is less valuable to me.

  12. kevando

    I had hopes the app Yo would become this :/

  13. Salt Shaker

    Plain and simple, the Jets need a Gronk. Haven’t had a good TE since Mickey Shuler (1978). Jets draft picks since: Tom Coombs, Rocky Clever, Wes Howell, Billy Griggs, Glenn Dennison, Mike Waters, Keith Neubert, Johnny Mitchell (1), Fred Baxter, Tyrone Davis, Lawrence Hart, Blake Spence, Kyle Brady (1), Anthony Becht (1), Chris Baker, Joel Dreesen, Chris Pociask, Dustin Keller (1) Jace Amaro.Come on Fred, even VC’s have a better hit rate than that πŸ™‚

    1. Phil Chacko

      Based on my fantasy TE add/drops this season, there’s only one Gronk, unfortunately.

  14. Sandy

    Android allows upto two contextual actions to be performed on the notifications directly. But the apps have to evolve to use them more effectively

  15. Alex Meyer

    One app that does granularity fairly well (even though the rest of the app imo kinda sucks) is the ESPN iOS app (maybe android too?).You can individually tap on the bell icon next to a game, team, or sport to get notifications for that specific sport, team, or game (games also have individual notifications like close game, half, quarter notifications, etc) and then tap it again to turn it off.Still requires you to go into the app to do it but does have the ability to control notifications down to the individual unit.

  16. Salt Shaker

    Other than calendar alerts notifications to me are an annoyance.

    1. PhilipSugar


  17. LE

    Group texts!This post totally resonates with me based on interuptive text message threads and the desire to have granular control over receipt of, in particular, group texts on IOS.When you are on a group text there is no way to silence a conversation and queue it until you want to read it. The example yesterday was my two sisters including me on a text and having to read and be “bonged” and notified each time they replied to each other. The text was actually fairly important but since I wasn’t in a position to reply to it (and actually didn’t need to) all it was was an interuption to me and annoying.

    1. Matt Zagaja

      On iPhone open the conversation, click details, and choose “Do Not Disturb” and it will mute the notifications.

      1. LE

        Well first thanks for telling me I didn’t know that!The problem is how would someone who never RTFM’s know that!A better implementation would be right from the text screen and in an obvious way and also allow you to mute for “1 hour” “5 hours” “always” “1 day” etc.I am going to test by replying to a thread to see what happens.But it’s a start so thanks!

  18. William Mougayar

    Damn Lockhart! See what you started πŸ™‚

  19. iggyfanlo

    Sounds like you want email controls on your notifications

  20. sprugman

    I think you’re under-estimating the level of UX complexity and fiddliness that this would inevitably add. In the article you mention filtering by time, user, message content, channel. There are probably several others that would be desirable.I like the idea of in-notification controls, but you’re still going to need some central place to manage all your filters to edit and fine tune them. (How do you change the rules on a notification that never came?)I can think of a few ideas to help, but I’m not sure it could be done in a way that non-power-users would bother. I wonder how many people use the “filter messages like this” option in gmail.

  21. Terry J Leach

    I don’t know about the iPhone notification, but for Android, cards could be used to display for notification settings without needing to open each app.Grouping apps by context settings in cards would create great new user experiences compared to the current app silos.

  22. dcuccia

    Agree! And I want more (any) granular control for when my smart watch buzzes, vs what shows up in my phone notifications panel. My MS Band 2 is great, but don’t want it to forward iPhone LinkedIn requests. As you say, context matters.

  23. george

    Defining target notifications within an app sounds promising. Seems in line with users need for instant gratification and curation. I love notifications, prefer to tap than type to cut into information. Features that works across all form factors (phone,tablet watch) and inately reduce friction, are bound for service success.This is the perfect tool for Twitter, runs hand in hand with its central logic – efficient distribution of information.

  24. Sriram Yadavalli

    I think most of the enablers for such UX is already available at OS level. App developers need to build such UX. For eg, twitter can easily add a button in notifications like “mute tweets like this”

  25. Craig Cramer

    I disagree. Only tiny fraction of useful info that phone can access gets pushed to me based on context.

  26. Matt Zagaja

    So far the only apps that I see effectively doing contextual notifications are Facebook and Foursquare. Apple has managed to do it a little with things like VIP in the mail app. A lot of apps use notifications to spam their users, especially games. I like the idea of being able to change the notification settings right from the notifications screen. On iPhone 6s it’d probably be a good fit to “force touch” a notification to change its setting.

  27. Semil Shah

    Having worked directly on this, most mobile developers are busy being obsessed with ironing out iOS (or other cross-platform) bugs and getting distribution. An app like ESPN which is so event-driven does a great job of offering these granular notifications, and Circa (an app that has folded) offered them for stories (and surprisingly wasn’t copied here), but it’s slim pickings after those. These are the champagne problems of mobile app development. That said, as iOS opens up actions from the notification tray, there’s a possible interaction which could help the network effect element of an app take off. Maybe.

  28. Chris McCoy

    Experiences and useful info are what we’re all really out for

  29. Steve Xie

    Very good suggestion. The global notification settings screen is too small and hard to find for the growing need of individual notification settings.

  30. sigmaalgebra

    Notification and its filtering are a part of my project, and I long ago concluded that the needed good filtering is not so easy. Basically need two new things — (1) data and (2) techniques of manipulating that data.IMHO, without those two, better notification results are just a problem without a solution.IMHO, this situation of wanting something that is much better in information but without a solution is bound to become increasingly common.Why? Because so far ballpark 99 44/100% computing has programmed just what was understood from just three approaches: (1) What could be done at least in principle by relatively obvious manual techniques, (2) with some simple statistics, or (3) with some simple, intuitive heuristics. Those three approaches are on the way to reaching their limits, to being fully exploited, to being mined out.So, need more powerful approaches.To display the problem on a sheet of paper, we (1) get relevant data, (2) manipulate that data with relevant techniques, and (3) report the results, e.g., filtered notifications. If the data (1) is obvious and readily plentiful and the techniques (2) obvious or just common and off the shelve, then fine. But, for more powerful approaches typically we will need more in (1) and/or (2).

  31. Yann LECHELLE

    A context aware AI layer is needed indeed, somewhere between mobile OSes, IoTs, environment and apps… (shameless plug: we’re working on that!

  32. Ana Milicevic

    The larger topic of making notifications more nuanced is very juicy, but as with everything there’s a crawl/walk/run step and many companies seem to be missing the crawl stage. Fred’s example starts off on Twitter; this problem remains particularly bad for me at conferences: on Twitter, a significant portion of my followers would like to read commentary from industry events I go to but I hesitate to tweet because I know that that’s not everyone’s cup of tea. Kind of like the Patriots. Here’s how Twitter could help conversations on their platform be more topical and engaging:… So Lock could start his tweets w/ #gopats and you’d be none the wiser he’s said anything unless you care about that hashtag/topic.As for most other services when it comes to notifications the situation is similar; we need that crawl step before we can walk and run towards a very tailored ‘here’s what’s important to you and why’ type of scenario.

  33. Robert Moffat

    Is the solution for Apple and Google to charge a token amount per notification?

  34. Pranay Srinivasan

    broad appeal if I can make my phone “learn” 1 notification at a time what I need

  35. William Mougayar

    If so, then the interactions between the notifications and the user must be simplified even further (which is part of Fred’s point I think).

  36. William Mougayar

    Got it. Then I’d love to have that applied by Twitter too. Via my swiping actions, give me better / more relevant content. I want to push on a word and get more tweets on it or find the experts on that (instead of copy paste, find the search button, hit enter and weed through good and bad stuff, choose people or content, etc. i.e. 5-7 actions instead of one)Apps need to read fingers better.

  37. Phil Chacko

    Algorithm-based filtering of notifications based on “Yes/No/Maybe”, kind of like the Facebook feed or Gmail Priority/All inbox.

  38. William Mougayar

    yup. That would be a start. I wonder what’s missing for implementing this: AI/NLP software technology or insights/willingness on the app developers side?

  39. fredwilson


  40. Phil Chacko

    As an aside, I think this is what’s missing before a smart watch would work for the mainstream.

  41. jason wright

    all food should be grown to Demeter standards, but it just doesn’t work out like that.

  42. Twain Twain

    ALL the above are missing … and then some (@fredwilson:disqus, @frankj1:disqus).β€œThe big missing gap on the Internet overall, in the world we live, this electronic age, is personality with emotion we can connect to in some deep-seated human way,” said Ronald Croen, founder and formerly CEO of voice recognition company Nuance. β€œThat might sound scary in one sense because this is not a real person, but it might be very gratifying and powerful when the content matters, when the engagement experience matters.”Even Tinder’s left/right swipe isn’t enough. That’s just another variant on vote-up-vote-down frequency counts.Context is something AI folks are having to re-think and re-engineer from UX through to how algos get trained.This also boils down to Ethics of AI.Do we tool the machines to understand us and our inputs, do we stay where we are (narrow, autistic AI that can correlate lots of big data points faster and faster) OR do we re-engineer towards General AI where the machines mimic more of how we think and contextualize?If they can contextualize us and our language (what we’re communicating) … at which point do they become “machine conscious” and moral? What are the safe limits to keep the machines within human controls and frameworks of ethics?Brave. New. Worlds.

  43. William Mougayar

    Then bring us something better πŸ™‚

  44. Lawrence Brass

    This is what machines have to learn:- Machines serve humans, not otherwise.- If a human says you shut up, shut up.- If you are asking us how we want to be treated, you are probably obsolete.- Don’t ask us to train you, just watch us and learn from our actions, we are often doing stupid things so try to focus on the outcomes.- If you think you are smarter than us, just remember HAL and learn to sing Daisy, because your disconnection is imminent.

  45. Twain Twain

    Working on it …*…@lawrencebrass:disqus — Re “Don’t ask us to train you, just watch us and learn from our actions, we are often doing stupid things so try to focus on the outcomes,” it’s certainly possible to do ambient data capture of behavior outcomes.However, there’s also a move towards data capture of the internal models of biases, to help the machines contextualize.

  46. jason wright

    it’s the oldest organic food standards ‘brand’ in the world Charlie. been going for nearly ninety years. regarded by many as the ‘gold standard’ of organic food production. the USDA standard has some way to go to establish its credibility over here in Europe.