200x Growth

Back in 2011, my partner Brad suggested that USV invest in a search engine called DuckDuckGo.

I laughed at the idea, “why would we ever want to compete with Google?”

“Because they do something Google will never do”, Brad explained.

That thing was private search, no storage of search history, no storing of personal information.

DuckDuckGo was doing about 100,000 searches a day when we had that conversation.

Last week, they had a day in which over 20mm direct searches were done.

That’s 200x growth over the six plus years we have been invested in DuckDuckGo.

Brad was right, of course, and I saw that pretty quickly as did everyone else at USV and we made that investment.

And I’m very glad we did.

I suspect I’ve told this story a few times now at AVC.

I love it so much.

#VC & Technology

Comments (Archived):

  1. JimHirshfield

    I like the way this dovetails with GDPR and the growing trends of consumer privacy.

    1. WA

      Agreed. It will be interesting to see how usage scales again after May of next year. What part of growth so far derives from EU or beyond US?

    2. LE

      GDPR: A new money making industry (created around it) and a mess just like HIPAA has been in many ways. Typical idea implemented by attorneys and academics with unintended consequences even for people in the US. (US companies have it in a wad about this one). Then those who write go around like chicken little about how bad things can be if you don’t comply.http://www.ibtimes.com/eus-…Bottom line: The cost of doing business increases and we all pay for it. Thanks Europe.

      1. PhilipSugar

        I agree and disagree. I agree it increases the cost of doing business. It makes me have to deal with lawyers and accountants. It gives idiots at big companies something they can justify their job with…..do you have SOC 1 SOC 2 GDPR? These desk jockeys don’t have a clue. I got this just this week…..what database do you use in your datacenter? MS-SQL Server, that’s not a real Relational Database we only want Oracle. Really? When was the last time you programmed, let’s just talk about a detailed project you did……you are describing a project what were the code obstacles you hit? Ummmmm, I wasn’t really programming.So you have a clue?But on the other hand, you are the one saying you can only be as honest as your competition. If you just take the info you have and sell it that means you win against somebody that doesn’t. If I know you have a medical condition which requires you to have a robotic assisted foot, should I sell that do any drug company I feel like?

        1. cavepainting

          Cost of doing business is one consideration. There are also real human needs around privacy and consent that need to be respected. Or else, it becomes a free for all world where our most personal data is up for grabs. When so much of our life is digital, it is critical that individuals have some semblance of control over their data.I am a big fan of GDPR and think it is in line with what individuals (people) want and aspire for. I believe similar regulation will eventually find its way into Asian countries. The US is the last frontier…

      2. cavepainting

        Maybe, but we have to draw a line somewhere. Companies have gotten away with murder by selling personal data indiscriminately to the highest bidder.If someone were to put stickers on your body as you enter and exit shops within the mall and if adverts were to dynamically pop up around the mall matching these stickers, you would not be exactly thrilled. At least, not if you did not provide explicit consent. But this is exactly what happens with web sites, ad-tech, and the data brokers.Why would we allow this on the internet when it is so creepy and inconceivable in the physical world?There is a human being behind every browser instance who deserves respect and privacy unless he/she chooses to explicitly trade it off for something else.

  2. pointsnfigures

    Search is still better on Google but over the past couple of years I have seen DuckDuckGo get significantly better. I put DuckDuckGo as my default search engine in my browser. Obviously it correlates to user base. I tried using Brave’s browser for the same reasons. I used it for a few months but it didn’t have the functionality of Chrome or Safari yet so I switched back to Chrome and will wait for Brave to get better.Not being tracked etc is going to be a bigger and bigger deal. Will be interesting to see what the DOJ does with Facebook and Google now that they are suing to block the ATT/Time-Warner deal. FB and GOOG are really in the advertising and communication business no?

    1. Vendita Auto

      Did the same as the default some time back but went back to Firefox ditto reasons, Chrome is better but !! Have to say back then it was a pain to shake DuckDuck off/ out of my systems.

    2. Lawrence Brass

      I have the same experience, duckduckgo results keep getting better, might be a consequence of the user base increase.Not so sure about Google search, I still use it for more specific searches. The first page of a Google search usually gives me 4 ads and 2 obvious results coming from wikipedia or stack exchange, I have to scroll down to find the useful stuff, if any.Not being tracked was always the deal, we have been conned into thinking otherwise.

      1. pointsnfigures

        True about the ads, and they have only gotten worse.

  3. Dan Pullman

    Curious how you think about exit on this company now that you are 6 years in? Is the Fund life 10 years?

    1. fredwilson

      I never worry about an exit. The opportunities always come for the good ones and this is most definitely one of them

      1. JamesHRH

        Someone has to say this – that astonishing growth still leaves DDG w <1% of The GOOG’s dailies.Do they do that comparison?Or is he metric just usage?

        1. Rob Larson

          GOOG is valued at $720B as of this morning. With $20B earnings on $90B revenue in 2016.Anything that is approaching 1% of that kind of business has to be considered a huge success, regardless of market share. (Obviously they won’t have margins as strong as GOOG, but the point still holds)

          1. JamesHRH

            I think it’s debatable, whether that ‘point holds’.I have heard countless investors say ‘Don’t tell me all you need is 1% of some huge market.’.They typically mean – don’t me too a huge market and think you will have success @ 1%. DDG – as Brad rightly pointed out – solves the search problem using a unique attribute.Just not a very popular one.

          2. Rob Larson

            I agree with your point that, in almost all cases, investing in hopes of getting 1% is foolhardy, because that first 1% is by FAR the hardest to get.This was a unique situation, with a realistic chance of getting that 1%, because they compete on a vector that is off-limits to the monopoly holder’s business model.

          3. JamesHRH

            You can tie yourself in knots on wording this stuff – it’s a new category of search and has found a niche market.I don’t know what the monetization to margin ratio is on $20M dailies, but I am guessing that if it sucked, this post wouldn’t exist.

          4. fredwilson

            that was basically the argument that Brad made to the rest of USV. if DDG gets to be worth $1bn (0.14% of GOOG’s market cap), it would be worth more than the entire USV fund it is in.

          5. Alessandro Cattelan

            Since DDG is unable to target ads, I assume that taking 1% of the search market, does not mean that it will also take 1% of the advertising money. How much are advertisers willing to spend to have theirs ads on DDG?Note: I’ve just run a few random searches on DDG for popular consumer products (e.g. jeans, OnePlus 5T, iPhone X) and it was presenting me with irrelevant ads (e.g. from target.com or other stores which I cannot get access to being based in Italy as I am).

          6. LE

            Sorry to beat up on DDG but to your point I just did a search for airline tickets and got a company that is in the UK while I am in the US (and not using any proxy etc.)… https://uploads.disquscdn.c

          7. Rob Larson

            Perhaps different horses for different courses – do your airline tickets searches with Google, do your “what is a dirty bomb anyway” searches with DDG. 😉

          8. LE

            This reminds me a bit about how when you talk to some people over the phone they don’t want to say certain things over the phone. As if there is someone tapping their phones, in particular the government, (without a warrant no less) that is ready and able (and has the budget) to spring into action with a raid confiscating everything and tossing their house. Meanwhile they regularly break traffic rules and cheat on their income taxes.Keep in mind that any info that comes out such as ‘and they found searches for X’ that you read about in the press is after someone actually does something that warrants the police having reasonable cause for a search warrant in the first place. And even that won’t necessarily hold up in court if there isn’t other evidence and of course if a crime hasn’t even been committed.Even with ddg what you search for is saved on your computer as web history which means it is backed up if you have those running (which you should) such as Apple’s time machine.For example:”https://duckduckgo.com/?q=a…”Appeared in my web history just now.

          9. Rob Larson

            All good points. That said, some people (including my wife) don’t trust google and probably never will. I am persuaded by both the points you mentioned as well as what I believe are the motivations driving Google’s behavior, so I am comfortable using them. A few people (less than 1% to date) are suspicious of Google enough to inconvenience themselves.

          10. Donna Brewington White

            That is a really good point.

          11. JamesHRH

            Having a unique attribute for doing the search job is the key, I think. If DDG were BING @ 1% market share……..not so much?One thing is for sure – which, again, Brad was likely thinking – hugeness solves a lot of problems.A very good bet and, as I recall, another USV investment that was not a bidding war. Most VCs didn’t get it, yes?

          12. pointsnfigures

            I would have passed as an angel six years ago. What are the odds of winning against Google? Other investments I see (like Carta) I would have jumped on in a heartbeat. It’s always tough to pull the trigger when there is a big kahuna in the market. DDG has a twist that Google will not match. Similar to other legacy companies when they ran into internet competition.

          13. LE

            DDG has a twist that Google will not match.Google could easily brand another engine under a different name if they wanted. Re-purpose their existing content. [1] There is literally no barrier to entry for them to do so. And it wouldn’t impact their existing user base either. That is I don’t think it would cannibalize any existing users.I am not seeing any barrier to entry for Google given that DDG has, what, 20 or 30 employees from what I read.Now would google waste time on this? Probably not. Not now. In the future though anything is possible.[1] Similar to a known brand that makes, say, washing machines also selling the same exact product through another channel.

          14. cavepainting

            It is not so much that Google cannot do this. It is that the people who use DDG value privacy very highly and Google’s brand simply does not fit with that. Even if Google were to build something with the privacy features of DDG and launch it with a different name, the DDG user base is unlikely to use it, because they might not trust the brand.

          15. LE

            has to be considered a huge successIt’s a huge success to the team that started and operates DDG. For the investors the only huge success would be if they get a huge return or some other huge tangible benefit.This also reminds me a bit about real estate BS such as ‘if this house were located in _____________ it would be worth 20x as much!’.

          16. Matt A. Myers

            Ad revenue streams will disappear sooner than later.

      2. Jonathan Libov

        When I’m out and about in tech circles and I tell people who work in investing that I used to work at USV, they usually ask me about dealflow vs. thesis-driven investing. The simple reply is that the former focuses more of their efforts on studying markets, the latter studying history. The more nuanced answer is that the former are generally thinking on five year horizons — i.e., what could be a solid business in five years with trajectory toward an exit? — vs. ten year horizons. The world changes so much every ten years that, when evaluating an investible opportunity, you need to suppress whatever instincts you have to assess whether the product or business is really going to be a thing, because when you do that you’re implicitly thinking on a five year horizon, a horizon that looks more or less like the present. But on a ten year horizon, how things look today is bordering on the immaterial. Hence it requires an understanding of macro cycles and a lot of faith in the quality of your study of those cycles, because you’re counting on a future world to embrace something that the present and near-term future may not value all that much.That’s what this graph of DDG search volume is screaming to me.

        1. Jonathan Libov

          To be clear, this articulation of dealflow/market-driven vs. thesis-driven investment is merely something that I understood after two years of working at USV, because for most of us, implementing that kind of advice itself would require a ten year horizon of learning how to invest.

        2. andrewparker

          I agree and had a similar experience while working at USV. That said, my biggest takeaway from this graph is the value of compounding growth. Eyeballing this, it looks log-linear to me. DDG figured out how to turn a 1 user into 2 on some reasonable time horizon, and then had the patience to let this cycle repeat itself consistently. It’s not just patience: All marketing channels generally deteriorate over time, so they found new sources (such as the browser distribution deals flagged on the chart) to keep the 1-into-2 compounding process fed. But often patience is the most mentally challenging part of the formula, because in the early years its hard to convince your gut that the small upward bend in your graph is actually log-linear and not just linear, and thus, patience will work.[insert your choice Warren Buffett quote about the value of compounding growth]Kudos to Gabriel, this is quite impressive.

          1. sigmaalgebra

            First-cut, ballpark, early on, which is so far, it’s exponential growth,y = a exp(bx)for x on the X axis and y on the Y-axis and for some two constants a and b.Why? That’s roughly, ballpark, first-cut, what viral, e.g., word of mouth, growth does.Why? Given time t and revenue y(t) at time t. We agree that for the present we have t = 0. Let b denote the total potential number of users. Then at time t, the number of users talking about the service is proportional to y(t), and the number of other relevant people hearing that talking is(b – y(t)).Then the rate of growth, the calculus first derivative,y'(t) = d/dt y(t)is proportional to the number of people talking and the number of people listening. So, for some constant of proportionality k, we havey'(t) = k y(t) (b – y(t))So we have an initial value problem (that is, we know y(0)) for a first order (we use only the first derivative) ordinary (no partial derivatives) differential equation.Then from calculus,y(t) = y(0) b exp(bkt) / ( y(0)(exp(bkt) – 1) + b))For k, we can fit to the current growth.When t is small, then exp(bkt) is close to 1, the denominator is close to just b, and we have abouty(t) = y(0) exp(bkt)for just simple exponential growth. That exponential growth continues until y(t) becomes a significant fraction of b; for DDG, that’s a long way off.But as y(t) continues to rise, it approaches b asymptotically from below.So, in total, y(t) is a lazy S curve.Below is a sample of solutions for several values of the constant k.Yes, once at FedEx, this analysis pleased our two BoD representatives from BoD Member General Dynamics, got them to stay (they had their bags packed and plane tickets back to Texas), and saved Fedex.

      3. JaredMermey

        Beyond track record, how do you convince LPs to take this mindset?

  4. CJ

    I’d probably use DDG more but I’m an Android, Chrome, Drive, Google Play, YouTube, Gmail, Maps user. What I mean by this is that Google already knows my life story, using DDG won’t buy me any privacy.If I get the urge to get out of this ecosystem though, DDG would definitely be in play.

    1. LE

      Also you are leaving cookies all over and wherever you go most likely. Third party or otherwise. So the search engine you use is only part of the privacy picture.

      1. CJ

        True, though I’m blocking most of those except where it benefits me. But that is a point well taken.

  5. Ronnie Rendel

    Ahh DDG was a key component of my Dark Web experience (Alphabay RIP)

  6. sigmaalgebra

    Gee, some VCs regard as success only an exit of $1+ B within 10 years. So, DDG is already 6 years old.Okay, let’s see: 20 million searches a day, two ads per search, $2 per 1000 ads displayed, give annual revenue of2 * 20 * 10**6 * 2 * 365 / 1000 = 29,200,000dollars. If count essentially all that revenue as pre-tax earnings and evaluate the company at 40 times annual pre-tax earnings (maybe optimistic), then have market capitalization of40 * 29,200,000 = 1,168,000,000dollars. Okay, that’s a good exit!

    1. cavepainting

      | If we count essentially all that revenue as pre-tax earningsThat’s not true unless DDG has no operating expenses, is not depreciating any of its development and data center investments, is paying no browser companies or apps to integrate search into their toolbars…I also wonder if the click through on ads is a lot lower because of the self selecting audience that is seeking privacy.

      1. sigmaalgebra

        is not depreciating any of its development and data center investments, My understanding of accounting and tax accounting 101 is old and never was very good. Just now I’m concentrating on getting revenue, not worrying about taxes on revenue I don’t have yet. So, my views on tax accounting likely not correct:But, for your statement, suppose a company buys a building for $20 million and pays cash and, because the building has long term value, does not get to call this $20 million a current expense. So, the $20 million comes out of after tax earnings, maybe even savings already in cash in a bank and counted as stockholder’s equity. Then, sure, the building will be listed as an asset of the company.Then with straight line depreciation, each year for 20 years the company gets to claim that $1 million is an ordinary business expense from pre-tax earnings so is not taxed. So, really, that $1M can be counted that year as after-tax earnings.So, net, depreciation does not cut into after tax earnings but adds to them.Same for your depreciation of software development, that is, if the company wanted to or was forced to count the costs as a capital expense, that is, to capitalize the costs, e.g., salaries, of the software development.But, IIRC, commonly companies regard software costs, e.g., salaries, as current ordinary business operating expenses, opex, that comes out of gross, top line revenue before counting earnings. Then in future years, the software has no costs, depreciations, etc.Let’s get some ballpark numbers and see: Maybe a DDG search takes 1 second of CPU time, that is, for DDG to pass the search to other servers, get back the results, manipulate them a little, and send them to the user.So, the 20 million searches in 24 hours would be an average of( 20 * 10**6 ) / ( 24 * 3600 ) = 232searches per second. But the DDG server farm would have to be sized for the peak rate. So, ballpark, call that twice the average rate for( 2 * 20 * 10**6 ) / ( 24 * 3600 ) = 463So, DDG would need 463 processors. Call that 500.So, maybe they have Intel Xeon E5-2697V4 Broadwell-EP with 18 cores and 36 threads each. Maybe they put two of these on each motherboard for 72 thread per board. Motherboard? Glad you asked: Supermicro dual socket X10DAL-i. Then they would need, presto, bingo, and may I have the envelope, please [drum roll, please] a grand total [trumpet fanfare], queue uphttps://www.youtube.com/wat…500 / 36 = 14motherboards. In mid-tower cases could get all those on one wire rack, 18 x 48 x 72″ from Sam’s Club for about $100. Or using standard rack mount 2U pizza boxes, that is, 3.5″ high,14 * 3.5 = 49inches, only 49″ of vertical space in a standard rack.We’re talking one corner of a spare bedroom with plenty of room for a queen sized bed, a desk and work station, big screen TV, refrigerator with bottled water, etc., that is, all the essentials of a successful Web site!Let’s see: We’re talking 463 Web pages a second. So, call that ballpark 1 Mb per page. So, we looking at 1/2 Gbps, peak upload data rate. Just locate in JLM’s neighborhood and have choice of several 1 Gbps providers, just for consumers. Sure, need a static IP address.But, sure also have to get the results from Google, Bing, Yahoo, etc. So, maybe we’re talking a few Gbps. Okay, then just need a commercial building or a colo. The last time I checked at my local colo, it was dual 10 Gbps standard. Sure, want 10 Gbps NICs — those are routine now.For ad targeting, a Web server always gets the “user agent string;” for an old example I use in CURL,Mozilla/5.0 (Windows NT 5.1; rv:35.0) Gecko/20100101 Firefox/35.0Can use that as a user finger print. So, how much opex does that one rack in the corner take?So, we’re talking one guy plus one more to order the Chinese food carryout for lunches?Remember, we were talking top line of $29 M, and with that the opex is in the rounding error.Oh, it’s the capex of those 14 Supermicro puppies?Just checked, at Amazon, theXeon E5-2697V4has “FREE Shipping”! Can you fathom that!!!And the price, quantity one, is$2,979.23.Did I mention the free shipping?Call that $3000. And the server farm would need 28 of those babies. IIRC, the motherboards were under $500 each.14 * 500 + 28 * 3000 = 91,000dollars plus some rotating iron storage, power supplies, cases, cables, fasteners, maybe some copies of Windows Server, UPS boxes, Window AC units, a generator on a pad out back, a Cisco box, call it $150,000 capex. IIRC so far can expense something that small in one year but maybe need two.We’re still mostly in the roundoff error.I know; I know; I know; it’s unfair, so UNfair to look things up on the Internet and use arithmetic!So, it’s market cap of $1 B.Okay, but all that data is just keyword/phrases and only about 1/3rd of search.I’m going for the other 2/3rds so that’s very different. Also I’m going for only the “safe for work” parts, so that may be somewhat different. And I’m going for $1T instead of $1B.

  7. jason wright

    “I laughed at the idea” – precisely.

  8. creative group

    CONTRIBUTORS:It emphasizes that so called conventional wisdom isn’t always wisdom and smart is smart. Don’t allow past performance to get in the way of new streams of ROI.Brad is a smart MF.Captain Obvious!

  9. João Gonçalves

    I don’t use DDG because of one main reason: its lack of context. If you’re searching for a restaurant and there’s a restaurant in the other side of the world, a more busier side of the world for that matter, that has the same name, it’ll fail miserably at showing me the result I want.Most of the times, DuckDuckGo ranking algorithm places the results I want after the first 5 or 6 links. Sometimes it’s simply not there. Google acknowledgement of my own context gives them the advantage that all results are contextual and to the point.

  10. jortpossel

    A great story, well told. Congrats

  11. Francois Royer Mireault

    Am I right in thinking that search is a tactic for DDG? Consumer privacy is the strategy. And I feel it explains why they’re ready to sacrifice search results relevance to appeal to an audience that will ask for more than ‘just search’ in the future. @fredwilson:disqus, we need a follow up post on this… 😉

  12. george

    How does DDG search results (product) grow and become better if personal data is not stored? Seems counter intuitive…