Step Function Growth
I was at a board meeting yesterday and we were looking at a chart of growth. It looked something like this:
It was not straight line growth. It was stair step growth. Back in my math geek days, we called this a step function.
And this is what the user growth charts of most of our portfolio companies look like.
Nick Denton published a manifesto on blog design and blogging earlier this week. In it he showed this growth chart.
You'll note a similar stair step function in this chart.
Then Nick delivers the money quote:
We learned our lesson: aggressive news-mongering trumps satirical blogging. Gawker.com's growth since 2008 — from 300,000 people a week in the US to 1.4m — came in steps. After each story-driven spike — Tom Cruise's Scientology pitch video, Montauk Monster, Eric Dane's hot-tub non-orgy, iPad security breach, Christine O'Donnell's Halloween sleepover, etc — the audience settled back down, but at a higher level. The same pattern holds for Deadspin, which has ridden a sensational series of scoops culminating in the revelation that Brett Favre had stalked a buxom sideline reporter.
Growth comes in steps. There's a big event. Shaq joins Twitter and brings his fans with him. There's a spike. Things calm down, but they don't go down. Then a plane lands in the hudson. Another spike. Things calm down, but they don't go down.
That's how it was with Twitter and that's how it has been for most of our portfolio companies. The big events drive user growth. For Etsy it might be a seller being featured on Martha Stewart's TV show. For Tumblr, it might be Newsweek starting a tumblog. For Disqus, it might be Techcrunch joining the network. For Nick Denton, it is clearly "aggressive news-mongering."
The key learning is that, as Nick says, "the audience settled down, but at a higher level." Big events will drive audiences and some of them will stay. And you will grow in steps.
This is somewhat similar to stock charts when there’s support at certain levels and breakaways at others. I wonder if some of the same stock analytics might apply to these web traffic charts.How about time on site. Another important metric.What do you do when the chart is trending down. Time for a pivot if not too late?
depends on how long it is trending down fora month or two is no cause for panic
This underlines the importance of great PR/one-off marketing. Events, launches, stories do drive traffic and the value of that traffic is not in the spike, it’s in the people who stick around.It reminds me of the models we used to build for brand equity. A campaign may drive top-of-mind awareness higher in the short term, but the real value is in the long recall of those initial users that become loyal customers.
but relying on big, unpredictable publicity events (like getting on Martha Stewart, or Shaq using a service) is pretty difficult and, well, unpredictable. Are you arguing for more PR?I worked for a company in which we’d have these events – the spike (and corresponding step) would happen but the step would tapper off after a few days or weeks back our “normal” traffic level. If we didn’t follow quickly with another spike to compound on the first (and on and on…) we’d be back to square one. So it was a business of chasing hits, and since we couldn’t control their frequency, it was extremely difficult to keep up.
there was something about the product or service you worked on that was not right then
that’s true. The big bang events help new people try the product, but the product itself has to be good enough to retain those first timers into long term customers (some – mostly not all). If the product/service is not good enough to retain the customer, traffic/use will go back to pre-event traffic levels.
Have you seen this type of growth with AVC Fred? It’d be interesting to observe if there were any particular blog topics that caused a similar step change.
yes, but mostly driven by my occasional redesignsthe cleaner and faster i make AVC the more UVs we get
Same thing happens for Google – http://goo.gl/7zYuv – makes you just realize just how important speed is.
Ah yes, so it is not necessarily the design but the speed that makes a difference. Testing this would be interesting. Is it the design change, the increased speed or both that leads to increased traffic for AVC
Content is also important.You could test this with this new design, simply by adding a delay in everything loading and see the difference in bounce rates.It would have to be a longer term test see if bookmarker rates drop. But with quality content people are more tolerant as well. If you can get the same kind of content anywhere, then that will increase the importance of speed.
Sounds like you find a key traffic driver for your business when you start seeing these steps. The problem is making sure you really understand what is causing them and then repeating the process. I would not have expected a redesign to cause a big jump in traffic on a blog. If you go into constant iteration you could be bigger then the NY Times!
That’s interesting. This is definitely your cleanest design yet has there been a subsequent growth of engagement?(it’s also sort of interesting in the history of digital design -there seems to be equal pulls back and forth between hyperrealism and a clean modernism -and I have no idea where design will go)
Fred,Not sure if you chart is a dummy representation to drive home your point or an actual representation.If it is the latter, it is implicitly a straight-line function. Just change your time buckets on the x-axis to two-month intervals and you end up with a perfect linear equation!Re your broader point about “big bang events” shifting the base metric to progressively higher levels, while it does seem to be intuitively true (and possibly empirically proven as well), don’t you think this growth curve would be a lot more jerky/staccato?Cheers,Sumanth
it’s a dummyand if you change your time buckets to a year, then its stair step againhere’s the time series that generates the chart11.51.5222.52.5333.53.54
So are you taking KP’s $ at $4B?That’s the real questionBack to the topic at hand: you have to be ready for the stair step. That big spike – plane lands in the Hudson – might not have given you the stair step if you weren’t prepared for the growth on the tech side – and weren’t able to provide people with something useful on the product side.Too many companies are not ready for their 15 minutes – and just flame out.
twitter was not prepared for the growth until relatively recentlyintellectually i agree with you about preparation but twitter’s success does not prove that out
Twitter might be the exception. The idea was big enough for people to waitthrough the “fail whale” but I wouldn’t suggest that as a good strategy foreveryone.
I don’t think practically you can do that when you are a start up. It was the case in 1998-2000 when people were raising huge amounts of money and building systems that could cope with bucket loads of traffic from the day you launched. But that doesn’t seem to be the start up model at the moment. Having said that the current valuations and amounts that are being raised from VC’s might make that more of a possibility.
On the tech side perhaps not – but on the product side you can certainly beready. Ready with a great experience, ready with more functionality, readywith virality built in. On the tech side – thinking about a path of growthearly on will likely allow you to add on quicker (perhaps using AWS asopposed to a server in your closet). Not every site is going to be likeTwitter that becomes a global phenomenon overnight.
So maybe the plan is to build for a step or two ahead until your steps surge in frequency -Overinvestment can cause flameouts, same with underinvestments. A sense of balance is needed -to be just forward looking enough seems to be the right idea.
You know the old saying hope for the best prepare for the worst? Inthe start up world we live in today I think it is prepare for the besthope for the google takeout. I have to believe that with services likeaws, architecting a service to scale is as easy as it has ever been.Now I’m talking out of my ass here since I’m not an engineer or coderbut I have to think that there’s a fine line between lean start up andtoo lean for efficient success. Duster had a good post on thisrecently.Harry DeMott917-439-6600
I have the same experience with my blog. More spikey like Nick’s about posts that were a hit on Hacker News/Reddit. I’ve never had a “twitter spike” but some posts have made nearly a hundred shares.There are components I don’t understand, like why my juicy well researched posts fall flat, and my one liner video embeds explode. One off hand post about Minecraft has generated thousands of ongoing reads from search. I tried jazzing it up a little over a couple of weeks then left it. I figure if that’s the landing page people will see, I better make it a good one.
“There are components I don’t understand, like why my juicy well researched posts fall flat, and my one liner video embeds explode.”Like always, I believe there is an old saying that will ring true for you: “You don’t sell the steak, you sell the sizzle”
I’ve never been a great salesmen, likely because I’m mostly honest (wife:”I feel pudgy”, me:”you look great honey”) and I didn’t love the product, including stuff I’ve worked on. Give me a product worth believing in and I’ll sell waves to the ocean.
This step lift is what online marketers drive towards. Spikes that rise and fall are a fail and were invariably noise not value. Spikes that create new plateaus which translate to extended growth are the wins.This model is the core of seasonal businesses especially where a large chunk of traffic and revenue is around the holidays or Valentines Day for example. That higher plateau is the core to next years expectations as well and a key to understanding whether you are targeting and finding the returning customers, driving (hopefully) a higher value over time per customer for a single acquisition.
My favorite topic for right now traffic and spreading the “word”…Was wondering if Dunbar numbers play into it ergo you get 150 people visiting your site and say 15 of them visit a week and if you post something descent they communicate with the rest and to some degree you grow your network at the growth they grow their friendships as long as they keep coming back.Perhaps its just if you got good stuff on the site people check it out a few times.
Shhhh! You’re giving away all the secrets Fred!
Flip side…..how to give your competition a spike that paralyzes them: http://bit.ly/9U8TUc
Great story Andy! Get the manager in as a strategist 🙂
Double-flip – link to your site in the AVC comments and enjoy the spike and higher baseline traffic. 😛 More posts please Andy! I’ve long since read through the archives.
That is a great story on ingenuity…figures it came from the Midwest ;D
Got to love the the Mother Wit of such a move. Best story I have heard in a month at least. Thanks.
Actually this ripple and splash pattern is the exact same when it comes to social media spaces. those charts and that quote will be very useful for client presentations. thank Denton for me 🙂
This is the experience what i have from my own projects and from the internet companies i am invested in: “the audience settled down, but at a higher level.”
Completely agree with this. It’s how my blog has grown. The challenge or red flag is if the step falls back to or below the previous volume.You never want to fall too far.
I’ve used Google Analytics for a long time and started using the ‘annotations’ function lately to keep track of these ‘trigger events’ that cause a step. I’ve noticed over time I generally check occasionally until such an event happens (or receiving a notification via e-mail ala chartbeat) then slog through different segments, traffic sources, etc until I find the source of the spike. In essence, 99% of my time is spent ‘analyzing’ the data to just determine what caused it!Are there any web analytics tools that focus on this type of external event-driven mentality than pure segmentation and funnel/goals?
that looks way too symetrical
Excellent post. A lot of folks forget this when they create projections.Can’t wait to see how you will integrate the rest of Nick’s comments. AVC the video edition… 😉
i think he’s nuts about the front page and the emphasis on video
It seems like, for certain types of services, there’s an explanation for step growth to be found in looking at network graphs. Adoption may accelerate as it penetrates a local “network component” – a densely connected subnetwork of the larger/global graph (e.g. your friends or colleagues) and then level off until it finds a “local bridge” that connects one component to another. These local bridges are the so-called “weak ties” popularized by folks like Granovetter http://is.gd/i6HUW.A little abstract/academic, but maybe particularly relevant for services whose growth is driven by social distribution.
I’ve found the same thing is true for blogs. Although, the events aren’t necessarily some news thing. The events are usually finding a new method to market your blog which steps you up and generally doesn’t involve the spike.
This is interesting – I’ve seen this same adoption curve in healthcare, i.e. med device adoption / better hand washing practices / drug prescription behavior changes. The big/important new developments tend to get adopted in a step-wise manner…that is unless the gov’t/payor comes in and forces a change, in which case it looks like a 1 big step in terms of adoption, but upon “deeper” analysis, there’s usually poorer quality utilization of these forced adoptions.
I wonder if a firm like USV would invest in a startup that states their business model and growth strategy is based on events helping spur growth and show a step function graph something similar to what you have.It is nice to see these things from a historical perspective but if the businesses growth is based on random events happening and helping build up their growth I am doubtful they get funded, but then again VC’s would never invest until they see this graph to begin with, after all they are very risk averse.I wish there were examples of this kind of growth in sustainable energy ventures.
some are random, some are notetsy and disqus actively worked on the events i mentioned
Yes they may have worked actively, but they themselves did not create the events to begin with.What I am trying to state is if the business model for growth is centered around events happening that you take advantage of then one could say there are events happening all the time, one is unable to take advantage of them all the time too.Maybe it is possible to define that if certain types of events occur the business is geared tot ake advantage of it for it’s own growth.I personally don’t believe that one can draw investors based on growth from a step function where the variable for the growth is random, and I doubt that investors accept that and say ” yes I will invest in that because your growth is based on the step function”I would say in the case of disqus or etsy your investment was based on other factors, this was at the most just maybe icing on the cake.
This is exactly what we see at Rypple. Use of the app tends to spread from one team to another in companies. The various companies aggregated show a cleaner line, but the step growth is key to internal spread of the product!
This is something that I have preached in the past but didn’t have a name for it. Raising the bar over and over again is the key issue. You do it once and you better do it again. That is hard. That is why most sites get little to no traffic. As soon as you take your eye off the ball you will lose visitors and never get them back.
Great post and great point. While growing Box Office Mojo (which we sold to Amazon.com in 2008), we certainly found this to be the case. When Spider-Man broke the $100 million opening weekend record, we had a huge spike (thanks largely to a headline link from The Drudge Report). Things calmed down after that, but settled at a higher level than our previous norms. This pattern happened over and over again — big spikes followed by big drops that leveled off higher than our previous norms.
Insightful. Has been true for my blog.
This is a great post and thank you for sharing this nice experience,and hope you can give another posts as soon as possible
Thanks for your article, Now there is more reason to comment than ever before!