Sunday, July 14, 2013

Distribution's Day - Why the Rise of the Growth Engineer (or Growth Hacker) is now

In the late 1990's when you wanted to build a startup, a good chunk of your first expenditures went to building the tech, including renting a cage, putting boxes and switches in it, software for databases, code, and content management systems, storage space, and all of the talent to architect, build out, and manage these systems, back ups, and load peaks.  Additionally, you were building all your code, front and back end, from pretty much scratch.  It was rare to even run into a problem that a commercial solution existed for, much less at a reasonable price with fast easy integration.  But today, we benefit from a ton of what used to be the overhead of a tech startup being commoditized, virtualized, and democratized.  For example, servers on AWS, database at Amazon on Red Shift, open source content management systems, open source front end libraries (ok, mostly jquery), hosting for API's at Engine Yard or Heroku, with low cost pre-built off the shelf modules for all of the above, vastly reduces operational overhead.  And you can leverage solutions to send SMS with Twilio, email with a bevy of providers, take payments with Stripe or half a dozen mobile payment providers. Even services to manage all these cloud tools have been virtualized by companies like Pager Duty and many others.  The time and cost to build a tech startup are both dramatically down (Moore's Law dramatic) and the quality has increased simultaneously as these scaled point solutions grow up.

Now I would never suggest that there is not still bust your tail hard tech work that differentiates companies.  And I have seen the right platform choices give companies significant advantages over competitors plenty of times too.  But what I am saying is that the relative landscape is changed.  I am seeing more and more companies with great tech products stalled on distribution than ever before.  With product and marketing being intertwined, and design being a big part of the fabric that binds that together, I have lumped the whole lot into the overarching label "distribution," and the merger of these areas is probably the subject of another blog post another time.  But customer acquisition, or growth, or quantitative marketing, and the speed of it, can make or break two companies competing now.  Company A starts before company B but does not have strong distribution talent.  Company B starts later, but with tech hull speed so high now, they won't be far behind.  And if company B engineers the magic growth button for the market, company B wins.  As I have said many times before, you don't need the best product to win, and you don't need to be first to market to win, you just need to be first as the market consolidates, and then by definition you are winning, and displacing you in that market which you now own will be expensive and difficult.  At that point, you might not have to outrun the bear, just be an unattractive enough market that tech talent will go elsewhere to make their mark rather than try to do battle with you.

Several smart tech leaders I have worked with predicted this time to me.  Rob Meyer at Washington Mutual, Sunil Bopardikar at, and John Malek at Practice Fusion all foreshadowed this pretty clearly in conversations we had.  And some of my NextCard friends like Dave Schwartz at Cold Creek Technology, Rebecca Lynn at Morganthaler, and others way back when even talked about the markets that had been ripped open by scalable distribution techniques like the ones we helped develop in the late 1990's, and the ones that hadn't YET.  I think we all believed it a foregone conclusion that one by one every market would roll eventually.  And so at last we have Agile Marketing (go to the Agile Marketing Meetup in San Francisco if you can), Growth Engineering or Growth Hacking depending on your flavor, and the rise of Quantitative Marketing all employing user experience design, behavioral design, virality, direct marketing, inbound marketing, and optimization to try to win the distribution battle first.

So as you are spinning up your tech, make sure you either have a co-founder with some solid distribution background, or get help fast from an adviser or investor that is a growth engineer.  And avoid being one of those companies wondering how that competitor just ran you over with a lesser product.

Tuesday, July 2, 2013

Why I like 12 Month Break-Even as a Hurdle for Paid Marketing

A lot of people ask me how much they should spend to get a customer.  And I tell them there is no one magic number.  In the early days of a startup you might use one number, but that is only because there might not be enough volume to segment your return measurements meaningfully yet.  But as soon as possible, you would like to measure the return and cost of every campaign you run so long as you can do so at negligible cost, which you should be able to do with modern tools.

Anyway, as a rough rule of thumb, I like to suggest that startups might like to use a 12 month break even for a spending hurdle, so long as it is applied on a campaign by campaign basis and not an average for all channels and campaigns.  There are three reasons I think a roughly 12 month break even is a very comfortable place to start.

1) There is so much spillage in marketing measurement that I am certain that even in pretty tight channels, you are going to get a bunch of customers that you are not able to attribute to the spend.  Want to learn about channel interaction and spillage?  I think my friend Dave Schwartz is going to share some of his immense knowledge and experience at this month's Agile Marketing Meetup, so follow it and sign up when his talk is announced.  But basically, for every account that clicked through on your banner ad, setting a utm (urchin still really?) tag, there was another that just typed your domain into their browser or a google search bar that showed up as branded search or direct type in traffic.  And unless you have a solution that can nail down every banner impression to later traffic like Cold Creek Tech's or you are paying an agency like AdRoll that can do something similar when they serve your ads, you are going to identify that customer as direct type in traffic or branded search if you are on your game enough to track granularly with blunt instruments like Google Analytics.  Similarly, even if you are using a best in class conference lead tracking solution like Bloodhound, there will be some people that saw your booth, maybe even talked to you or approached it while you are busy, and they Googled your brand later.  So you won't be able to attribute every lead to their source.  The same can be said about every channel.  I will write another post sometime about measuring spillage.

2) There will likely be many customers that continue to generate revenue beyond 12 months.  So if you break even at 12 months, you will be building out the gravy train for all those longer term customers.  That will help you put snow on the revenue snowball and barrel down the hill.

3) Using the value that a new customer brought in over the trailing 12 months should be conservative.  I hope that any startup worth their salt can provide immensely more value and extract more revenue out of customers in the next year than they did in the last year.  If you can't, stop reading this blog and go fix that before you figure out how to ramp up paid marketing : )  In my favorite vernacular, I would say that your forward looking cohorts should be a lot better then your backward looking ones.  So if you spend to break even on your backward looking ones, you should be fine.

Now if you know you can beat these things, I would say do whatever you are comfortable with.  For instance, if you know that you have 50% spillage on a banner campaign, then consider relaxing even more to account for that.  And if you know that over the next twelve months your cohorts will generate 50% more revenue than the past 12 months, consider raising the bar 50% higher.  And if your attrition is really low and only getting lower, consider using more than 12 months of revenue for a customer.  Heck, consider discounting a long tail of years of revenue if you can model it and feel confident in it.  If you can make investors confident in it then more power to you, right?  And if 90% of your revenue arrives in the first 7 days of a cohort, then there is likely no need to wait twelve months to call it.  But I have seen 12 months applied to a lot of different industries and it has worked out pretty well as a rough starting point.

What?  You can't wait 12 months to see if something worked out or not?  I will write another post about working out proxies for 12 month value.

Happy distribution and growth engineering  : )