Lean Your Marketing: Find The Metric That Pays Your Bills


The customer funnel. We know what this is, right? It’s the progression of people through stages of your marketing, from (for example) searching on an ad, clicking on the ad, visiting your site, and signing up for or buying your wonderful product. All of those stages produce opportunities for tracking.  Any one-horse ad product is going to give you impressions, clicks, that kind of thing.  And the most basic GA implementation will give you site visits.  And because those numbers are easy to get and also large and easily juice-able… well, you’d be surprised at how many people consider those to be very important metrics.

Well, no less an authority than Mark Andreessen has a message for you: your metrics are bullshit.

Eric Ries, author of Lean Startup, is more genteel, so he calls them vanity metrics.  If you pay attention only to things at the top of the funnel– ad clicks, site traffic, page views– you can feel like you are making progress when in truth your growth is stalled or worse.


Instead, smart organizations look deep into the funnel to find the one KPI, the key performance indicator, that really shows the health of their business.  What is it? Well, it’s going to be different for different organizations.  The way I like to approach it with different clients is to ask them: what pays your bills?  If there is revenue coming into the organization, what is its main source?  Is it subscriptions? Product sales? Donations?  Then that’s the KPI.  Even if the company is pre-revenue you are likely to still have a business plan in place for how money will come in eventually; is it by acquiring active users? Is it by showing ads to a big user base?

The exciting thing about identifying this core metric is that it gives everyone in the organization the same focus.  Once you’ve agreed what the single most important number is for your organization, everyone should be adjusting their performance to making that KPI grow.  You won’t have your marketing people off looking at ad clicks and while your product people are looking at user engagement.  If your organization has agreed that Daily Active Users is your KPI, then that’s what everyone should be trying to produce.  Your product people will be tuning the site to increase this number, and your marketing people will judge their projects solely on the number of daily active users they generate.  Your customer service people might have targets for how quickly they handle tickets and amount of positive feedback from customers, but the overall goal of their group will be helping to maintain daily active users.  Everyone in the organization will be rowing the same direction.

Is it easy to track deep into the funnel?  Generally no, and that’s why so many companies don’t bother.  But is it worth it?  Absolutely.  Having a real understanding of what’s going on in your organization will lead to smarter, more confident decisions.

Once you’ve identified your KPI you’ll want to turn it into a goal; that’s the topic of my next couple of slides.


Lean Your Marketing: The Slide Deck

Slide 1 | Slide 2 | Slide 3 | Slides 4, 5, 6 | Slide 7 | Slide 8 | Slides 9 & 10 | Slides 11 & 12 | Slide 13 | Slide 14 | Slide 15 | Slide 16 | Slide 17 | Slides 18, 19 & 20



Lean Your Marketing: Start Early And Stay Focused


Aaaah, the Measure portion of the presentation. *cracks knuckles, gets ready*

So, it was kind of funny trying to figure out how to organize this presentation, because Measure kept wanting to come first, despite being clearly second in the Build-Measure-Learn cycle.  And the reason it kept wanting to come first was because of this recommendation: start early.  Many’s the company that literally tackles things in that order: build it first, then figure out how to measure it.  But you should think about measuring and metrics very early in the game because, people: Google Analytics is a great tool, especially considering its price, but it isn’t magic.  Not a few people have asked, once I got some tracking set up on GA, if it could look at past results.  Yes, let me just drape a little blue cape on it and make it fly around the world until time flows backwards.  You want a baseline of your early performance?  You’ve got to set up performance tracking early.  It’s pretty simple.

As for what you track, it can be surprisingly hard for companies to be disciplined when it comes to looking at numbers.  When I’m talking to potential clients I’ll ask them if they’re interested in more traffic, more leads, more sales, etc, and they’ll generally answer yes to everything.  But it’s not that helpful to try to optimize 15 different stats at once– instead you should focus on the numbers that really matter, which is the subject of the next couple of slides.


Lean Your Marketing: The Slide Deck

Slide 1 | Slide 2 | Slide 3 | Slides 4, 5, 6 | Slide 7 | Slide 8 | Slides 9 & 10 | Slides 11 & 12 | Slide 13 | Slide 14 | Slide 15 | Slide 16 | Slide 17 | Slides 18, 19 & 20


It’s Statistical Analysis, Not Jesus Appearing In Toast

Jesus Wept

Caring about stupid stats like impressions makes Jesus cry. Stop it.

Before my new boyfriend Nate Silver was forecasting elections, he created a baseball stats system called PECOTA that used advanced statistical methods to predict future player performance.  This was before Brad Pitt starred in Moneyball, so his fancy new stats were widely derided.  But he has at least one new fan, Sports Illustrated writer Phil Taylor:

I have surrendered to the numbers. I will make no assessment, athletic or otherwise, without rigorous statistical analysis…. I reject your anecdotal evidence, your hunches, your wishes disguised as predictions.  I will keep my gut instincts away from my brain and suggest you do as well.

And what has brought this swoon over Phil?  My new boyfriend Nate Silver’s election-night success.

Silver not only forecast President Obama’s reelection, he did it with uncanny precision, calling which candidate would win each of the 50 states despite weeks of heckling from more than a few pundits… That’s like hitting every jumper in a three-point-shooting contest while opponents rain trash talk on your head.

Well, actually, no it isn’t.  I’m not running down MNB Nate Silver’s skill or anything, because he’s my boyfriend and I support him unconditionally, but his miraculous achievement was the result of looking at polls that asked people which of several well-known choices they might make in a short time period coming up, and then believing the answer.  There weren’t a lot of unpredictable factors.  To be amazed by this is like asking people on your wedding invite if they want chicken or fish and then being impressed when the same number of chicken and fish dinners are ordered on the big day.

It’s really a very different creature than sabermetrics, aka nerdly analysis of baseball stats, where you’re doing a more straight-forward regressive analysis, trying to comb through figures to find ones that might be predictive of future performance.  “His baseball predictions weren’t as spot-on as his election projections,” notes Taylor.  Right!  Because making a prediction of what people are going to do based on what they said they were going to do and making one based purely on past performance are two entirely different creatures.  To paraphrase Pulp Fiction, “it ain’t the same ballpark, it ain’t the same league, it ain’t even the same sport.”

People who don’t necessarily understand numbers tend to think that one is pretty much as good as the other.  This is the kind of thinking that leads to getting excited over the number of impressions a banner ad got, or the number of clicks a search term got, even if neither of those leads to revenue.  But with any analysis it’s important to figure out which ones matter and which ones don’t.  Then getting good results might not seem like such a miracle.

Picture by piratetuba

Revenge Of The Nerds: Election Validates The Quants

538 Electoral College Prediction

Wow, has it really been so long since I posted?  I REALLY need to get this hangnail situation under control.  Seriously, I need some body butter or some really rich cuticle cream.  Something with olive oil.

So, one of the big stories in the aftermath of the recent presidential election, besides the minor sideshow of someone was re-elected, was the utter triumph of Cold Hard Numbers over Gut FeelingsNate Silver became a veritable It Boy as his blog FiveThirtyEight correctly predicted the totals of the electoral college, including the always-kooky Florida (official state breakfast: the waffle).  Other pundits who did a good job of predicting the outcome included Sam Wang of Princeton Election Consortium and Markos Moulitsas of The Daily Kos.  And how did they manage this feat? Through the wild method of looking at the polling data.  I know, that’s crazy, right?  Ask people what they’re going to do and then listen when they tell you.

Now obviously there was a little more to it than that—there was a wealth of polling data available at both the national and state level, and each of these pundits created their own proprietary models that included how reliable they felt the polls were, adjustments for likely turnout, etc.  But in general, the models amounted to looking at the data and seeing where it lead.  Contrast this with methodology of former Reagan speechwriter Peggy Noonan:

All the vibrations are right… there’s the thing about the yard signs. In Florida a few weeks ago I saw Romney signs, not Obama ones. From Ohio I hear the same. From tony Northwest Washington, D.C., I hear the same.

Is it possible this whole thing is playing out before our eyes and we’re not really noticing because we’re too busy looking at data on paper instead of what’s in front of us?

Turns out this was not possible.

Now, we in California love our vibrations, but in this case the data on paper was the result of calling people and asking them which of a few well-known choices they might make in a week or two.  As far as surveys go, it’s pretty damn predictive.  And yet a broad swathe of people who didn’t like what the survey results were telling them did whatever they could to instead find an answer they liked better.  Dean Chambers created a whole site, UnSkewed Polls, dedicated to the idea that all the polls oversampled Democrats and thus were unrepresentative.  He suggested that Nate Silver wasn’t to be trusted because he is “a thin and effeminate man with a soft-sounding voice.”  Peggy made her prediction based on vibrations and yard-sign anecdotes.  Newt Gingrich was probably just fluffing up the person signing his paychecks.

I get where this comes from.  I love numbers, and one of the hardest things to do is to listen to the numbers when they are telling you something you don’t want to hear.  This is probably even harder at startups, because the idea of starting a new company is fairly irrational to begin with.  So to jump from the mindset of irrationally hoping, working, and striving to rationally assessing a dashboard of numbers, that can be tough.

But if you don’t?  Well, you can find yourself putting a lot of time, money and effort into heading the wrong direction.

Creating a Week Label in Excel

I just love Excel.  I remember when I was introduced to pivot tables—it was like the clouds of data ignorance had parted to reveal golden-robed angels singing of cross-tabbed insights.  And I wasn’t even drunk!  Except on the beauty of Excel, of course.

Just recently I was looking at some daily performance data, and I decided I wanted to group it by week.  To start I just applied the WEEKNUM formula:


which renders “6/22/11”, for example, into “26”.

This worked fine for grouping, but “26” doesn’t really mean anything intuitively to me.  When I wanted to figure out why the numbers jumped in a particular week, I had to go back to the daily sheet, find the week number in question, and see what the associated dates were—then I could say, “Oh, yes, 9/2, that’s when we had a piece run in the Wall Street Journal.”

So instead I decided to create a label that would display the first and last days of the week in question.  After a bit of tinkering I hit on using the WEEKDAY function, which delivers a number for the day of the week a date falls on.  I figured I could use that number to calculate how far from the beginning or end of the week the date is, then subtract or add the right number of days to deliver the two numbers. I used TEXT to make them labels:


This yielded an output like this:

Date Week
6/22 06/19-06/25
6/23 06/19-06/25
6/24 06/19-06/25
6/25 06/19-06/25
6/26 06/26-07/02
6/27 06/26-07/02
6/28 06/26-07/02
6/29 06/26-07/02
6/30 06/26-07/02
7/1 06/26-07/02
7/2 06/26-07/02
7/3 07/03-07/09

I formatted the labels to include the “0” in front of single-digit numbers because otherwise they don’t sort properly—October (10) comes before February (2).

Did I mention that I did all this on a Saturday morning?  And in fact considered it to be a bit of a treat to myself for doing some other work?  Yep, I party hard.