On Friday, I gave a 20-minute talk at a product design conference organized by Ty Ahmad-Taylor (CEO of FanFeedr) and Hard Candy Shell (the talk shared the same title as this post). I discussed mistakes and lessons from Aprizi’s journey. I don’t think it was videotaped, so I am going to take advantage of this flight to California to write down some of the salient points here (and include thumbnails of the slides). Longtime readers will recognize many of these statements.

First, we assumed that the audience had read Steve Blank and Eric Ries. Second, if you are unfamiliar with the term product-market fit, it was coined by Marc Andreessen, and I think of it as basically meaning that you have the right product for the right market at the right time.

I want to stress that this was focused on PRE product-market fit, when you are still iterating on the problem you want to solve and how you can best solve it.  Here are the six points that I covered:
1. Focus on the Value Proposition
2. Qualitative not Quantitative
3. Focus Groups are Evil
4. Own Customer Development AND UX
5. Let Some Things Suck
6. Beware the Siren Song of Investors

I’m going to skip discussing my background in this post — for that, you can click here.

As for the context of Aprizi (currently in open beta), let me ask you a question: have you ever walked down a cool city street and enjoyed discovering a cool boutique with beautiful, unique products? Well, we want to bring that experience and feeling to the Web. Aprizi gives people a fun and (increasingly) personalized way to discover boutiques, independent brands and emerging designers online. We focus on design and lifestyle-centric products. We’re not a marketplace like Etsy or Boticca, but rather a discovery engine.

When Liz Crawford and I started Aprizi last December, we knew that we wanted to make online shopping both smarter and more fun, but it has been quite a journey from that point to the present. Some of our initial hypotheses held up, and some died under the customer development sword. Our journey consisted of hundreds of interviews across an arc of paper testing, manual alpha (i.e. me behind the website as the hamster on the wheel), a crude first beta, and finally now, a baseline beta product which I am really happy to have as our true starting point.

We made some mistakes and learned lessons along the way, so let’s dive in.

1. Focus on the Value Proposition

This sounds like such an obvious statement but teams can get lost thinking about their own needs rather than what the customer needs. Let me highlight with one of my moments of idiocy. At the start of Aprizi, we got obsessed with email receipts. We thought, “there is item-level purchase data locked up in people’s email inboxes — we can parse it and that would open up so many opportunities!” I was doing lots of customer interviews to figure out which were the best opportunities, which was fine and good, but on the coding side, we started building and testing this email infrastructure.

Thankfully, we woke up after a few weeks and realized that building this infrastructure had *nothing* to do with testing what the customer wanted. It had to do with what *we* wanted, and the data we thought we needed. We put the code on the shelf, got our alpha up to test things that the user might care about, and never looked back. Our business has evolved and we have not touched that code since.

I will give you another example, which I saw in effect with a social games company. The team was so concerned about monetization that they built a virtual goods system before they had even properly tested whether their game was fun, or iterated it to the point that it actually *was* fun. Monetization is important, but “fun” had to be their core foundation. They put their needs ahead of the customer.

Many ask “which hypothesis should I test first?” Definitely think through which assumptions are biggest and most risky, but I think it is always wise to begin with a focus on the value proposition to the customer.

2. Qualitative not Quantitative

I’ve said this before — metrics are really important, but the startup echo chamber can over-emphasize their importance in the early days. The closer you get to product-market fit, the more critical metrics become, but they are not all-important while you are still trying to figure out the right product for the right customer.

If you are trying to solve a gnarly problem, you need to look people in the eye, read their body language, hear their tone, and understand their deeper motivations. You need to watch people using your application, or talk to them right after, not stare at sterile numbers. Beware the “local optima” problem (optimizing for a small problem/market). Yes, install metrics and try to think about which metrics you should really care about, but dig into the human side.

On this line of thought, I think surveys are awesome for objective data, but troublesome for subjective data. Example: at the very start of Aprizi, I did a survey of about 60-70 people and got great factual data about their shopping habits, but also got a huge red herring on desired value proposition. By far, the most popular “solution” the survey takers said they wanted had to do with deals/discounts. However, in a classic case of “what people say is often different from what they do,” when I really dug into their true behavior, very few of the people I had targeted actually oriented their shopping behavior around deals.

Aprizi’s true “eureka” moment did not come from metrics. Instead, I was watching a New York City school teacher try out our alpha. We had given her a bunch of different kinds of recommendations to see what got a reaction. She spotted a tote bag from an an independent artist with a small online store. Her face registered curiosity and she clicked through to the store. Then she lit up and I watched her spend 10 minutes gleefully browsing around this woman’s website. At the end, she turned to me with excitement and said, “I *never* would have found this!” At that moment, all these little things I had been hearing and seeing finally sunk in, and I thought to myself, “THAT is what we need to bottle!”

3. Focus Groups are Evil

To properly express my disdain for focus groups, I asked my 5 year old daughter what the most evil thing in the world was. The answer, of course, was Cruella De Vil. And Cruella is actually an apt metaphor, because that’s one of the problems you can get: a dominant personality taking over. OR you get terrible group think. We tried a focus group early in Aprizi’s journey and it really was not effective, at no fault to the participants.

In a customer development interview, you want to start by talking generally about behavior, then talk about a possible solution to a problem, and throughout you want to keep your antenna really sensitive so you can drill down into the “whys” of people’s responses. You simply cannot do that effectively when you are managing a group of people.

4. Own Customer Development AND UX

Don’t let your company and product design become a game of telephone. You need the person designing the UX to be the same person (or people) talking to customers. You do not have to be a graphic design expert — anyone can use Balsamiq, and anyone can look at 5 comparable websites or mobile apps and decide what works and what does not work.

We had a short-lived experiment. I am an adequate UI designer, but not the best by any stretch of the imagination. At the start of Aprizi, I thought to myself, I know some really good UX/UI folks, and they’re willing to help me out, so why not start with the product being *that much better* with the hand of a pro!

Mistake. While this designer was super helpful in giving us a start in the right direction (for which I am highly grateful), the problems quickly became evident. The designer knew tons about usability, but just wasn’t close enough to the user’s problem. Even more problematic was that my thoughts about the product and marketing were evolving far too quickly to do anything but drive this designer absolutely batty. We needed too many iterations, and I knew I had to take back over the design or else threaten a good friendship.

5. Let Some Things Suck

When you are still hunting for PM-Fit, you have to let some things suck. It is hard for people like us, who take pride in our work, but it is necessary. You really could spend a lifetime optimizing the wrong product and go nowhere.

At this super-early stage, focus on learning. Only do what is “good enough” to learn. The definition of “good enough” varies tremendously depending on the product and customer, so you have to use your own judgement here.

In Aprizi’s first beta, we had a web form that I’ve written about before, so going to skip details here. I came to hate this form with a passion. We learned some useful things through the questions asked to the user, and the metrics showed that conversion was not a huge problem, but when I actually watched people get to this page, I saw them stop dead in confusion. Obviously, you never want that to happen; you want to always give people an action and get them to the sexy sauce as fast as possible. So did we fix or remove this page? No we did not. It killed me, but this form wasn’t getting in the way of learning, which was our true goal. Once we had learned what we wanted to learn from beta-1, and got to work on beta-2, this form was happily taken out back and shot.

6. Beware the Siren Song of Investors

I got a lot out of talking to investors early on in our journey. We weren’t trying to raise money, just looking for advice. You can learn about competitors, business model ideas, and the many ways similar things have failed before (VCs never have a shortage in this department). However, for all the well-intentioned, intelligent advice you will get, you need to be careful around the product.

Investors know what is hot right now. That is part of their job. Digging deep into a gnarly customer problem — that is your job.

When we talked to investors, we kept on hearing “you should focus on deals!” Why? Because Gilt, Groupon and Woot were out there killing it. Deals were the rage. However, in startups, as hockey players say, you want to: “skate to where the puck will be, not where it is.”

So my caution is don’t act like a cat chasing a lure perpetually out of reach. Sometimes the new hotness can reveal opportunities — take Yipit, for example — but you need to really think about the problems and customers *you* care about.

The reality is that investors don’t want you to build what they say you should build; they just want you to build something successful.

- – -
So there you go — roughly my six points for the talk.  The folks in the room then asked a whole bunch of great questions, and you should not hesitate to do so either, whether in a direct email to me (giff.constable at gmail) or in the comments below.

Update: you can still download the model if you want and see if there are any pieces you can use, but I am working on a new version that is a little more cash-oriented. Right now, it isn’t explicitly modeling out annual subscriptions but rather averaging a monthly revenue level, and I think that should be fixed.

Mark Suster of GRP just wrote a post on the importance of financial models, and in an effort to be helpful to new entrepreneurs finding their way around Excel, I thought I would post a business model Excel template for anyone to download and customize (bottom of post).  I completely agree with Suster.  The one time I told myself “I don’t need a financial model yet because it will all be made up bullshit,” I completely embarrassed myself a few months later and swore never ever to make that mistake again (seriously, it still hurts years later).

Why must you create a financial model? Beyond the obvious of needing to speak your investor’s language, here are a few reasons:

• it helps you better understand your assumptions and goals
• it helps you better understand your business’ pressure points
• it forces you to  better understand your scaling cost structures
• it forces you to examine the realism of milestones and cash needs
• it gives you a working document to track and structure/focus for your business analytics (like a product, it should be a continually iterative process and tool)
• it gets everyone on the same page as to what the business is shooting for and sets expectations as to what needs to be accomplished and how long it might take. EVEN IF IT IS GUESSWORK, you still need to make sure your team and investors are on the same page when it comes to assumptions, goals and timing!
• hopefully this one isn’t necessary, but it also creates a document everyone can sign off on, which can reduce finger-pointing later if memories become revisionist should things turn sour

Focus your efforts on a model that feels ambitious but doable, but don’t forget to create a conservative “oh shit” version, and if it has to be a sales document, you may need a “$best case ka-ching$” version as well.

I find that with every startup/product I model out, a lot of customization goes into the user acquisition/growth model, the revenue model, and to a slightly lesser extent, the infrastructure/hosting scaling model.  I like building user growth from the bottom up, focusing on active users (someone who uses your product at least once a month).  I don’t much like simplistic “viral multiples” or top-down approaches like market share penetration.

This particular model came out of an exercise for a freemium Web application startup that wanted to monetize through premium subscriptions.  The model takes a “cohort” approach (more information on cohort analysis over at Fred Wilson’s blog) and applies churn and conversion assumptions to each month’s new users over time.  It also allows you to model out the growth of your team by role and employee type (full time vs contractor).  I have zeroed out most of the assumptions and the staffing plan so that you can start with a clean slate.

Every business is different and you’ll no doubt end up ripping some stuff out and adding some stuff in, but I hope the attached model is a useful starting point for some of you out there.

Here are just a few of the things you’ll need to research/guess if you are just starting out:

• comparables on user growth rates, virality, and churn
• customer acquisition / sales methods and costs
• price points and conversion rates for customers, given your business model and sector
• payment processing rates and fraud/chargeback levels
• contractor and salary rates (both “startup discounted” and market) for your hires
• customer support method and costs , and how it scales with customers
• true travel costs (often underestimated)
• expected concurrency rates and visit frequency
• download size per visit
• architecture scaling: owned vs managed hosting vs cloud and how servers and costs scale

Here are a few key things you want to examine out of your model:

• are there any variables where a slight variation makes a massive difference?
• are your fundraising milestones well matched to your product and financial milestones?
• when do you hit monthly breakeven?
• how many users do you need to have acquired / kept to get to breakeven?
• how much money have you spent in total to get to breakeven?
• are you adequately scaling costs to meet the demands of a large user base in later years?
• can you believe any of it!  You can make a model spit out whatever numbers you want, so cross-check, second-guess, and don’t buy too deep into your own assumptions and guesses!

I’ve tried to pepper the model with notes to make it a little easier to understand.  Please let me know your comments on the model, whether they be improvements, corrections, or just a note on if it helped.

Freemium model v1-2.xls

While the sales of new games in retail have struggled recently, Gamasutra recently reported that the digital download services continue to see strong growth:

• Valve’s Steam service revealed a 97% increase in download sales, year-over-year
• IGN’s own direct download service, Direct2Drive, has also surged year on year, according to the company, posting gains of 56% compared to twelve months previously.
• paid downloads on Xbox Live increased 73 percent over the 12 months through June, as a total of 20 million users are now on the Xbox 360′s online service.

It is good to see areas of resilience in gaming/online entertainment.  There are headwinds, but more of a slowdown than a plummet.

Of course, on this topic, I spoke to one small PC game developer last night who has found online to be an extremely poor distribution channel, but has been thrilled lately with their sales performance at Walmart, Target, etc. So it is clearly hard to extrapolate micro situations from the macro data.