jamiequint’s posterous

Lesson Three - Spend Time on Things That Matter

I tweet random thoughts and articles I like a lot more than I blog. If you like these posts you should follow me on Twitter here.

In my last post I mentioned how there were multiple ways we wasted a lot of time at Snaptalent. The first was indecision, the second was spending time on things that didn't matter. At Snaptalent we spent way too long on things that had absolutely no effect on ultimate success of our business. 

Stupid Things We Did

- Spending time discussing whether the company should comp founders cell phone plans. (This discussion actually included emails to investors. I still can't believe we wasted our investors time with something this menial.)

(download)

- Switching bank accounts three times.

- Having 4 closings on our financing and 18 investors

- Prematurely optimizing our server setup. (we had 5 servers at softlayer pushing 20-50k uniques a month)

- Many more things I don't remember right now.

Lessons Learned

1. Block off time for administration at the front end of your startup after your financing closes. Have one person focus on it and get everything done all at once. Don't do it over weeks and weeks, jump on it and remove the distraction. This includes finding office space, setting up utilities, buying and setting up furniture, setting up health insurance, general liability insurance, payroll, accounting, phones, business cards, company credit cards, etc. 

2. When you get into a back and forth about something with your co-founders there is only one question you need to ask

Is this decision material to the outcome of this business?

If the answer is no make a decision on the spot and don't look back. Even if the answer is yes you probably won't have much more data in the near future so do your best to make a decision on the spot.

Financing Climate and its Impact on Startup Business Models

Dave McClure's post today reminded me of Reid Hoffman's talk at Stanford where he mentioned their failure while building SocialNet was trying to do a subscription model and undercapitalizing while their competitors raised 8-10 times as much money as they did and focused on free offerings and growth.

Reid Hoffman's theory of "your strategy has to reflect what your financeability is" explains at least some of the "lazy, ad-happy, Web-Tards with crappy ROI" that happened when the capital markets got frothy and revenue became a bolt-on afterthought secondary to growth. With VC investments the lowest in 12 years its not surprising that the majority of companies are trying to get to profitability earlier than they might have five or six years ago.

Lesson Two - Indecision Kills

I plan to make these more often than every five months, sorry for the long delay. I tweet random thoughts and articles I like a lot more than I blog. If you like these posts you should follow me on Twitter here.

I was going back through my old Snaptalent emails getting some ideas for my next blog post and one of the things that was immediately apparent was how much time we wasted. There are two main ways this happened. I will address these in separate posts. First, indecision.

After weak results from the Snaptalent Ad Network became obvious in mid-2008 we spent a lot of time discussing our many different options for changing strategies. We debated a talent auction model in May, job board strategy in June, and a Facebook application recruitment advertising network in July. Our analysis here was naive, as we were not customer focused and were not making data driven decisions based on customer needs or feedback. We were just guessing what we thought the market wanted from our own point of view. However, the biggest problem we had was not making bad decisions, it was not making decisions at all. We also had no real framework for evaluating decisions once they were made. This led to more guessing, and more wasted time.

Once it became clear our blog advertising network was not working well we spent three months figuring out what to build next. In the meantime, we did not admit to ourselves that the recruitment advertising network was not working. We kept the network running while we built another product on the side. This was a big waste of time. We had to maintain and incrementally improve a dying product while building another product at the same time.

(download)

We decided to shift focus away from single rec recruitment, which targeted mainly passive candidates, to recruitment branding campaigns that targeted mainly young, college age, active candidates in June. We began engineering the product in early July. However, three months later we were still internally debating the merits of our approach.

"In the end I believe the issue with these lacking clickthrough rates is more a timing issue than anything. Many niche ad networks operate on content level relevance (e.g. Ruby, sports, gay, etc.) while we were able to target at the same level we faced the problem of timing. Standard content level relevance seems to work pretty well because many people on sites in these networks are potentially interested in the product or service at any given time, whereas interest for jobs is not steady on a per-person basis and peaks depending on that person's situation at the time which they are viewing the ad. Instead of having steady clickthroughs and results from a relatively homogenous audience, we have sporadic clickthroughs and applications because even though the general topic matches, timing does not. (i.e. an interested passive candidate ad viewer today may not be one tomorrow) Because of this, we essentially have to charge a higher CPM to the advertiser to make up this difference in interest.

The core problem we have discovered is that (with tech jobs in a small network) the economics to make a purely passive candidate advertising network interesting to companies does not exist. So, how do we fix our economics?

Much of our recent discussion has involved switching to a college network where we focus on selling recruitment marketing campaigns rather than single rec recruitment advertising. I think that there are two key things in this strategy that shift the economics in our favor. First, college students are an active audience. Second, recruitment marketing shifts focus away from filling recs and towards secondary benefits like branding. While we have not fully explored the economics of this market yet with our facebook trial and other projects it seems safe to assume that these two key differences may swing the economics in our favor."

Jamie Quint - email to team@snaptalent.com - September 8, 2008

Looking back, our decision to focus on the college network and recruitment advertising campaigns rather than single rec recruitment advertising never should have been argued for 5 months. We should have made our decision, launched the product, and measured external results. 

Argue only hard data. Make decisions quickly. Evaluate based on external feedback not internal pontification.

We should have not have gone into another product line while trying to maintain our existing product. We should have either focused all our effort on our existing product, or admitted to ourselves that it did not work and shut it down immediately rather than letting it distract us.

Realize a sunk cost when you see one.

In the end, this product never launched. We got distracted and shifted focus to a different college strategy before this product ever made it beyond a single-company pilot. 

Indecision kills.

 

Resources:

Mark Suster - JFDI (http://www.bothsidesofthetable.com/2009/11/19/what-makes-an-entrepreneur-four-lettersjfdi/)

Fred Wilson - Action Oriented (http://www.avc.com/a_vc/2009/12/action-oriented.html)

Ben Yoskovitz - Indecision Kills Startups (http://www.instigatorblog.com/indecision-kills-startups/2009/12/16/)

Lesson One - Know Your Market

**Genesis of the Idea

Before starting Snaptalent in October 2007 the founders and I had spent the summer working on an idea for contact management software that was very similar to 37 Signals' product Highrise. We stopped working on this soon after Highrise launched, but kept exploring ideas around what people find such software useful. This led into a lot of discussion about recruiters and the recruitment market. From our analysis we determined that recruitment was a large market with very little technological innovation, ripe for disruption. We thought it was big problem was that there were a lot of places to post jobs on the internet, but knowing the right place to post was difficult. We decided that we could add value by helping people find the best place to post their jobs. We wanted to make a site where people could upload a job posting and would be told where to post it based on past performance of given sites in filling similar jobs. The tricky part about this was actually figuring out what job posts were successful and which ones were not, which is why we decided to build our own network. We needed to have more complete data in order to measure and optimize the distribution of postings. This is what became the first version of Snaptalent. (Techcrunch Review - http://www.techcrunch.com/2008/03/12/snaptalent-targets-job-candidates-where-they-work-and-spend-time-online/)

**Post-Launch

We were able to get a fairly good number of blogs signed up initially to cover programming verticals, which is the market we decided to tackle first. However, as a few months passed the novelty of a different type of display ads wore off and our click through rates started to drop off. We initially started at around a $1.50 eCPM and fell off to well below $1.00 after a few months (calculated based on a 75/25 split of the revenue if my memory serves me right) as our CTRs dropped dramatically. We also experienced the inability to convert the clicks we did get into actual candidates. We were charging $500 for 1000 clicks ($0.50 CPC) and delivering on average only about 2 candidates per position, a completely unsatisfactory number amongst an audience of recruiters who often measure the success of a channel by the bulk number of candidates it generates per dollar spent. We had a funnel with a less than 1% conversion rate from view to click and again from click to candidate. Even when we did acquire candidates we still got a resume only 50% of the time and often emails to the submitted email address would go unanswered. After 3-4 months our numbers were at the point where we would have almost needed an order of magnitude improvement on total conversions in order to keep our current customers happy and to further grow the ad network the way we had planned. There were certainly some notable successes, both Justin.tv and Weebly found hires through Snaptalent ads, however these were unfortunately only merely happy exceptions to the bleak economic picture that the aggregate numbers painted.

**Mistakes

Before starting Snaptalent we completely failed to evaluate what was going on in our market. We thought there was a problem that could be made better with technology and decided to tackle it without talking to *any* potential customers about if what we were doing was even useful or necessary. We did not have a deep understanding of what recruiters did on a day to day basis, none of us had ever recruited or even really talked to anyone who had. We decided to attack a market because it appeared to be inefficient, without gaining internal understanding of what was actually happening in the market and possible reasons for that inefficiency. We didn't understand what recruiters based their decisions on when making vendor purchasing decisions, we didn't even know who the decision makers and budget holders were within the organizations we were targeting. We were completely blind from a market perspective.

This problem was not only limited to the recruiting side of the market. We also had very little understanding of online advertising. For example, at the time we raised money, we didn't even know it was possible to purchase inventory from bigger networks like advertising.com to run our own ads. Our misunderstanding of online ads also led us to very largely overestimate certain parts of our business model, our clickthrough and conversion rate assumptions were incredibly high leading to overestimation of number of candidates we would generate and underestimate of what our payouts to publishers would have to be in order to maintain acceptable eCPMs.

One of the few firms we pitched to who passed on the Snaptalent deal was First Round Capital (I now have a lot of respect for them because of this). They basically predicted how the first version of Snaptalent would go down in flames, but at that point we were not ready to listen. The following email and attached spreadsheet (sent to our lead investor then forwarded to us) are pretty telling:

"I think we are going to pass on SnapTalent. I like the guy and the concept, but the model we put together just doesn’t play out the way we would like it to.  I’ve attached it here, and you can see that if we assume $5k cost per hire, we get to a very reasonable $2.50 eCPM. But I don’t believe $5K is the right number. I believe it is somewhere between $75 (what Craigslist charges) and, say, $2K.  At the high end, that leads to an eCPM of $1 which is not going to get over the bar of most publishers.  As the low end, the eCPM and the CPC converge....way bad ;) I’m sure we screwed something up and you are going to make a ton on this deal, and I thank you for giving us time to look at it, but it is not going to work for us right now."

(download)

 

**Lessons

You can't just apply technology to markets and expect to succeed. The technology must solve a problem that your target market is looking to have solved. When you understand technology it is easy to see how there are many incremental improvements that can be made with it. Lots of people don't care about fancy new technology and incremental improvements unless it solves a burning problem they already know they have. Do not delude yourself and tell yourself that this does not apply to you. Get out and talk to potential users, see if they will pay money for what you are making, see if it actually matters to them, see if they know the difference between what you are offering and what your competitors offer. Find out why the people you are trying to sell to buy things, figure out what the important metrics are to them. You absolutely *must* have a deep understanding of your market, competitors, and your customers' perception of where you fit in that landscape. Web startups are almost purely market risk, mitigate that as much as possible with customer understanding. Iterate as necessary.

**Related Reading

 

Snaptalent Lessons Learned

There are many prominent stories of startup successes, I know because I've read many of them. While reading the war stories of the hard times that successful entrepreneurs have gone through to reach their end goals is interesting, I always have wondered if there aren't more lessons in failure than in success.
 
Since detailed stories about failure seem to be few and far between I wanted to start a series of in-depth posts chronicling what we went through at Snaptalent. Hopefully this blog will help some people avoid making the same mistakes that we made.