My buddy at Redfin, Glenn Kelman, decided he wanted to bare his financial soul so that other entrepreneurs could get greater insight into the witchcraft called financial modeling. In this two-part posting, he reveals his numbers and his lessons. They are eye-opening for most entrepreneurs.
Part I: Numbers
Startups face one primary challenge: To never run out of cash. So when projecting costs, we heeded Guy’s advice that “the three most powerful words you can utter at a board meeting are, ‘We beat projections.’” This convinced us to develop the worst possible financial model that could still be used to raise money.
We’re glad we did. True underachievers, we’ve performed at or just a bit better than this worst-possible plan almost every month, raising revenue projections only when forced to in December 2006. We’ve been able to stick to our plan mostly because absurd assumptions in opposite directions cancelled one another out. As the real estate market tanks, we may not be so lucky in the future.
When first putting together our financial model, we looked online to calibrate spending assumptions. So many people have blown venture capital, we thought, there must be a manual somewhere on how to do it, at what rate, avoiding which follies. We couldn’t find anything. So we took some wild guesses and figured we’d see how they turned out. And now two years later to the day that we built our first model, here are the projections and actual results. Hopefully, you can learn from our experiences.
Rent, Per Employee, Per Month
Redfin Model: $250. Actual Redfin Cost (Last Month): $336
Our actual costs are high because we just moved last month into an office with room to grow, which seems to happen every eighteen months. When people were sitting in hallways at the old space, we were paying about $200 per employee, per month. Class B space on well-traveled mass transit lines is roughly $20 per square foot per year in Seattle, $30 in the Bay Area. You need 165-200 square feet per person or more.
At the extremes, Adobe supposedly allocates 435 square feet per person while Yahoo! allocates 220 square feet per person. The startup cult of cramming people into small spaces is counter-productive: people are what’s really expensive, not space. The cost Redfin really didn’t anticipate was for tenant improvements which you mostly have to fund yourself when signing sub-three-year leases. In September, we spent more than $100,000 to add private offices for our engineers on the hope that our current office will last us longer. It was probably too much money.
Initial Per-Employee Equipment Cost
Redfin Model: $6,500. Actual Redfin Cost: $5,700
Computers, 20” monitors, Ikea desk, decent chair, VOIP telephone, and cell phones for field employees. Our first phone system came from Craigslist, and we had to upgrade after a year.
Monthly Benefits, Per-Employee
Redfin Model: $600. Actual Redfin Cost: $471
Redfin benefits are competitive, but many employees are Seattle-based. Costs are 10% higher in California.
Annual Payroll Tax
Redfin Model: 12.5%. Actual Cost: 8.5%
We added 4% here to our plan, just to pad per-employee costs. In ways you can’t anticipate, people cost money. Payroll taxes are the same nationwide. California’s state payroll tax is, for example, negligible.
Quarterly Bonus Payout, as a % of the Total Possible
Redfin Model: 85%. Actual Cost: ~85%
We pay quarterly bonuses, mostly based on customer satisfaction objectives. Maintaining discipline on bonus payouts has been difficult. When business booms, everyone wants to be paid for it, even if you haven’t yet turned a profit.
Annual Payroll Increase for Existing Employees
Redfin Model: 6%. Actual Cost: ——-
We can’t disclose actual costs here, but they were higher than planned. When we set the plan, many employees were being paid below-market rates, which is not uncommon for startups; as a startup raises more capital and people go into their second year of sucking it up, you have to pay the piper at the employees’ annual review.
Percentage of Candidates for Which Redfin Paid a Recruiting Fee
Redfin Model: 35%. Actual Percentage: 20%
If you can’t build an engineering team through your own network, recruiting fees can become a significant expense at an early stage. Most of the folks Redfin paid a recruiting fee to hire still came through our own employees who got a $2,000 bonus for every recruit they bring on board.
I assumed colleagues would encourage friends to apply to Redfin without a fee, but for $2,000, people start nagging their cousin’s friend’s wife to apply. We saw an immediate increase in candidates. The occasional party, done on the cheap with kegs and pizza, has also worked well for us.
For Employees Recruited for a Fee, the Recruiting Fee as Percentage of Annual Salary
Redfin Model: 20%. Actual Cost: 4.5%
Because we want top-of-the-stack candidates, we do pay 20% to professional headhunters, but most recruits only required a $2,000 employee referral bonus. We’ve also experimented with in-house recruiters working on an hourly wage, but they tend to focus on managing the hiring process rather than adding candidates to the pipeline.
What really wrecks the budget is a retained search for executives ($30,000 – $50,000), a cost we didn’t even include in our calculations above. A retained search as an agreement to work exclusively with one search firm is reasonable, but we recommend negotiating aggressively to defer most payment until placement.
Incremental Amount Paid to Contractors, as Percentage of Payroll
Redfin Model: 5%. Actual Redfin Costs: 3%
Our contractors have mostly been an in-house recruiter, a graphic designer, and a web programmer; no big-shot consultants.
Monthly Travel Costs, Per Field Employee
Redfin Model: $300. Actual Redfin Costs: $369
The $369 includes mileage for field agents who drive clients to listings as well as travel between our San Francisco and Seattle engineering offices. Your costs may be lower. Or not: on the road, some of us still stay with friends.
Monthly Telephone Costs per Field Employee
Redfin Model: $125. Actual Redfin Costs: $261
We’ve started equipping real estate agents with cellular modems, so costs are unusually high here, too.
Monthly Legal Costs
Redfin Model: $12,500. Actual Redfin Costs: $9,406
The $9,406 per month excludes legal costs for a round of financing, usually about $50,000 for company counsel and investors’ counsel (more late stage, less early stage). We have saved money by dividing the monthly work between a more-expensive tech-focused firm (Orrick, very good) for board resolutions, minutes and stock administration and a general corporate practice (Lasher, also quite good) for employment and real estate law.
We also handle on our own most of the repetitive paperwork like option grants and, perhaps unwisely, vendor contracts. We don’t spend much on patents. Incidentally, we pay a service $14,000 per year to set a quarterly price for our stock options. This is a cost that we didn’t anticipate and didn’t project as part of our legal costs.
Annual Accounting Costs
Redfin Model: $45,000. Actual Redfin Costs: $32,912
Once you’ve raised money, your investors will want a year-end audit of financial statements. This can be done for less money when the business is small and if you keep your books in good order, but we commissioned our first audit only after Redfin had generated over $1 million in revenues. Your accounting expenses will also be a bit higher (and your payroll significantly lower) if you can hire a book-keeper to come in twice a month to pay your bills, which makes sense for the first year or two until you need someone permanent.
All-Company Meeting Cost, Per-Meeting, Per-Employee
Redfin Model: $350. Actual Redfin Cost: $560
Almost half of Redfin works outside of Seattle, so our meeting costs are unusually high. But we can’t avoid meeting at least once a year, which is a significant expense that we forgot to plan for in our original model.
There are of course all sorts of other costs that are unique to our business as an online real estate broker: how much we spend to attract a home-buyer to our site, for example, or what it costs to sell a listing. What you see here are just the costs common to every startup.
And now, looking this over, I worry at every turn that we’ve spent too much money. When I first worked at a venture-backed company, someone told me that Sequoia liked to see entrepreneurs “dive in the toilet for nickels.” I’m not even sure that’s true, but it was an image that always stayed with me. At the time it was actually comforting. I couldn’t do anything else right but, thinking about the toilet and the nickels, I said to myself, “This, I can do.”
Part II: Lessons
Here are a few other tips for building a financial model:
Focus on headcount. Outside of marketing programs, the basis for all cost in Internet software is headcount. Just figure out whom you’ll hire and how much you’ll pay and you can’t go far wrong.
Plan slow, run fast. The most likely scenario is that you won’t be able to hire engineers fast enough, and that revenues will come more slowly too. Investors expect their money to drive artificially accelerated growth rates, but signing up for that sometimes just blows a company up before you’ve had a chance to figure everything out. At least in the financial model, give yourself as much time to grow as you can.
Run top-down sanity-checks. To estimate what a company is likely to spend each year, try doubling the average salary and multiplying it by the number of employees. A 100-person company might spend as much as $15 million per year.
Forget economies of scale. The biggest whopper is that a business will magically become more efficient as it grows. If you really believe this, just walk into the headquarters of Amazon or eBay. Bureaucracies grow. Salaries float away. Straining to make a model work, I always forget that per-employee costs rise every year.
Admit that revenues are a mystery. If you don’t have any revenues yet, you can’t say what they’ll be. The point of a model is to prove you can make money if people buy your product, not to insist that they will. By developing different scenarios based on different levels of demand, you can later calibrate hiring and spending according to which scenario fits reality best.
Build from building blocks. Nearly every model is the sum of smaller units. In Redfin’s case, our unit is a market like the San Diego real estate market, which we plan to grow to a certain size in a certain number of months, hopefully returning a certain amount of profit to the overall business. We can then gauge whether the model works by just looking at whether San Diego works, and then asking, “Now what if we had twenty San Diegos?” For another company, it may be a user-created website, with so many page-views and so many ads, or it may be the productivity of a single salesperson, with a million dollars in quota per year.
Take out “hope.” Think about what is most likely to happen, so that a bookie would say you’re as likely to out-perform the plan as under-perform it. Generally speaking, “hope” is not a strategy.
Flag your assumptions. Rather than burying your assumptions in Excel formulae, call them out in a separate tab of the workbook, so that you have a control panel for adjusting the model. This is especially important if you plan to share your model with potential investors.
Hit $100 million in revenues within five years. The premise of most venture investments is the possibility of generating ten-fold returns in five to seven years, which is hard to do if you spend $5 million to build a $25 million company.
Keep market-share under 20%. Most startups reach a jillion in projected revenues by assuming that the business grows by leaps and bounds for five years. Since there’s a natural limit on growth, be ready for the question: “What would your market-share be in year five?” If it’s over 20%, take the jillion-dollar projection down a notch. Even a hit like iPod doesn’t have 20% market-share. You’ll be lucky to come close to 20% of any market.
So these are our costs, and that’s our advice. What’s your take? We’d love to see how our costs compare to other startups'; please leave a comment and let us know where your numbers differ from ours, especially in markets outside the U.S. We could also post a sanitized version of our financial model, if enough people ask for it to make it worth the trouble.
(Thanks to Redfin’s Chris Roske, Chris Neitzert, Matt Goyer and Angela Cough for their help with the numbers in this post.)