Search Before You Execute

The Core Insight

Webvan raised more than 800 million dollars and burned it in three years going after the 450 billion dollar grocery business.

It spent 18 million dollars on proprietary software and 40 million dollars on its first distribution center before it shipped a single item. Six months after launch it took about 2,000 orders a day against the 8,000 its plan needed for profit.

Steve Blank uses Webvan to indict the model that produced it. Product development runs a schedule from concept to first customer ship, and the board measures the company against that date. First customer ship is the day engineering thinks the product is done. It carries no information about whether a customer wants it.

Most founders believe the hard part is building the product, so they run the plan and staff sales before launch. Blank argues the risk sits on the customer side, and the launch schedule is what hides it. His count is that more than nine of ten new products fail, funded or not.

The Four Steps to the Epiphany, published in 2003, carries the argument, the four steps and the market-type doctrine. The Startup Owner's Manual, written with Bob Dorf in 2012, turns the same method into a workbook. It adds a business model canvas, 45 checklists and a separate track for web and mobile.

Eric Ries sat in the first class Blank taught at Berkeley and became the first practitioner at IMVU. Osterwalder's canvas later supplied the front end. Blank wrote the discipline down, and the lean movement packaged it.

The Framework

Customer development runs beside product development in four steps. Each step is a circle with arrows pointing backward, and going backward is the normal case.

• Customer discovery turns the founders' guesses about the customer and the problem into facts gathered outside the building.
• Customer validation test sells to prove the sales process repeats, before anyone hires a sales team.
• Customer creation builds end-user demand and drives it into the channel, in a shape dictated by market type.
• Company building replaces the learning team with departments that execute a model somebody already proved.

The seam runs between steps two and three. The first two search for a business model. The last two execute one. A startup is a temporary organization searching for a scalable, repeatable, profitable business model, and a company runs a model it already has.

Spending stays low until the seam. Fred Wilson's bands, printed in the manual, put discovery at 50,000 to 75,000 dollars a month, validation under 100,000, and creation held near 250,000. The board funds two to three pivots before the seam, so a pivot is a budget line rather than an emergency.

Key Ideas

First Customer Ship Is an Engineering Date

The product development model came out of manufacturing, reached consumer packaged goods in the 1950s, and arrived in technology last. It fits one case, which is a new product entering an established market with known customers and known competitors.

Blank lists ten flaws in the first book and nine deadly sins in the second, and they collapse into one. Every milestone before launch belongs to engineering. Sales and marketing work to a plan written a year earlier, with no checkpoint to stop and fix it.

The collapse has a printed order. Sales misses the number and the board raises an eyebrow. The VP of Sales exhorts the team, the presentation changes weekly, and then the VP of Sales is fired. The replacement calls the positioning wrong, so the VP of Marketing goes next. The half-life of a startup VP of Sales after first customer ship is about nine months. More than half of startups with first-time CEOs replace the founder.

Market Type Sets the Budget

Blank sorts every startup into four market types. A new product enters an existing market, creates a new market, resegments an existing market on low cost, or resegments it into a niche. The manual adds a fifth, cloning a model proven abroad, which it limits to countries above 100 million people.

The type decides the launch, the positioning, the year-one goal, the revenue curve and the cash requirement. In a new market the year-one objective has nothing to do with market share. Blank's figure, from hundreds of high-tech startups over twenty years: a new market produces no profit until three to seven years after launch.

The competitive read comes from military operations research and takes an afternoon. One company at 74 percent is an effective monopoly, and a head-on attack fails. A leader and a number two above 74 percent combined, with the leader inside 1.7 times the second, form a duopoly. A leader at 41 percent and at least 1.7 times the next largest is very hard to enter and ripe for resegmentation. A biggest player at 26 percent means the market is unstable, and below 26 percent nobody has real influence.

The cost of entry is priced the same way. Attacking one dominant player takes three times its combined sales and marketing budget. In a crowded market it takes 1.7 times the budget of the company you attack.

Earlyvangelists Are Defined by Budget

The only people who buy an unfinished, undelivered product are the ones already in pain. Blank gives them five characteristics, and the last one does the work.

• They have a problem or a need.
• They understand that they have it.
• They are actively searching for a solution, on a timetable.
• The pain pushed them to cobble together an interim solution of their own.
• They hold budget dollars, or they can get them quickly.

Willingness to pay is the test. A believer without a budget is a conversation. The hit rates are printed. About one prospect in twenty engages at all, which means 95 percent rejection. Of the ones who engage, one in three or fewer signs a purchase order.

Sell at or near list price. Anyone can give a product away, and a giveaway measures nothing about buying intent. The named error is discounting an early version to a blue-chip logo.

Every Hypothesis Gets a Pass or Fail Number

For a physical product the loop runs on people. Start with a list of 50 prospects. Each founder holds at least ten conversations a day and books three visits a day. Every 50 calls yields five to ten visits, and reaching 50 people face to face usually takes contacting 200 or more.

The test is binary and its number is set before it runs. Take the hypothesis that three of ten sales calls move to active consideration. Run the same pitch at 30 prospects, and pass at nine or more orders or letters of intent.

For web and mobile the loop runs on traffic. A hypothesis of five million reachable customers for 1,000,000 dollars of AdWords implies 20 cents a click. The test puts up three landing pages at 500 dollars each, runs them on alternate days, and passes at 2,500 clicks a page.

One stop rule covers both tracks. When four of the first ten customers try to take it out of your hands, the test is over.

Get, Keep and Grow Are Three Machines

Customer relationships split into three jobs with separate arithmetic. Get brings people in, keep stops them leaving, and grow raises what each one spends.

The physical funnel runs awareness, interest, consideration, purchase. The web funnel collapses to acquisition and activation, plus a viral loop. Activation is the choke point, and a customer counts as activated once you hold permission to contact them again.

The manual works one example end to end for a 39.95 dollar download. Two hundred lookers at 5 dollars a click cost 1,000 dollars. Forty of them activate, which prices activation at 25 dollars, and none of them has bought anything. Half the activated group buys, so twenty sales at about 40 dollars return 800 dollars. Paying 25 dollars to collect 40 is 63 percent of the sale, and the authors grade it a loser.

Grow is what rescues the arithmetic. Ten valued customers taking 2.5 packages at 100 dollars each add 1,000 dollars, which puts 1,800 dollars against the same 1,000 dollars of spend. The book calls that a modest victory. Keeping a customer costs five to ten times less than getting one, by their estimate, and you read retention by cohort rather than by average.

Pivot or Proceed Is a Board Meeting

Both books gate the spending with the same ritual, run twice: at the end of discovery and at the end of validation. The founding team locks itself in a room for a day or two with the canvas, the hypotheses and the diagrams on the walls. A pivot is a substantive change to one of the nine canvas boxes. An iteration is a small one, like moving a price from 99 dollars to 79.

The severity test comes as two scales. Most customers must rate the problem 8, 9 or 10 out of 10. Most must rate the pain at or near 5 out of 5.

The manual then answers how many customers count as enough, by product type. Enterprise software needs three or four earlyvangelists with as many again in the pipeline. A consumer product needs 20 or 30 or more, because fifteen skateboard buyers do not convince a Walmart buyer. A web or mobile app must reach several thousand prospects and activate at least 100 downloads. A network-effect product needs 500 to 1,000 active engaged users, and three to five times that when it is free.

The honest part is what happens next. Validation closes only when a dollar spent on acquisition returns more than a dollar of incremental revenue, and the funnel is predictable. That bar sends 90 percent of startups back into customer development, by the authors' own count.

Practical Applications

Write down your market type before you write a budget. The choice is existing, new, resegmented on cost, resegmented into a niche, or clone. Pick new, and plan for three to seven years to profit and stop measuring yourself on share.

Turn the loudest assumption in your plan into a pass or fail number this week, and write the threshold before the test runs. Without a threshold set first, you will negotiate with the result.

Build the list of 50 and start calling. Ten conversations a day, three visits a day, and a referred introduction instead of a cold one. Founders make these calls themselves, because employees hate delivering bad news and only a founder can call the pivot.

Score every prospect against the five characteristics and keep the ones with a homemade workaround and a budget. Then ask for money at close to list price, because a free pilot measures nothing.

Agree the burn ceiling and the pivot count with your board before discovery starts. Two to three funded pivots is the printed number, and burn rate is the issue boards most often fire founders over.

Who This Is For

Founders selling something a buyer has to be convinced about get the most from the pair. Read The Four Steps for the argument, the market types and the sales roadmap. Read The Startup Owner's Manual when you want the worksheet for Tuesday morning.

Both books show their vintage. The Four Steps landed in 2003 and the manual in 2012, and the tactics date faster than the method. The worked examples lean on enterprise software and hardware, where a 250,000 dollar sale pays for 50 face-to-face interviews. The same manual notes that few smartphone applications sell for more than 10 dollars, where 50 interviews never pay for themselves.

The manual also bloats. Its 45 checklists mostly restate the prose, and two are printed with duplicate text. A patent primer sits inside it with no customer development in it. The web and mobile chapters are reprinted almost whole inside the physical edition.

Skip both if you already have a repeatable sales process and you are hiring against a proven model. These books are written for the search.

The Decision

Two questions settle whether this pair earns your week. Do you know in writing which market type you are in? Do you hold a number that proves your central assumption wrong?

A no to the first means an afternoon with the share thresholds and the cost-of-entry multiples. A no to the second means writing one pass or fail test and running it before Friday.

Blank sets one exit number for validation: three to five purchase orders for a product that does not exist yet. Count how many you hold today. Below that number the job is still search. Search costs a list of 50 names and the calls to work it.