Live from Business of Software 2011 - David Cancel (@dcancel), serial entrepreneur and CEO of Performable (acquired in June by Hubspot), talks about how to measure results and what really matters...

"In God we trust; all others must bring data."

 Former CEO of Performable, acquired by HubSpot in June.

 Always has been "the data guy." At HubSpot it's taken to a whole new level.

 Focus: It's not enough to talk, read, and dream. You have to DO THINGS.

 Favorite hashtag: #JFDI - Just Fucking Do It

Data alone is useless.

Optimize your business for LEARNING not data.

Use your data to validate your assumptions, not to make decisions for you.

Why is it important to create a data-driven team?

Data-driven teams are stronger. A culture of measurement and feedback everyone knows how their work impacts the business, whether support, sales, marketing, etc.

Shared metrics = clear feedback loop.

Always be testing. Just don't be stupid.

ABtests.com - A/B testing alone won't lead you to the answers.

The good news about testing: the average conversion rate is 2%. It's hard to be worse than average if you're doing any testing.

Failure.

Data-driven teams can fail faster.

Rovio failed 51 times before Angry Birds.

Formula 409: It's called that because the first 408 failed.

Harmoix/Guitar Hero: A 10-year overnight success.

iRobot: A 12-year overnight success.

Start Today. Be data-driven now.

The Operating Dashboard.

A single place to measure your business health.

Transparency -- let everyone see it.

UI doesn't matter, it can be an Excel spreadsheet. The data just needs to be good.

Once everyone is fully onboard reviewing the data, you can improve the usability.

Everything you do should improve your operating metrics.

Sales funnels: there are tons of ways to look at them. HubSpot loves funnels.

Draw your funnels, with numbers, and share them.

Daily Sales Performance

The ultimate in accountability.

Large inside sales team, HubSpot shares with everyone who wants to see it performance of every single rep

Churn Performance

Net At-risk MRR

NPS (Net Promoter Surveys)

"How likely are you to recommend to a colleague or friend?"

NPS = % promoters - % detractors

Measure via simple surveys to customers.

Graph NPS score over time.

Cohort Analysis

HubSpot gangs its users by month/year joined and tracks their retention all together.

Without a cohort analysis you are flying blind.

"Are your new customers better or worse? Are you bringing in the right types of customers?"

What really matters?

It's about creating happy customers.

Success = Happy Customers and a Happy Team.

Q&A

Q: You said to not do anything that doesn't directly affect metrics. What about disruptive ideas? How do you include those?

A: The important thing isn't to always succeed, it's to measure what's failing worse or succeeding better than you expected.

Q: What are the teams you look at every day? If you only have one metric you can look at what would it be?

A: Sales and churn, usage metrics, operating metrics.

We've had huge swings in NPS. When that happens we go out and talk to those customers that are in those surveys and interview them to look for patterns. There's always a pattern.

Q: When you send an NPS survey, how many people respond and what biases do you see (e.g. only angry people respond)?

A: Data is always about biases, but we tend to have low multiple hundreds to respondents.

Q: How many responses do you need to consider the NPS valid?

A: Depends on the size of the cohort. We have 5500-6000 customers, and we want over 100 respondents before we feel good about it. Don't know how many we field to get that response rate, but they're surprisingly responsive.

Q: Seen several curves this week that go from 0 to millions. For those of us who are below the 2% average, where do we start with something like AB testing?

A: That's my favorite time to test -- at the early stage when you're still looking for the product/market fit. Would recommend sticking to simple message testing, not detail stuff like button color, etc. Customer has to connect with the message, or not. Also test one variable at a time so you know the differences are meaningful.

Q: Do you advocate giving incentive to respond to surveys, e.g. a gift card?

A: It doesn't work. I'd do it sometimes if it's your customer base, but if it's outside your customer base I wouldn't spend any time -- they're professional survey takers :).

Q: How do you handle customer feedback?

A: Customers can't design software. Being customer-driven doesn't mean doing what customers ask for, it means knowing how they work and fixing that. We take customer requests/feedback to other groups to validate the requests, then start to prototype, etc. Easiest way to validate a feature is to charge for it: "This will cost you $1." "OK, I don't want it anymore."

Q: (from livestream) How many visits do you consider to be significant for AB testing?

A: Totally depends on the type of change you're testing. There are free calculators to help you figure it out.

Q: Feel like I don't have insight on my product for what's working and not. Is it your advice to start recording everything?

A: No. It's not about recording every click, but rather the flows you're trying to solve for. "My customer is trying to do X. How successful are they?"

Got your own measurement story to share? Leave a comment or ask a question!