Tales from the Activity Log: How to Measure and Optimize Database Health

Editor's Notes

• #2: Hi, I'm Jeff, here's what I do, here's what Nintex does.
• #3: We as marketers wanted to improve what we were doing in the email marketing and automation realm. We had some big questions, and I started to realize, they all centered around a common thread - monitoring trends in activity in our database.
• #4: Database growth - Are we growing or shrinking? Its a leading indicator to downstream performanance and its also a metric we must monitor with all the database-size metered applications we use today.
• #8: The problem – basic reports in Marketo and SFDC can’t get us there
• #9: The activity log has the potential to show the underlying trends, but the problem, is you can’t view it from a traffic cam across multiple leads. And that’s where we went to solve the problem.
• #10: Gotchas: Marketo throttles APIs – so need to make sure that your application can pause or have a “catch all” to make sure all Leads are synced at the end. Took almost 9 months to stabilize and bug fix the application
• #12: 235K records in 1 year (includes duplicate Leads) Activity log - 47.3 million records in 1 year
• #14: Measuring true net database growth (And churn) over a period of time By segment, By Region and any combination of the two Correlate churn with activities on your Marketing Calendar We could monitor how the database was truly growing – how to identify our subscription requirements when we’re likely to hit a new level.
• #15: Aim to keep churn rates below 2% - for customers for instance, this means no more than 1 email a week. Meanwhile, our prospects don’t get emailed enough – we’re far below the tolerable threshold – I’d be willing to tolerate even higher rates for them, and we aren’t even close, and this is telling, given how little we talk to them. But this varies by region, too. For example, our customers in APAC are a lot more tolerant to more emails – we’re able to send 1-2 emails a week without seeing a major issue with churn. Our partners, on the other hand, tend to get a lot of emails from us, but this is especially challenging in our americas region, where we’re seeing it having a big effect on churn rates. With this knowledge, we know that the built in communication limits won’t be sufficient for us in Marketo – we have to devise a way to customize how we throttle email sending. Haven’t built it yet, but have an idea for a score field on all records that increments each time an email goes out and resets every 30 days, and then use in smart campaigns as filter criteria (only send to leads who have a score less than 2)
• #16: Once you have this kind of insight, it starts to look silly to even think about communication limits as an adequate solution.I haven’t built this yet, but based on what we’re seeing at a macro level with email trends, we can build a simple scoring field to keep track of emails sent to a particular contact, and then use it as filter criteria for emails.
• #17: Lead scoring is a tricky system to build reliably because while you can set a weight for each value, you’re often flying blind to how frequently each occurs until you launch your program, and by then, your scores may already get pretty skewed. But if you could look at the frequency of each occurrence beforehand, you could get a better sense of how to weight each scoring rule. And that’s exactly what we did.
• #18: For us, we wanted to look at the best converting leads and the behavior they exhibited. What we found was that these leads were more likely to perform relatively “rare” activities, like visit our community, sign up for a knowledgebase account, or fill out what we call a “high value form”. This behavior was much less likely to be found in other inbound leads. So we identified these rare activities, and were able to weight them much higher than we otherwise would have, which can send a clear signal to our sales team.
• #19: Product telemetry data Email engagement by time of day – best time to send by segment What you can do to take first steps: