Thoughts on Lagging, Leading, and Predictive Indicators

Dave Kellogg

Foundry CFO Summit

10/27/22

Revision 1.6

​

www.Kellblog.com

https://twitter.com/Kellblog

Who is This Guy?

• Independent consultant, EIR at Balderton Capital, and blogger.
• Former operator
• CEO: MarkLogic, Host Analytics (Planful)
• CMO: Versant, BusinessObjects, Alation (gig)
• GM: Salesforce.com
• Independent board director*
• Alation, Aster Data, CyberGuru, Granular, Nuxeo, Profisee, Scoro, SMA
• Advisor*
• Examples: Bluecore, GainSight, Tableau, MongoDB, Pigment, Recorded Future
• Angel / investor
• Examples: Alation, Cube, Cuein, DataGrail, FloQast, Growblocks, Hex, Saurus, Skyflow

​

2

* List includes both current and former roles

3

While the CEO is on the bridge looking forward,

many finance teams are on the stern, offering in-depth analyses of the ship’s wake

I ❤ SaaS Metrics

• ARR growth
• Net dollar retention (NDR, aka NRR)
• CAC ratio
• Magic number
• CAC payback period
• Churn rate
• LTV/CAC
• Rule of 40 score

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

4

I ❤ SaaS Metrics, But Which Are They?

• ARR growth
• Net dollar retention (NDR, aka NRR)
• CAC ratio
• Magic number
• CAC payback period
• Churn rate
• LTV/CAC
• Rule of 40 score

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

5

Forward-looking or

backward-looking?

Forward-Looking Metrics Have Never Been More Important

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

6

Forward-Looking Metrics Have Never Been More Important

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

7

2022 To-Be-Named Crisis

​

(Except for the other two times)

Why Are Leading Metrics So Important Right Now?

Potential Inflection Point

​

​

​

​

• Cannot extrapolate recent past
• Need to look up and out
• Up the funnel and outside the company

​

Fear-Greed Meter Recalibration

​

​

​

​

• Some VCs flip too fast to red
• Some founders perma-stuck on green
• This is good! We’re counter-cyclical!!

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

8

Only the Paranoid Survive, Grove

The real question isn’t, “what’s happening?” �it’s, “what’s happening to us?”

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

9

What’s a Leading Indicator?

• Let’s take an example
• Annual recurring revenue (ARR)
• Leading?
• Lagging?

​

• IMHO, it’s both
• ARR leads subscription revenue
• ARR lags new ARR bookings (and churn ARR)

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

10

Let’s Try Again

• Stage 1 oppties
• Leading?
• Lagging?

​

• Again, IMHO it’s both
• S1 oppties lead S2 oppties and … closed deals
• Insert s2-to-close rate and sales cycle length
• S1 oppties lag MQLs and leads

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

11

The Marketing Funnel View

• Each layer leads the one below
• And lags the one above

​

• Please note the potentially considerable irony in telling marketing to focus up-funnel
• For 30 years we’ve been moving them down: stop celebrating leads, celebrate MQLs; no, s1 oppties; no, s2 oppties; no, closed deals; no, closed deals that don’t churn, …

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

12

And The Two Funnels Tie Together

13

Marketing

Sales

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

The Worst Day of My Marketing Life

• In the early CRM days, we watched pipeline coverage
• Current-quarter pipeline / current-quarter target
• We knew about the 3x rule
• Pipeline coverage was a pretty good predictor of bookings
• Maybe better than the CRO forecast
• We could use pipeline coverage to manage the business

​

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

14

And Then This Guy Came Along

• We have determined that reps with 3x pipeline coverage generally hit their number

​

• Ergo, I will pummel anyone unless they have 3x pipeline coverage

​

• At all levels in the organization

​

​

​

• And then what happened?

15

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

Everyone Had 3x Coverage!

• And it was mostly ruined as a predictive metric

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

16

“What gets measured, gets managed.” -- Peter Drucker

(Well, it’s fairly clear he didn’t says this. See note.)

​

Who else, did precise or roughly?

See The Dysfunctional Consequences of Performance Measurements – VF Ridgway (1956)

​

“When a measure becomes a target, it ceases to be a good measure.” -- Charles Goodhart (1975)

​

“When you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind.” Lord Kelvin (1883)

​

So I’ll say my take on it

​

“Managing a metric -- e.g., setting OKRs on it, putting into it standard reports -- ruins it as predictor, as a free indicator of business performance.” – Kellogg (2022)

​

17

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

Nota bene: thanks to Ronny Kohavi for calling out some mistakes in the prior version.

So When We Say We Want Leading Indicators

• Is that to predict business performance or to manage it?

​

• Because you don’t really get to choose both

​

​

​

18

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

Demos, Ice Cream, and Drownings

• Deals that get to demo have a 31% chance of closing
• Therefore, we need more demos

​

• Wait, do they close at 31% because we did the demo or because we filtered out so many tire kickers along the way?
• If we reduce the filtering to do more demos won’t that reduce the close rate?

​

• Drowning deaths and ice cream production are strongly correlated
• Does that mean that ice cream production causes drownings?

​

​

19

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

So What Do We Actually Want?

Operational Metrics To Manage and Predict

• Manage -- metrics on which we can assign OKRs to managers
• Predict – what can best tell us where we’re going to land?

​

​

​

​

• Example: it’s a bad idea to tell marketing to go worry about names because they’re leading indicator

​

Analytics To Determine Where To Go

• Focus – on what should we focus time and money?

​

​

​

​

​

• Example: it’s a good idea to ask the data science team which customers close bigger/faster/higher

​

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

20

Operational Metrics, The Classics

• Funnel volume, conversion rates
• Traffic
• MQLs
• MQL > S1
• S1 > S2
• S2 > S4
• S4 > close
• Pipeline progression
• Opportunity histogram
• Triangulation forecasting

​

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

21

Now is the time when you benefit from having done it right all the way along

Note That It’s Actually A Rotten Time To Change How We Do Everything

• Compounds extrapolation risk with
• Invalidation of historical comparison data

​

​

​

• That is, let’s not go change pipeline stage definitions, MQL definition, pipeline scrubbing process, …
• (Let’s do that in fair weather, not foul.)

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

22

Operational Metrics, A Little More Creative

• Activity as predictor of closing
• The non-obvious Gong use-case
• Renewal intent as predictor of churn
• Better than NPS (loosely coupled)
• Post-deployment CSAT
• Product usage
• Win/close rates segmented by industry and use case
• Who buys your product and why is quite possibly changing
• Example: selling CI based on onboarding vs. productivity
• Relationship score as predictor of close
• Build some score that indicates if you’re selling higher and to the business

​

​

​

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

23

Analytics to Guide Us

• Ceteris paribus, we want deals that close
• Bigger – more ARR per unit work
• Faster – more velocity is “like adding a month to your year”
• Higher – win rate, i.e., probability of close
• Ceteris paribus, we want customers who
• Renew – need to at least pass CAC payback period
• Expand – everybody loves NRR

​

• How do we find them?
• This is an analytics / data science problem

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

24

Build Data Models (ICP Re-evaluation)

• Use an internal team if you have one, find an MSDS intern if you don’t
• Techniques: logistic regression, random forests, …
• Goal: model that predicts outputs given inputs
• Inputs are attributes of a customer (as many as you can get)
• Outputs are what “success” means to you – surprisingly elusive
• What is success?
• Renewed? Expanded? Deployed? Landed big? Landed fast? Is a reference?
• Beware two things
• Interpretability problem: these can score customers, but typically can’t say “go look for blue eyes”
• Extrapolation problem: we are at inflection point -- this might produce Web3 as a great target

​

Dave Kellogg, Creative Commons Attribution, Non Commercial, No Derivatives 4.0 International

25

Conclusions

• This is your company
• The CFO is on the right
• Now is the time you get rewarded for having done metrics right
• (If you haven’t, the best time to start is today.)
• We want leading indicators to help us with our instrument landing
• A funnel view is inherently leading when you “look up”
• Beware using managed metrics for prediction
• But we also want to figure out which airport to fly to
• That’s where model-building (ICP re-evaluation) can help

​

​

https://www.getsurrey.co.uk/news/surrey-news/cockpit-view-shows-pilot-land-10373724