Dealer’s Choice: A Deep Dive on Picking the Right NDR Formula for Your Business Model
Net Dollar Retention (NDR) rate is perhaps the most telling metric when it comes to a company’s future revenue durability. It’s a window into how SaaS companies embed future growth into their model. In my humble opinion, it’s the highest signal into how much operating leverage you have.
What’s more, each company’s calculation sheds light on the most important levers within their business: how do you retain and expand customers?
Because NDR is not a standardized GAAP (generally accepted accounting principles) metric, there are subtle changes companies make when it comes to measurement around the edges. Some of these moves are logical, given their customer behavior and selling motion. Others are purely playing ‘hide the ball’.
For Reference: This is the most “vanilla” way to calculate NDR
Let’s dive into how MongoDB, Snowflake, and HubSpot calculate their NDR, deconstructing the formulas, inputs, and nuances that set them apart.
One formula is geared towards multi product, multi channel businesses (MongoDB)
One formula is geared towards usage heavy, seasonal businesses (Snowflake)
One formula is geared towards subscription heavy, seat based models (HubSpot)
And then we’ll discuss how your choice impacts both how you forecast your business, and how fluctuations potentially impact your valuation (you won’t want to miss that part, as it’s not widely written about).
Letttt’s ride…
Three Proven Methodologies (And When to Use Each)
MongoDB's Segmented ARR Expansion: For Multi-Product, Multi-Segment Businesses
How it works:
Tracks ARR changes from existing customers over 12 months
Segments direct sales (90-day usage) vs. self-serve (30-day usage)
Excludes professional services
Choose this approach when:
You have distinct customer segments with different expansion behaviors
Your direct sales motion differs significantly from self-serve growth
You need to isolate which segments drive your larger accounts (and also have higher S&M costs)
Operational implications:
Higher reporting complexity but superior actionability (you'll know exactly which segment to double down on)
More volatile quarter-to-quarter due to usage sensitivity, so budget for 3-5% swings in your forecasts
Earlier warning signals when segments start diverging in health
Trade-off you're accepting: Increased reporting burden for granular insights that directly inform resource allocation decisions. Do you invest more in product? Or direct sales?
Snowflake's Long-Term Consumption Model: For Usage-Heavy, Seasonal Businesses
How it works:
Measures product revenue changes over two full years
This is the longest measurement period of the companies I studied
Focuses purely on product consumption, excluding services
Choose this approach when:
Your customers have long onboarding cycles (6+ months to full adoption)
Usage patterns are seasonal or project-driven
You want to smooth out short-term volatility for cleaner board reporting
Operational implications:
Cleanest story with less quarter-to-quarter noise
Delayed reaction time, as problems show up 6-12 months later than other methods
Best for consumption-based pricing models where usage directly correlates to value
Trade-off you're accepting: Later problem detection in exchange for smoother, more predictable reporting that supports higher valuation multiples.
I’d be remise if I didn’t mention this - because it’s a two year look back, it’s possible for your NDR (less one) to be higher than your annual growth rate. This is an odd phenomenon that’s happened a few times in Snowflake’s history, driven by measurement differences (e.g., NDR of 170% but annual growth of only 60%). And when it happens, it takes some explaining.
HubSpot's Traditional NDR: For Subscription-Heavy, Predictable Models
Choose this approach when:
You have stable, seat-based pricing with clear expansion paths
Your business model emphasizes cross-selling and upselling over usage expansion
You need benchmark comparability with traditional SaaS companies
How it works:
Standard formula: (Starting ARR + Expansion - Contraction - Churn) / Starting ARR
Emphasizes upgrade and cross-sell revenue from existing customers
Typically measured annually for consistency
Operational implications:
Highest benchmark comparability: VCs and public market investors understand this immediately
Most sensitive to expansion team performance: Directly reflects sales execution
Clearest correlation to traditional SaaS unit economics
Trade-off you're accepting: Less granular insights for maximum comparability and investor familiarity.
How Your Choice Impacts Valuation
Usage-based models (MongoDB/Snowflake approach): Usage based models are a cheat code for higher NDR. Because customers take time to ramp, the model is primed for higher rates, as you measure off a smaller base compared to subscription, which starts at a fixed per seat price (and needs to benefit from more hiring). Usage based models typically command 15-20% higher NDRs and revenue multiples when executed well, but investors discount when there’s volatility. Your NDR needs to stay above 120% to maintain premium.
Traditional models (HubSpot approach): Easier to underwrite and predict, supporting more stable valuations but potentially lower peak multiples. Benchmark is 110%+ for strong performance.
Implementation Decision Tree
Start here: What drives most of your expansion revenue?
Customer usage increasing → Snowflake model
Seat expansion within accounts → HubSpot model
Multiple expansion vectors → MongoDB model
Then consider: How much reporting complexity can you handle?
High complexity, high insight → MongoDB
Medium complexity, smooth reporting → Snowflake
Low complexity, high comparability → HubSpot
Finally: What's your board's sophistication level?
Usage-based methodology requires education but tells richer story
Traditional NRR needs no explanation but provides less operational insight
Red Flags to Watch By Methodology
MongoDB approach: When your segments start diverging (>10% gap), it signals fundamental business model stress requiring immediate attention to underperforming segments. It’s understandable that self serve is below what the field team touches, but it can’t fall off the rails completely.
Snowflake approach: If consumption growth stalls, you won't see it for 6-12 months. Monitor leading indicators like seats deployed, data volumes, or compute hours. The feedback loop is long and wide.
HubSpot approach: Declining upgrade rates hit this metric immediately. When expansion revenue drops 20%+ quarter-over-quarter, your growth engine needs immediate intervention. You’ll see a decline in seat based expansion like a lead balloon falling from the sky.
Food for Thought
Your NDR methodology is a long term choice that shapes how your business gets valued, understood, and operated. Choose based on your business model, reporting capabilities, and the story you want to tell, but choose deliberately. The companies getting premium valuations aren't just growing efficiently; they're measuring and communicating that growth in ways that maximize investor confidence and operational clarity. That means sticking to the same formula over multiple years and consistently improving the output.
Wishing you NDR with low downsell,
CJ
