How to Choose a Value Metric
A four test method for picking a value metric that grows with customer value, with examples from Slack, Snowflake, HubSpot, and Figma, plus failure signals.
On this page 7 sections
- The four tests a value metric has to pass
- Worked examples, including the two that caused backlash
- Generating candidates from your own usage data
- What the ten customer interviews should actually ask
- Migrating when the current metric is wrong
- Pick the forecastable metric
- What to do next
- Frequently asked questions
The short answer
A value metric is the unit you charge for: a seat, a contact, a credit, a resolution, a task. The right one passes four tests. It tracks the value the customer actually realises, it grows as the account matures, it is predictable enough for a buyer to budget, and it is measurable and auditable inside your product. When those conflict, favour predictability, because unpredictable bills cause more churn than high prices do.
Key points before you start
Pricing projects usually start with the wrong question. Teams argue about whether Pro should be $49 or $59 for three weeks, then ship a page that charges for the same unit it always did. The price point is a dial you can turn any quarter. The unit you charge for is the architecture, and it determines whether an account that doubles its value to you also doubles its spend.
That unit is the value metric. Here is a structured way to pick one, with the four tests I use and the worked examples worth studying, including the two that went badly in public.
The four tests a value metric has to pass
A candidate metric has to clear all four. Failing one is usually fatal, and teams tend to notice only after launch.
Test one: does it track realised value? The customer should feel that paying more coincides with getting more. Figma charges for editors, not viewers, because editing is where the work happens. Charging for viewers would have taxed the exact behaviour, sharing designs widely, that made the product spread.
- Test two: does it grow as the account matures? A metric that plateaus caps your expansion revenue permanently. Storage grows. Seats grow, slowly. Transactions grow with the customer’s own business, which is the strongest version of this test: your revenue rises without a salesperson involved.
Test three: can the buyer forecast it? This is the test teams underweight and the one I weight highest. A procurement lead has to put a number in next year’s budget. If the honest answer to “what will this cost us in 2027” is “it depends on your usage”, you have created a renewal risk that no amount of product quality repairs.
Test four: can you measure and audit it? The metric has to be countable in product, visible to the customer in real time, and defensible in a dispute. Anything requiring manual reconciliation will consume your finance team and produce arguments you lose on principle even when you are right.
Where the tests usually conflict
Tests two and three fight each other constantly. The metrics that grow fastest are usually the least predictable. My rule: when they conflict, take the more predictable metric even if it captures slightly less value, then recover the difference through tiering and commitment discounts.
Worked examples, including the two that caused backlash
Six products, six different answers, and the reasoning behind each is more useful than the metric itself.
| Company | Value metric | Why it works | Where it strains |
|---|---|---|---|
| Slack | Active users in the period | Removes the pay for dormant seats objection outright | Revenue dips when customer headcount falls |
| Snowflake | Compute credits | Tracks the work done, separate from storage | Bill spikes from a badly written query |
| Figma | Editor seats | Taxes creation, not sharing, so files spread freely | Editor and viewer boundary gets gamed |
| HubSpot | Marketing contacts | Grows with database size and marketing ambition | Punishes freemium list growth, renewal friction |
| Zapier | Tasks executed | Directly proportional to automation value delivered | Failed and retried runs cause billing disputes |
| Intercom | Resolutions by the AI agent | Charges for an outcome, not an interaction | Defining a resolution is contentious |
Slack’s fair billing policy is the cleanest example of test one done well. Charging only for users who were active in a billing period removed the single biggest objection to seat pricing, which is the customer paying for licences their ex employees still hold. It cost Slack revenue in the short term and bought enormous goodwill in procurement conversations.
Snowflake’s separation of storage from compute is the cleanest example of test two. Storage grows slowly and cheaply. Compute grows with how much value the customer extracts. It also produced the most famous failure mode in usage pricing: an engineer writes an inefficient query, leaves it running, and the bill arrives with three extra zeros. Snowflake now ships resource monitors and spend alerts specifically because test three was the weak leg.
Intercom’s shift to charging per AI resolution is the most interesting recent case, because it is an outcome metric rather than a usage metric. It aligns beautifully with test one. It is fragile on test four: customers and vendor can disagree about whether a conversation was really resolved, and that disagreement lands in a renewal call.
The two that drew public complaints
HubSpot’s marketing contacts model and Zapier’s task counting both generate recurring customer complaints in public forums, for the same underlying reason: the metric grows for reasons the customer does not perceive as value. A list that fills with dead signups, or an automation that retries a failing step, both increase the bill without increasing benefit. Watch for any metric that can grow while the customer’s outcome does not.
Newsletter launch list
The Friday SaaS Marketing Brief
Join the list for the upcoming SaaS Marketing Brief. Get the marketing planning worksheet immediately.
Generating candidates from your own usage data
Do not start from a whiteboard. Start from the product database.
Pull every countable object and event in your product for the last twelve months, per account. Records created, users active, workflows run, integrations connected, API calls, documents generated, gigabytes stored, messages sent, hours tracked. You are looking for a list of maybe fifteen candidates, most of which will be obviously wrong and that is fine.
Then run three tests on each candidate before you score anything.
Correlation with retention: do accounts with high values of this metric renew more? A metric that does not correlate with retention is measuring activity, not value.
Correlation with account size: plot the metric against ACV or seats. You want a positive slope without extreme outliers. If your top account has 400 times the median, that metric will produce a pricing page nobody can read.
Distribution shape: plot the histogram across your base. A long tail with a thin middle means your tier boundaries will be arbitrary and half your customers will sit awkwardly between plans.
The selection process, start to finish
- Extract every countable object and event
Twelve months, per account, from the product database. Aim for ten to twenty candidates. Include the ones you think are silly.
- Plot each candidate against retention and ACV
Kill anything with a flat or negative slope against retention. You are looking for metrics that rise as accounts get healthier.
- Check the distribution shape
Histogram across the base. Reject metrics where the top decile is more than fifty times the median, unless you plan to sell only to enterprises.
- Score the survivors against the four tests
Weight realised value at 30 percent, growth at 25, predictability at 30, measurability at 15. Force yourself to score, not discuss.
- Interview ten customers about the top two
Smallest, median, largest, plus two recent churns. Ask each to forecast next year's number and to explain the metric to you in their words.
- Model revenue under each metric on real data
Reprice your existing base under both candidates. Look at who wins, who loses, and how much revenue moves. Nobody should move more than 40 percent.
- Design the migration before you decide
If you cannot describe a grandfather path and a cap on increases, you have not finished choosing. The migration plan is part of the decision.
The scoring matrix is where discipline enters. Give each of the four tests a weight, score each candidate one to five, and multiply. It is not science, and it forces a team that has been arguing in circles to commit to relative importance. My default weights put predictability equal with realised value, which is deliberately unusual.
If you want the arithmetic on what each candidate does to revenue, the value metric pricing calculator and the Usage Based Pricing Simulator both let you model the base under alternative units before anything reaches a pricing page.
What the ten customer interviews should actually ask
Interviews go wrong when they turn into preference surveys. Nobody will tell you they want to pay more. Ask about mechanics instead.
Three questions do most of the work. First: if we charged by X, what number would you put in next year’s budget, and how confident are you in it? Second: who else in your company would have to approve that line, and what would they ask you? Third: describe how this pricing works to me as if I were your CFO.
The third question is the killer. If a customer who has used your product for two years cannot explain the metric back to you, no prospect will understand it on a pricing page. That alone has killed more candidate metrics in my experience than any scoring matrix.
Include two recently churned customers. They are far more honest, and billing surprises are a common contributing cause of churn that never gets recorded as such because the exit interview records “budget”.
Self-paced learning
Build your SaaS marketing study plan
Choose a free course and work through its published lessons at your own pace. Save the course index for later.
Migrating when the current metric is wrong
Most companies reading this already have a metric and suspect it is wrong. The migration is harder than the selection.
Sequence it over two to four quarters:
- Launch the new metric for new business only. You get clean data on conversion and deal size without touching revenue you already have.
- Publish both models for a quarter. Let existing customers opt in. The ones who benefit will move immediately and become your proof points.
- Migrate cohorts with a cap. No account sees an increase above a stated percentage, typically 15 to 25 percent, in the first year. Write the cap into the communication.
- Grandfather the holdouts indefinitely. The revenue you protect by not forcing the last 10 percent is worth more than the consistency you gain by forcing them.
The cost is real. You will run two billing models simultaneously, your finance reporting gets messy, and sales has to explain two price lists for a year. Budget engineering time for billing changes; it is always more than estimated because proration, mid cycle changes and refunds all have to work under both models.
The failure signal that means you chose wrong
If more than one in five renewal conversations includes an argument about the bill rather than the value, the metric is wrong regardless of what the revenue chart says. Track that ratio. It is the earliest available signal and most companies never measure it.
Pick the forecastable metric
The position, stated plainly: when two candidate metrics are close, take the one customers can forecast, even if it captures a few points less value per account.
The reasoning is about who makes the renewal decision. The person who loves your product is rarely the person who signs. The signer sees a line item that moved unpredictably and asks why. Your champion then has to defend a number they did not control, which is an exhausting position to put a supporter in twice a year.
Predictable metrics also make expansion easier to sell. “Add ten seats” is a conversation a champion can have internally without procurement. “Your compute spend may rise somewhere between 20 and 200 percent” is a conversation that requires a budget cycle.
This is why hybrid models have become the standard answer in usage heavy categories: a committed platform fee that gives the buyer a predictable floor, plus usage above it with alerts and caps. You give up some upside on the smallest accounts and you remove the objection that kills the largest deals. The full argument between the two shapes is in Seat Based vs Usage Based Pricing, and the underlying philosophical split in Value Based vs Cost Plus Pricing.
Two more resources worth having open while you do this work: the definition and edge cases in Value Metric, and the market data on how quickly usage models have spread in Usage Based Pricing Adoption.
What to do next
Run the candidate extraction this week. It is a query, not a project, and seeing fifteen candidates plotted against retention usually settles arguments that have run for months.
Then book the ten interviews before you score anything, because interview answers change the weights you thought you believed in. When you are ready to build the page itself, the Pricing Page Spec Template covers how to present the metric without burying it, the guided version of this exercise is in Lesson 1: Pick Your Value Metric, and the surrounding decisions live under SaaS Pricing Strategy.
Editable CSV worksheet
SaaS Pricing planning worksheet
A practical pricing planning worksheet: decisions, owners, evidence and next actions.
Frequently asked questions
What is a value metric in SaaS pricing?
A value metric is the unit a customer is charged for: seats, contacts, API calls, storage, tasks, credits, resolutions or transactions. It differs from a pricing tier, which bundles features. The value metric determines how revenue scales with each account, which makes it the single decision that most shapes expansion revenue, net revenue retention and how customers feel about their bill.
How do you choose a value metric?
Generate candidates from usage data, then score each against four tests: alignment with realised value, growth as the account matures, predictability for the buyer, and measurability inside your product. Validate the top two with ten customer interviews and a usage distribution plot across your base, checking that the metric does not concentrate revenue in a handful of outlier accounts.
What makes a bad value metric?
One that grows when the customer gets less value, one they cannot forecast, or one you cannot measure reliably. Charging for storage in a product where value comes from collaboration fails the first test. Charging for unpredictable API volume fails the second. Any metric that produces a surprise invoice creates a renewal conversation you will lose even when the product is working.
Should SaaS use seat based or usage based pricing?
Seat based when value scales with the number of people collaborating and headcount is stable and predictable, which describes most workflow tools. Usage based when value scales with volume processed and the buyer can forecast that volume. Hybrid, a platform fee plus usage, is now the most common answer in infrastructure and AI adjacent products because it gives the vendor a floor and the buyer a ceiling.
How many customer interviews do you need to validate a value metric?
Around ten well chosen ones, spread across your smallest, median and largest accounts, plus two recent churned customers. You are not measuring anything statistically. You are checking whether buyers can explain the metric back to you, forecast next year's number, and defend it to their own finance team. If they cannot do all three, keep looking.
Can you change your value metric after launch?
Yes, and it takes two to four quarters done properly. Grandfather existing customers on the old metric, launch the new one for new business, then migrate cohorts with a clear value story and a cap on the increase. Companies that force a same quarter cutover reliably lose accounts that were otherwise perfectly happy.
What is the difference between a value metric and a pricing tier?
The value metric is the unit of consumption, the tier is the feature bundle. A company might sell Starter, Pro and Enterprise tiers that all charge per seat. Tiers segment customers by need; the value metric determines how each account's spend grows over time. Getting tiers wrong costs you some conversion, getting the metric wrong caps your expansion revenue permanently.
The saas-marketing.net editorial team Research and editorial
We research, write and maintain every page on this site. The library explains marketing decisions through practical frameworks, explicit assumptions and references. Corrections can be requested through the contact page.
Published September 11, 2026. Last updated .