What is the right way to compute true gross retention vs net retention when half your customers are on multi-year contracts with annual escalators in 2027?

Two ways to treat the escalator: base-year anchor versus contracted-step anchor

The entire disagreement about mixed-tenure retention math reduces to one question: when a multi-year contract steps from $200,000 to $214,000 because of a 7% annual escalator written into year two, is that $14,000 *expansion* or is it *the same customer paying the same deal*? Both answers are defensible, both are used by real finance teams, and they produce retention numbers for an identical book of business that can differ by 8 to 15 points. You have to pick one, and you have to know exactly what you are giving up.

Option A — the base-year anchor (escalator-as-expansion). Every customer's denominator is their contracted ARR as of the cohort-start date, exactly as billed. When the escalator fires in month thirteen, the incremental dollars land in the numerator and count as expansion, precisely the way a seat upsell would. Under this method, a customer on a three-year deal at $200K with a 7% escalator contributes 107% to net retention in year two and 114.5% in year three, having done nothing but honor a contract they signed before the measurement window opened. Gross retention still shows 100% for that customer, because gross retention is capped at the base and only ever counts losses. The virtue of the base-year anchor is that it is mechanically simple, it ties directly to billed revenue with no adjustments, and it matches what most SaaS metric definitions actually say when read literally: net retention is ending cohort revenue over starting cohort revenue, and the escalator is genuinely in the ending revenue.

Option B — the contracted-step anchor (escalator-as-baseline). Every customer's denominator is their *contractually scheduled* revenue for the comparison period, escalators included, rather than their prior-period actual. The same three-year customer has a denominator of $214,000 in year two — the number the contract obligated them to pay — so honoring the contract produces exactly 100% net retention, and only revenue *above* the contracted step counts as expansion. The virtue here is that the metric measures decisions rather than arithmetic: net retention above 100% means somebody actively bought more, and net retention at 100% means the book performed exactly as contracted. The cost is that you now need a contract-schedule table alongside your billing system, and you have to defend a denominator that does not appear anywhere on an invoice.

Where they diverge most. The gap between the two methods scales directly with escalator size times escalated share of the book. With half your customers on multi-year deals carrying a typical 3–5% CPI-linked escalator, the base-year anchor inflates net retention by roughly 1.5 to 2.5 points versus the contracted-step anchor. Push the escalators to 7–10% — common in deals signed during the 2022–2023 inflation window — and the gap widens to 3.5 to 5 points. If the escalated cohort is 70% of revenue rather than 50%, multiply accordingly. A company reporting 118% net retention under the base-year anchor may be reporting 113% under the contracted-step anchor, and nothing about the business changed.

Gross retention is far less contested, and that is why it matters here. Both methods produce nearly the same gross retention number, because gross retention caps every customer at their baseline and ignores upside entirely. The only place the two methods diverge on gross retention is in how you treat a customer who *should* have escalated but negotiated the escalator away at a mid-term amendment — under the contracted-step anchor that is a measurable contraction against the scheduled $214K; under the base-year anchor it is a flat 100% renewal that shows up nowhere. This is the single most important asymmetry in the whole comparison, and RevOps teams miss it constantly: escalator suppression is real revenue loss that the base-year anchor is structurally blind to.

Choosing between the anchors for your specific book

The decision is not a matter of taste. It is driven by four properties of your contract book, and once you inventory them the answer is usually obvious.

Property one: what fraction of revenue carries a contractual escalator. Below roughly 20% of revenue, do not build a parallel measurement system — use the base-year anchor and footnote the escalator contribution in basis points. Methodological purity on a small cohort creates false precision. Between 20% and 60%, the escalator contribution is material enough that investors will ask about it, so use the base-year anchor as your headline but *disclose the escalator contribution as a separate line in the net retention bridge*. Above 60% — which is where a book with half of customers on multi-year deals often lands once you weight by revenue, since multi-year customers skew larger — the contracted-step anchor becomes the more honest primary metric.

Property two: escalator magnitude and variability. Fixed 3% escalators are boring and predictable; you can model their contribution in a spreadsheet and disclose it. CPI-linked escalators with floors and caps ("the greater of 3% or CPI-U, capped at 8%") are not predictable, and their contribution to net retention will swing with macro conditions you do not control. If a meaningful part of your book carries CPI-linked terms, the contracted-step anchor protects you from reporting a net retention *improvement* in a high-inflation year that reverses when inflation cools — a pattern that reads to a sophisticated investor as a metric you do not understand.

Property three: whether you sell against the escalator at renewal. Some companies treat the escalator as a floor and negotiate upward from it at renewal; others treat it as a ceiling and routinely trade it away to secure a multi-year commitment. If your sales team is *giving escalators back* — waiving year-three uplift to close a renewal, or resetting to base pricing on a re-signature — you need the contracted-step anchor, because that giveback is a real economic concession that the base-year anchor cannot see. Run the diagnostic: pull every multi-year contract that reached an escalator date in the last four quarters and count how many actually billed the escalated amount. If more than about 10% did not, you have an escalator-realization problem and you need the anchor that measures it.

Property four: how you are benchmarked. If your investors and board compare you against pure-subscription SaaS comparables, the base-year anchor is what those comparables are (mostly) using, and deviating from it makes you incomparable without a bridge. If you are compared against enterprise-contract businesses where multi-year escalators are the norm, the contracted-step anchor is closer to peer practice.

The tiebreaker when the four properties conflict. Compute both, permanently. The marginal cost of maintaining a second denominator once you have the contract-schedule table is close to zero, and the *difference* between the two numbers is itself a valuable metric — it tells you exactly how much of your net retention is contractual mechanics versus customer decisions. Report the base-year number as the headline for comparability and the contracted-step number as the internal management metric. What is not acceptable is computing whichever one looks better each quarter.

The numbers behind each option, worked through a real book

Abstraction hides the stakes, so here is a concrete 20-customer book with the structure the question describes — half on multi-year contracts with escalators, half on annual deals — carried through both methods.

The cohort. Ten customers on annual subscriptions with no escalator, averaging $120,000 each, for $1,200,000 of base ARR. Ten customers on three-year contracts averaging $280,000 each, for $2,800,000 of base ARR. Multi-year customers skew larger, which is typical — the escalated cohort is 50% of *customers* but 70% of *revenue*, and that distinction is the first thing that trips teams up. Total cohort baseline: $4,000,000. All ten multi-year deals carry a 5% annual escalator; six of them are fixed 5%, four are "greater of 3% or CPI, capped at 6%" and landed at 5% this cycle.

What actually happened over the measurement year. Among the ten annual customers: seven renewed flat ($840,000 retained), one expanded from $120K to $180K on a seat addition (+$60,000), one contracted from $120K to $90K at renewal (−$30,000), one churned entirely (−$120,000). Among the ten multi-year customers: eight billed their escalated amount without incident, one had its escalator waived in a mid-term amendment in exchange for adding a two-year extension, and one contracted its committed base from $280K to $230K in a restructuring while keeping the escalator on the reduced base.

Under the base-year anchor. Denominator is $4,000,000, the billed base ARR at cohort start. Numerator: annual segment contributes $840,000 + $180,000 + $90,000 + $0 = $1,110,000. Multi-year segment: eight customers at $280K × 1.05 = $294,000 each = $2,352,000; the waived-escalator customer at $280,000 flat; the restructured customer at $230K × 1.05 = $241,500. Multi-year total = $2,873,500. Cohort ending revenue = $3,983,500. Net retention = 99.6%. Gross retention counts only losses: the annual segment lost $30,000 to contraction and $120,000 to churn; the multi-year segment lost $50,000 of committed base on the restructured customer, and *nothing* on the waived escalator because it never fell below base. Retained = $4,000,000 − $200,000 = $3,800,000. Gross retention = 95.0%.

Under the contracted-step anchor. The denominator becomes contractually scheduled revenue for the comparison period: the annual segment stays at $1,200,000 (no escalators to schedule), the multi-year segment becomes $2,800,000 × 1.05 = $2,940,000. Total denominator = $4,140,000. The numerator is unchanged at $3,983,500. Net retention = 96.2% — 3.4 points lower than the base-year anchor, and every one of those points is the escalator being reclassified from expansion to baseline. Gross retention now catches the waived escalator: that customer's scheduled $294,000 came in at $280,000, a $14,000 shortfall that is a genuine, permanent, negotiated revenue loss. Total losses = $30,000 + $120,000 + $50,000 (restructure, now measured against the scheduled $294K as a $64,000 shortfall — the $50K base cut plus the $14K escalator that never applied to the lost slice) + $14,000 waiver. Retained = $4,140,000 − $228,000 = $3,912,000. Gross retention = 94.5%.

Read the spread. Net retention moved 3.4 points on methodology alone. Gross retention barely moved — 0.5 points — but the *composition* changed materially: the contracted-step anchor surfaced $14,000 of escalator suppression that the base-year anchor recorded as a perfectly healthy flat renewal. Scale that to a real book. If 8% of your escalator-bearing contracts get their uplift waived each year, and escalators average 5% on a $2.8M escalated base, you are quietly forfeiting roughly $11,200 annually per cycle in the first year and compounding it forward — because a waived escalator does not just cost this year's uplift, it lowers the base every future escalator compounds against.

The compounding math is the part practitioners underweight. A 5% escalator on a $280,000 contract yields $14,000 in year two, $14,700 in year three, and — if the customer renews for another term at the escalated level — $324,135 becomes the new base for the next cycle. Waiving year two's escalator once does not cost $14,000; over a five-year customer relationship it costs closer to $77,000 in cumulative billings and leaves the base 5% lower forever. That is why escalator realization deserves its own metric rather than being buried in a retention aggregate.

Sanity ranges to check your output against. For a book structured like this, healthy gross retention lands at 88–95% under either anchor — the anchors agree closely on gross by construction. Net retention under the base-year anchor typically runs 3–6 points higher than under the contracted-step anchor when escalators average 5% on 70% of revenue. If your two methods differ by more than about 8 points, either your escalators are unusually large, your escalated revenue share is above 80%, or you have a computation error — most commonly, double-counting the escalator by putting it in both the denominator and the expansion line.

Implementing the calculation and sequencing the rollout

Getting the methodology right on paper is the easy half. The implementation is where mixed-tenure retention math actually fails, almost always for the same three reasons: the contract schedule does not exist as data, mid-term amendments are not versioned, and co-terming silently rewrites baselines.

Step one: build the contract schedule table. This is the prerequisite for everything and the reason most teams default to the base-year anchor — they cannot compute the contracted-step denominator because nobody ever stored the schedule. For every active contract you need: customer ID, contract start, contract end, base ACV, escalator type (none / fixed / index-linked), escalator rate or formula, escalator floor and cap, escalator effective dates, and the scheduled amount for each contract year. Most CPQ and subscription-management systems store the *current* billing amount but not the forward schedule; you will likely be reconstructing it from order forms. Budget real time for this — a few hundred contracts is a multi-week effort, and it is the single highest-leverage RevOps investment in this whole exercise, because the same table feeds forecasting, renewal planning, and escalator-realization tracking.

Step two: freeze the cohort and lock both denominators simultaneously. Pick the cohort-start date, snapshot every active customer, and compute *both* the base-year denominator (billed ARR at start) and the contracted-step denominator (scheduled revenue for the comparison period) in the same pass. Store both, versioned, with the snapshot date. Recomputing a denominator months later from a mutable contract table is how retention numbers become irreproducible — the contract record has changed since, and you will never get the original number back.

Step three: classify every escalator event that fires during the period. Each one resolves to exactly one of four states: realized (billed as scheduled), waived (negotiated away, base unchanged), partially realized (billed at a reduced uplift — a 5% escalator honored at 3%), or deferred (postponed to a later date, still contractually owed). Realized events are baseline under the contracted-step anchor and expansion under the base-year anchor. Waived and partially-realized events are contraction under the contracted-step anchor and invisible under the base-year anchor. Deferred events must not be counted as loss — flag them and resolve at the deferral date, or you will book a contraction that reverses next quarter.

Step four: handle the amendment and co-term cases explicitly. A mid-term expansion on a multi-year contract usually gets co-termed to the master end date and prorated, which means the incremental ACV shows up at an odd annualized figure and often *resets* the escalator schedule for the combined amount. If you do not re-derive the contract schedule after every amendment, your contracted-step denominator drifts from reality within two quarters. Rule: any amendment that touches price, term, or escalator terms triggers a schedule re-derivation, versioned, with the prior version retained so the frozen cohort denominator stays intact.

Step five: compute the two metrics in the correct order and never let them diverge in cohort or denominator. Gross retention first: sum each customer's retained revenue capped at their denominator, divide by cohort denominator. It is bounded at 100% by construction — if your gross retention exceeds 100%, you have a capping bug, almost always an escalator leaking into the gross numerator. Net retention second: same cohort, same denominator, numerator uncapped so expansion runs through. The gap between them is your expansion contribution, and it should decompose cleanly into escalator dollars plus non-escalator expansion dollars. If it does not decompose, you have a data problem.

Step six: publish escalator realization as a standing companion metric. Realized escalator dollars divided by scheduled escalator dollars, tracked quarterly. Healthy books run above 90%. Below 80% means your commercial team is systematically trading escalators for term length, which is a pricing-governance issue disguised as a retention issue — and it is exactly the failure the base-year anchor hides. Pair it with a count of contracts whose escalator dates fall in the next four quarters, so renewal planning knows where the exposure sits.

Step seven: report the full stack, not one number. Every board and investor package shows gross retention, net retention, the escalator contribution in basis points, escalator realization rate, and the escalated share of revenue. Disclosing the escalated share matters because if that mix shifts, net retention moves even with flat performance — a reader has to be able to separate mix shift from execution. Then freeze the methodology. A change to the anchor without restating prior periods is indistinguishable from a performance change, and that is how a finance team loses an audit committee.

Related questions

Should a mid-term price increase be treated the same as a contractual escalator?

No. A contractual escalator was agreed at signature and is scheduled; a mid-term increase is a new commercial decision. Escalators belong in the baseline under the contracted-step anchor. Negotiated mid-term increases are always expansion under both anchors, since nothing obligated the customer to accept them.

How do you handle a customer whose escalator is deferred rather than waived?

Flag it as deferred and exclude it from both loss and expansion until the deferral date resolves. Counting a deferral as contraction books a loss that reverses next period, creating false volatility. Track deferred escalator dollars as a separate disclosed line so the exposure stays visible.

Does gross retention change much between the two methods?

Rarely by more than a point in aggregate, because gross retention caps every customer at baseline. The composition changes materially though: the contracted-step anchor counts waived escalators as contraction, while the base-year anchor records them as flat renewals. Same headline, different diagnosis.

What if multi-year contracts are billed annually in advance versus monthly?

Billing cadence does not change retention math — retention measures contracted or earned revenue, not cash timing. Keep both segments on the same annualized basis. Do let billing cadence inform your cash forecast and deferred-revenue rollforward, which are separate reconciliations from the retention bridge.

How should RevOps report the anchor difference to a board?

Show both numbers side by side with the escalator contribution as an explicit bridge line. The spread itself is the insight: it quantifies how much net retention comes from contractual mechanics versus customer decisions. Boards that see the decomposition stop questioning the headline.

FAQ

Does counting escalators as expansion overstate net retention?

It overstates *decision-driven* expansion, which is what most readers assume net retention measures. Mechanically it is not wrong — the escalated dollars really are in ending cohort revenue. But a reader who sees 118% net retention generally infers customers actively bought more, and if 4 of those points came from contracts signed two years ago escalating on schedule, the inference is off. The fix is disclosure, not suppression: report the number and break out the escalator contribution in basis points.

Can gross retention ever exceed 100% with escalators in the book?

No, and if yours does you have a bug. Gross retention caps each customer's retained revenue at their own baseline by definition, so the numerator can never exceed the denominator. The most common cause is an escalator increase leaking into the gross numerator without the cap being applied per customer. Check the capping logic customer by customer, not at the aggregate level — aggregate capping masks the error when expanders offset contractors.

How do you compare against peers who use a different anchor?

Publish the bridge. State your anchor explicitly, give the escalator contribution, and let a reader adjust to whichever convention their model uses. Sophisticated investors discount an undisclosed methodology because they assume it is the most flattering one available. A company that volunteers "118% under base-year anchor, 114% under contracted-step, 4 points from a 5% escalator on 70% of revenue" is far more credible than one reporting 118% alone.

What escalator rate is typical in multi-year software contracts?

Fixed escalators most commonly sit in the 3–5% range, with CPI-linked structures usually written as "greater of a floor, capped at a ceiling" — a 3% floor and 7–8% cap is a common shape. Rates skewed higher in contracts signed during the 2022–2023 inflation period. Anything above 10% is unusual outside long-term infrastructure or specialized enterprise agreements and tends to attract renewal-time renegotiation.

Should the multi-year and annual segments be reported separately?

Yes, and for the same reason you separate any two populations with different retention physics. Multi-year contracts have renewal events every three years rather than annually, so their gross retention is structurally higher in non-renewal years and lumpier in renewal years. Blending them produces a number that describes neither. Report both segments plus the blend, revenue-weighted, and disclose the mix.

How often should the methodology be revisited?

Set it once and hold it for at least eight quarters. If your book's structure changes materially — escalated revenue share crossing 60%, or a shift toward CPI-linked terms — revisit, but any change requires restating prior periods on the new basis so the series stays comparable. Undisclosed, unrestated methodology changes are the single fastest way for RevOps to lose finance and board credibility.

Sources

• https://www.bvp.com/atlas/state-of-the-cloud
• https://www.saastr.com/
• https://openviewpartners.com/blog/
• https://www.klipfolio.com/resources/kpi-examples
• https://www.investopedia.com/terms/e/escalator-clause.asp
• https://www.bls.gov/cpi/
• https://asc.fasb.org/
• https://www.pwc.com/us/en/services/audit-assurance/accounting-advisory/revenue-recognition.html
• https://a16z.com/16-startup-metrics/
• https://www.sec.gov/edgar/searchedgar/companysearch

Related on PULSE

• How to compute gross revenue retention and net revenue retention for a standard subscription book
• Building a net revenue retention bridge that a board can actually read
• How to measure retention when half your customers are on usage-based pricing
• Escalator realization: tracking whether contracted price increases actually get billed
• Cohort definition rules for mixed contract-tenure SaaS books
• Why gross retention deserves equal billing with net retention in investor reporting