What's the difference between expansion ARR and net new ARR for forecasting in 2027?

The two lines side by side, and why blending them destroys the forecast

Start with the cohort test, because everything downstream depends on it: was this customer in your base on the first day of the period? If no, every dollar they sign is net new ARR. If yes, every incremental dollar above their starting run-rate is expansion ARR. That single question resolves 90% of categorization disputes, and it is the only rule that survives contact with an auditor.

The reason the distinction matters is that the two lines run on different physics. Net new ARR is a funnel output. It is governed by how much late-stage pipeline exists, what percentage of it converts, how ramped the reps working it are, and how long procurement takes. You forecast it with coverage ratios and close rates, and the inputs live in your CRM. Expansion ARR is a relationship-and-product output. It is governed by how deeply the customer has adopted what they already bought, whether the executive who signed is still in the seat, how close the renewal date is, and whether support has been quietly poisoning the account for two quarters. You forecast it with adoption telemetry, health scores, and the renewal calendar, and the inputs live in a product-analytics tool and a customer-success platform that your CRM does not read.

Blend them and you get a number that cannot be diagnosed. Suppose you forecast $4M of gross new ARR for the quarter and land $3.2M. A blended line tells you that you missed by 20%. It does not tell you whether pipeline creation collapsed two quarters ago, whether a marquee account's champion left and took a $600K cross-sell with them, or whether both lines were fine and contraction was under-reported. Each of those has a different owner, a different fix, and a different time-to-recover. The split is not accounting hygiene — it is the diagnostic instrument.

There is also a third bucket that must never be silently netted into either: contraction and churn. Churn is a full logo loss, ARR to zero. Contraction is the customer staying but shrinking — fewer seats, a tier downgrade, a lower usage commit. Teams routinely under-report contraction because a CSM will log a cancellation but soften a downsell, and that quietly flatters gross retention. Both belong on the waterfall as their own negative lines.

Before any of this works, the org needs one agreed definition of ARR itself. ARR is the annualized run-rate of contracted recurring subscription revenue at a point in time — end-of-period, not an average. It is not GAAP revenue, not bookings, not total contract value, not billings. A $120K annual subscription signed on the last day of the quarter is $120K of end-of-period ARR even though ASC 606 has recognized almost nothing. That timing gap is the single most common reason finance and RevOps show different growth numbers on adjacent slides. Committed platform fees and contracted usage commits count; one-time implementation, professional services, hardware passthrough, and uncommitted overage do not. Write it down once, get every analyst to sign it, and stop re-litigating it in board prep.

Choosing which model to run, and when the split earns its cost

The honest answer is that split-stream forecasting is the right default at roughly $10-20M ARR and above, and largely overhead below it. Below $10M, the customer count is too small for either line to be statistically stable — three deals swing the mix twenty points quarter over quarter, and a formal two-model stack produces false precision. Track the split in the warehouse from day one so the history exists, but forecast combined until the sample size justifies the machinery.

Above that threshold, the decision tree is straightforward. Ask whether the business has enough installed base for expansion to be material — if expansion is under 15% of gross new ARR, the new-logo model carries the forecast and expansion can ride as a simple attach-rate assumption. Ask whether the go-to-market motion is product-led, sales-led, or consumption-based, because that determines which expansion sub-model to build first. Ask whether renewals sit with a dedicated team, because that creates a third forecast owner who must be pulled into reconciliation explicitly.

There are genuine counter-cases. A pure consumption business priced entirely on metered usage — no seats, no tiers — partially breaks the categorical frame, because new-customer onboarding and existing-customer workload growth are both just meters spinning faster. Those businesses should forecast off usage cohorts and net usage retention, treating the new-versus-expansion split as a secondary reporting view. Large multi-product enterprise deals that land a new logo and an immediate second-product attach in the same signature can defensibly be booked as one strategic-account line, provided the practice is documented and genuinely subsequent expansion is still tracked separately. And a mature, low-volatility SaaS business with a stable renewal base will see a smaller accuracy gain from splitting than a fast-changing one — the split is still worth it for the diagnostic and comp-plan reasons, but the pure accuracy argument is weaker.

The genuinely ambiguous cases deserve a stated tie-breaker rather than a fresh argument every quarter. A customer acquires another company that was already your customer — one logo or two? A division spins out and signs its own contract — new logo or expansion of the parent? A customer who churned fourteen months ago comes back — new logo or reactivation? Most disciplined taxonomies use a twelve-month cutoff for reactivation and treat it as its own category, because win-back economics resemble neither new-logo nor expansion. Pick conventions, write them into a one-page definitions document, and apply them uniformly. An investor will forgive a defensible convention; they will not forgive an inconsistency they find themselves.

The numbers behind each line: CAC, timing, and mix

The cash economics are where the difference stops being philosophical. New-logo ARR carries a CAC payback measured in the high teens to low twenties of months once you load in sales, marketing, BDR, and solutions-engineering time. Expansion ARR carries a payback in the single digits of months, because the acquisition work — the trust, the security review, the procurement relationship — was already paid for. Pure seat expansion into an already-deployed account is cheaper still; the incremental cost is a CSM conversation and an order form. The practical implication is that a dollar of expansion ARR is worth several times a dollar of new-logo ARR in cash terms, which is exactly why LTV/CAC and Rule-of-40 math fall apart when the two are blended.

The timing profiles differ just as sharply. New-logo ARR in enterprise SaaS is heavily back-half weighted — roughly a fifth of the annual number lands in Q1 and something like a third in Q4, as budget flush and rep accelerators converge. Coverage requirements should move inversely to that seasonality, because Q4 pipeline converts at a higher rate than January exploratory pipeline. A team carrying a flat 3x coverage all year is over-covered in Q4 and dangerously thin in Q1. Expansion ARR, by contrast, clusters around renewal anniversaries, which means its seasonality is a function of your own signing history rather than the buyer's fiscal calendar. Pull the renewal calendar forward two quarters and the expansion phasing writes itself.

For the new-logo model, five inputs explain most of the quarter-to-quarter variance. Late-stage pipeline coverage is the strongest single leading indicator — below roughly 2.5x on the quarter, the miss is already baked. Time-in-stage matters enormously: a deal parked in the final stage past about 45 days closes at a fraction of the blended rate and should be re-staged or stripped from the commit. Discount creep in the prior quarter predicts optimism in the current one, because reps who priced into the close last quarter will do it again. AE ramp state needs explicit weighting — a first-quarter rep should not be forecast at a fully-ramped rep's rate, and a team that hired six people in January will otherwise show a phantom Q2. And large deals that have not reached redlines by roughly week nine of a thirteen-week quarter slip at a high rate.

A single blended win rate hides most of what actually moves the number. Decompose it by lead source (inbound converts at roughly double outbound), by competitive presence (a no-competitor deal converts at multiples of a three-competitor deal), by deal-size band (sub-$25K ACV converts far better than $250K+), and by rep tenure. A forecast built on a 25% blended rate applied uniformly to a pipeline that is mostly large multi-competitor enterprise deals is structurally optimistic, and the miss will be blamed on execution when it was arithmetic.

For the expansion model, the driver set is completely different. Product adoption depth — what share of purchased features the customer has actually turned on — is the strongest single predictor; a deeply adopted account expands at a large multiple of a shallowly adopted one. Executive sponsor continuity is next: when the champion who signed the original deal leaves, expansion probability drops sharply and often does not recover for two or three quarters. Seat utilization against contracted seats is a mechanical trigger for seat expansion. Rolling usage against the commit band is the equivalent trigger for consumption customers. Support sentiment is the negative signal nobody wants to model — an account generating a steady stream of high-severity tickets is not an expansion candidate this quarter no matter what the CSM says on the call.

Mix itself carries a valuation consequence that is easy to under-appreciate. Two companies can post identical gross new ARR and be worth very different amounts. A logo-heavy mix with net revenue retention in the low hundreds implies high CAC, funnel dependence, and a treadmill. A balanced mix with strong NRR implies a compounding base, cash efficiency, and a smoother forward line — and public-market comparables consistently show the expansion-heavy profile carrying a materially richer multiple. Consumption-led companies with well-above-100% net retention have historically traded at the premium end; companies hovering near flat retention trade at the low end. The mechanism is not sentiment, it is durability. A board deck that shows only a single gross-new figure hands the audience no way to tell the two apart, which means the company that earned the premium is the one failing to claim it.

One warning on reading these numbers: strong expansion can mask a churn problem. A business can post healthy-looking net retention while losing a meaningful share of its logos every year, because a handful of large accounts expanding hard mathematically covers a long tail bleeding out. That is why gross retention must always appear alongside net retention. High NRR with weak GRR is a red flag, not a green one — it means the headline metric depends on a small set of whales, and the day one leaves, the number collapses.

Building it: data model, sequencing, and the operating rhythm

Sequencing matters more than sophistication here. Teams that jump straight to a telemetry regression before fixing categorization end up with a beautifully fitted model trained on garbage labels.

Start at the data layer. Every closed-won opportunity should carry an explicit movement-type field — new-logo, expansion-seat, expansion-tier, expansion-cross-sell, expansion-usage, contraction, churn, reactivation — populated at close as a required field, not inferred later by a quarterly SQL job. Inference-after-the-fact is where categorization rots: an analyst who does not know the account history buckets a genuine cross-sell as a new logo, and now your mix reporting is wrong in the direction that flatters you. Mirror that taxonomy in the chart of accounts so the general ledger splits new-logo, expansion, renewal, reactivation, and contra-revenue for downgrades. When the GL mirrors the taxonomy, the quarter-end reconciliation is a mapping exercise instead of a forensic project.

Then build the two models as three layers each. On the new-logo side: a rep-level judgment forecast (commit, best case, pipeline), a statistical layer applying historical close rates with time-decay and a discount haircut, and a pipeline-creation layer modeling future quarters from top-of-funnel conversion. On the expansion side: an account-level CSM forecast by sub-type with confidence ratings, a telemetry-driven model fed by adoption depth and utilization, and a renewal-tied layer that models expansion probability against contract terms for every account renewing in the quarter. Variance between the judgment layer and the statistical layer should stay within roughly 12% on the new-logo side and can run wider on expansion, because telemetry is genuinely noisier than late-stage pipeline. Persistent variance above that band means reps are sandbagging or pumping, and that is a conversation to have before the board call.

The reconciliation layer is where most teams cut corners. Its job is to catch three specific failures. First, double-counting: an AE and a CSM both forecasting the same cross-sell, which requires a single designated forecast owner per opportunity in the CRM. Second, timing errors: expansion booked in the quarter its telemetry signal fired rather than the quarter it will actually contract. Signal-to-signature lag runs roughly 30-60 days for a usage true-up and 60-120 days for a cross-sell, so a signal firing in week three of a quarter frequently closes in the next one. Teams that skip this phasing over-forecast the current quarter, under-forecast the next, and produce a sawtooth that erodes board trust faster than a straight miss would. Third, unapplied renewal risk: a $200K expansion attached to a contract with a 60% likelihood-to-renew belongs in the roll-up at $120K, not $200K.

Accept that no single tool covers both sides. Revenue-intelligence platforms score deals well and know almost nothing about seat expansion, because the signal they need sits in a product-analytics tool they do not read. Customer-success platforms carry health and usage rollups and have no view of late-stage pipeline. The practical answer is a two-tool stack with an explicit coverage map, and a pipeline that lands both sides in one warehouse table before the roll-up. Do not wait for a vendor to forecast both halves well.

Comp design belongs in this section because a bad plan corrupts the forecast at the source. If new-logo pays roughly double what expansion pays and reps control the categorization field, some cross-sells will be labeled new logos, and your mix reporting — the thing your valuation multiple keys off — is built on a bias you paid for. The architecture that works at scale: AEs carry a primary new-logo quota with a capped share retirable from expansion, accelerators gated on new-logo only, CSMs or account managers carry a separate expansion quota at a lower rate with a health-score gate. Keep expansion accelerators off the shared curve until expansion is genuinely the primary growth engine, or reps will clear thresholds on the easy motion and coast. Credit timing should vary by sub-type: seat adds credit cleanly at signature, while a ramped cross-sell with a free period is better credited at contract start so nobody is paid in full on revenue that has not begun. And price-uplift or CPI-clause ARR — real expansion that no rep sold — should be tracked and reported separately so contractual-mechanism growth does not flatter the CS team's apparent performance.

The operating rhythm that keeps both models honest is weekly and exception-based. New-logo forecast call early in the week: pipeline hygiene, stage movement, discount review, and — critically — the pipeline-creation number for two quarters out, because a thin Q3 discovered in Q3 can only be fixed with discounting. Expansion forecast call the next day: account health movement, adoption-score changes, champion status, renewal-window accounts. Reconciliation midweek. FP&A integration and the board-grade dashboard refresh at the end. Run the forecast calls on variance, not on a rep-by-rep walk of every open deal — inspect what moved, pressure-test the commit rather than the upside, and make every rep name the single deal most likely to slip.

Finally, grade yourself. Record the week-three commit, the week-nine commit, and the actual for each model separately, every quarter, and trend it. A team that does this discovers within two quarters that its expansion forecast runs, say, systematically optimistic early and converges late — and can then apply a known correction. A team that never scores itself repeats the same systematic error indefinitely. Forecast accuracy is itself an improvable metric, and treating it as one is the cheapest accuracy gain available to a RevOps team.

Related questions

Does net revenue retention include net new ARR?

No. NRR measures only the cohort that existed at period start — beginning ARR plus expansion minus churn and contraction, divided by beginning ARR. Adding new logos to the numerator inflates it, and any investor modeling NRR from your cohort data will catch it immediately.

Where does a price increase belong?

Contractual price uplift is genuine expansion ARR, but it should be reported as its own line rather than folded into customer-success-driven expansion. Blending them makes an automatic CPI clause look like adoption-led growth and distorts both the CS scorecard and the forecast baseline.

Is overage revenue expansion ARR?

Not until it is contracted. Uncommitted overage is variable, not recurring — track it as a pipeline of variable revenue opportunity, then move it into expansion ARR when the customer converts it into a higher commit band at renewal or true-up.

How do you handle a reactivated customer?

Treat reactivation as its own category with a stated cutoff — commonly twelve months. A return inside the window is reactivation; outside it, a new logo. Win-back economics resemble neither new-logo nor expansion, so forecasting it as either introduces avoidable error.

Which line should a Series B company invest in first?

Usually new logo, because expansion compounds off base size and a small base caps it. But if expansion is under roughly 25% of gross new ARR at meaningful scale, that is a customer-success and product-adoption problem, and fixing it returns cash faster than adding reps.

FAQ

What's the simplest way to explain the difference to a non-finance stakeholder?

Expansion ARR is more money from customers you already had — more seats, a higher tier, an extra product. Net new ARR is the first money from someone who has never paid you. "More from the same" versus "first from a stranger." Everything else — the comp plan, the forecast model, the multiple — follows from that one distinction.

Can net new ARR ever be negative?

Under the strict definition used here, no — new-logo ARR has a floor of zero, since you cannot sign a negative number of new customers. But some FP&A teams use "net new ARR" to mean the total period change in ARR after churn and contraction, and that absolutely can go negative. This is exactly why every slide must label which definition it is using.

How long does it take to stand up two separate models?

The data-layer work — taxonomy, required fields, GL mapping — is typically a quarter of focused effort and is the part that cannot be skipped. The models themselves come together faster, often within a quarter after the data is clean. Expect two full quarters of running both before the reconciliation variance settles into a trustworthy band.

Our expansion forecast is always wrong. What is the most likely cause?

Almost always timing rather than magnitude. Teams book expansion into the quarter the signal fired instead of the quarter it will contract, and the sub-type lags differ by 30 to 120 days. Phase each opportunity by its sub-type lag before assuming the underlying opportunity assessment is wrong.

Does this apply to a usage-based pricing model?

Partially. For metered consumption businesses, the more natural primitive is workload growth and net usage retention, since new-customer onboarding and existing-customer expansion are both just meters moving. Keep the categorical split as a reporting and comp view, but forecast off usage cohorts.

Who should own the reconciliation layer?

RevOps, not sales and not finance. Sales owns the new-logo model, customer success owns the expansion model, and finance owns the roll-up into the corporate plan — which leaves a neutral party needed in the middle to arbitrate double-counting, phasing, and renewal-risk haircuts without a quota in the fight.

Sources

• https://www.bvp.com/atlas — Bessemer Venture Partners Atlas: cloud metrics, Rule of 40, and retention frameworks
• https://openviewpartners.com/blog/ — OpenView research on SaaS growth, expansion, and product-led benchmarks
• https://www.saastr.com/ — SaaStr operating content on ARR waterfalls and forecasting
• https://chartmogul.com/reports/ — ChartMogul SaaS benchmark reports on net and gross revenue retention
• https://www.key.com/businesses-institutions/industry-expertise/software-technology.jsp — KeyBanc Capital Markets annual SaaS survey
• https://fasb.org/ — FASB, ASC 606 revenue recognition from contracts with customers
• https://www.sec.gov/edgar/search/ — SEC EDGAR full-text search for public SaaS retention and ARR disclosures
• https://www.gainsight.com/resources/ — Gainsight customer-success and net-retention research
• https://www.pwc.com/us/en/services/audit-assurance.html — PwC guidance on software and subscription revenue recognition
• https://www.bridgegroupinc.com/research — The Bridge Group research on SaaS sales productivity and quota structures

Related on PULSE

• What is the difference between ARR and MRR for a SaaS business?
• Why is net revenue retention compressing in 2027 and how do you fix it?
• How do you operationalize net revenue retention in 2027?
• Should I Hire a Fractional CRO If My Net Revenue Retention Is Below 100 Percent?
• How do you separate NRR, GRR, and logo retention when board auditors ask which is 'real'?
• How do we comp reps on expansion/upsell deals when they're working alongside a CSM or account manager?