How'd you fix Fivetran's revenue issues in 2026?

The two options compared: re-price the core, or go vertical

Every "fix the revenue" conversation about a data-movement company collapses into the same fork, and Fivetran's is unusually clean. Option A is a pricing and packaging fix: keep the product roughly as-is, but change how it is sold so that bills stop surprising finance teams. Option B is a product and segment fix: accept that managed ELT is drifting toward commodity, and move the center of gravity toward workloads that resist commoditization — AI context pipelines, regulated-industry ingestion, streaming freshness guarantees — where buyers will pay for reliability rather than rows.

Option A exists because consumption pricing has a well-documented failure mode in RevOps: the bill is a lagging indicator of a decision someone made months ago. A data team adds a Salesforce connector with heavy custom-object churn, or a product team ships an event schema that triples row counts, and the invoice moves without anyone approving the move. Finance sees variance; procurement asks for a cap; the account manager spends a renewal cycle defending consumption math instead of selling expansion. The revenue issue is not that customers stopped valuing the product — it is that the pricing model manufactures a recurring argument. The fix is to sell a shape the buyer can budget: a committed tier with a defined ingestion allowance, overage that steps rather than spikes, and a written freshness or uptime commitment attached to the number.

Option B exists because the competitive floor moved. Open-source ELT is genuinely usable now, and warehouse vendors keep pulling ingestion into their own platforms. When the alternative to a paid connector is a free connector shipped by the warehouse the customer already pays for, the paid connector has to be worth something beyond "it moves rows." That something is usually operational: someone else maintains the connector when the upstream API changes, someone else is on the hook when a schema drift breaks a downstream model at 3 a.m., someone else carries the compliance paperwork. Option B says: stop competing on connector count, start competing on the workloads where that operational burden is expensive enough to outsource.

The two are not mutually exclusive, but they compete for the same scarce resources — engineering headcount, sales retraining time, and the political capital required to change a comp plan mid-year. Choosing both simultaneously is how companies end up doing neither. The trade-off is real: Option A produces faster, more defensible revenue stabilization and almost no new TAM. Option B produces new TAM and higher ACV but takes three to four quarters before it shows in bookings, and it burns cash in the meantime.

There is a third option that shows up in most of these discussions and deserves to be named so it can be dismissed: discount to hold the logo. It is not a strategy. It converts a pricing problem into a margin problem, resets the customer's anchor permanently, and — because discounts are contagious across a shared buyer community — spreads to accounts that were never at risk. If the only lever being pulled by Q2 is discounting, the revenue issue has become a leadership issue.

How to decide between them

The decision is not a matter of taste. It is a matter of reading three signals in the existing book of business, and the signals point clearly enough that most teams already know the answer before they run the analysis.

Signal one: where is the churn concentrated? Pull the last four quarters of logo churn and dollar churn separately, and segment by ACV band. If logo churn is heavy in accounts under roughly $25K ACV but dollar churn is mild, the problem is self-service economics and bill unpredictability — that is an Option A problem. If dollar churn is concentrated in the $100K+ band and the loss reasons cite "moved to native ingestion" or "consolidated into the warehouse platform," that is an Option B problem and no amount of repricing will fix it.

Signal two: what do the loss reasons actually say? Most RevOps teams have a closed-lost field that nobody trusts. Fix it before deciding. Force a structured taxonomy — price, predictability, native alternative, open-source alternative, reliability, missing connector, consolidation, no-decision — and backfill the last two quarters by listening to calls or interviewing the AE. The distribution across those eight buckets is the single most decision-relevant dataset in the company, and it usually takes two analysts about ten days to assemble. Teams skip it because it feels like homework and then spend two quarters building the wrong fix.

Signal three: what is the shape of expansion in surviving accounts? Net revenue retention tells you whether the existing motion still compounds. If NRR in accounts older than eighteen months is meaningfully above 100% and the growth comes from organic volume rather than upsell, the core product is healthy and the issue is at the front door — new logo acquisition and pricing friction. If NRR is flat or below 100% in that cohort, the product is losing ground inside accounts that already bought it, which is the strongest possible argument for Option B.

A fourth input matters and is rarely written down: how much runway the company has. Option A can be executed inside a single quarter with existing headcount because it is mostly contract terms, packaging, and enablement. Option B requires reallocating engineers away from work they are currently doing, which means someone must decide what stops. If the answer to "what stops" is "nothing, we'll do it with the team we have," Option B has not actually been chosen — it has been wished for.

The practical resolution for most companies in this position is sequenced rather than either-or. Stabilize pricing in the first two quarters so the base stops leaking, then use the stabilized base as the funding source and reference-customer pool for the vertical push. Reversing the order is the common mistake: a vertical push launched on top of a leaking base spends its first year replacing churn rather than adding growth, and the board reads flat net-new as strategy failure when it was actually sequencing failure.

Concrete numbers behind each option

Put numbers on both paths or the debate stays theological. The arithmetic below uses illustrative ranges — the point is the structure of the model, not precision about any specific company's internals, which are not public.

Option A, the repricing path. Start by segmenting the base into three tiers by monthly ingestion volume. For each tier, compute the trailing twelve-month bill variance: the ratio of the highest month to the lowest month. Accounts where that ratio exceeds roughly 2x are your predictability-risk cohort, and in most consumption businesses that cohort is disproportionately represented in churn. The repricing offer to that cohort is a committed annual tier priced at approximately the trailing twelve-month average plus a modest premium — call it 10-15% — in exchange for a hard cap on overage exposure and a written freshness commitment.

The math the customer sees: they pay slightly more on average and eliminate the tail risk of a 3x month. The math the vendor sees: revenue per account is roughly flat to slightly up, but forecastability improves dramatically and renewal conversations stop being about invoice archaeology. If a cohort of 200 accounts averaging $60K ACV converts at 30-35%, that is roughly 60-70 accounts and $3.6M-$4.2M of ARR moving from volatile to committed. The revenue lift is modest; the churn avoided is the actual return. If that cohort was churning at 20% annually and repricing cuts it to 12%, the saved ARR is roughly $960K per year on that cohort alone, and it recurs.

Costs to model honestly: legal and contract-template work, roughly one quarter of a contracts person; sales enablement to teach a new close motion, typically two to three weeks of ramp per rep with a visible productivity dip during it; and a billing-system change that is almost always underestimated because usage metering and committed-tier accounting are different data models. Budget a quarter of engineering time for billing alone.

Option B, the vertical path. The unit economics are inverted — fewer accounts, much higher ACV, much longer sales cycles. A vertical-depth motion aimed at AI and analytics platform teams targets accounts in the $200K-$600K range rather than $50K, but the sales cycle stretches from roughly 60-90 days to 150-240 days, and the win rate on first-time vertical deals is lower until reference customers exist. Model the first year at a low win rate and a long cycle, or the plan will show bookings in Q2 that cannot physically arrive until Q4.

The engineering cost is where Option B gets expensive. Deepening a vertical means building and maintaining connectors and semantics for that vertical's systems, plus the compliance surface — audit logging, data residency, access controls, and whatever certification the vertical demands. That is not a two-engineer project. It is a sustained allocation, and the honest way to fund it is to kill something. The obvious candidate is the long tail of low-usage connectors: in most connector portfolios, a large majority of usage concentrates in a small minority of connectors, and the tail consumes maintenance attention wildly out of proportion to the revenue it defends. Deprecating the tail is politically hard because every connector has one loud customer, but the maintenance savings are the funding mechanism.

The comparison that matters. Option A returns cash inside two quarters, has bounded downside, and does not change the company's competitive position. Option B changes the competitive position but does not return cash for three to four quarters and can fail outright if the vertical bet is wrong. A board evaluating both should ask a single question: is the current product losing to a cheaper version of itself, or to a fundamentally different architecture? If cheaper version, Option A buys enough time to matter. If different architecture, Option A is a delay tactic and the vertical bet is mandatory.

Adjacent to both is a channel angle worth pricing separately. Cloud marketplace listings let buyers spend committed cloud budget on the product, which removes a procurement step and shortens cycles measurably. Marketplace deals typically carry a platform fee, so gross margin per deal is lower, but the CAC is dramatically lower too — the deal arrives partly pre-qualified. Model marketplace as a distinct segment with its own CAC-to-LTV ratio rather than blending it into direct, or the blended numbers will hide both the margin hit and the efficiency gain. The same logic applies to systems-integrator resale: SI-sourced revenue carries a partner margin but arrives attached to an implementation budget the vendor did not have to create.

Implementation details and sequencing

The order of operations determines whether this works. Below is a sequence that assumes the sequenced strategy — pricing stabilization first, vertical depth second — and assumes a company with a functioning RevOps team but imperfect data hygiene, which is the normal starting condition.

Weeks 1-4: instrument before you act. Fix the closed-lost taxonomy and backfill two quarters. Build the bill-variance report by account. Compute NRR by cohort age and by ACV band. Nothing ships in this phase, and that is correct — the most expensive failure mode in a revenue fix is a confident intervention aimed at a misdiagnosed problem. Assign this to RevOps with a named owner and a date, not to "the team."

Weeks 5-10: design the committed tier and pilot it. Do not roll out repricing to the whole base. Pick 30-50 accounts across the variance cohort, offer the committed tier as a named pilot, and instrument acceptance rate, negotiation friction points, and which contract clauses draw the most redlines. The redline pattern is the single most useful output — it tells you exactly which commitment the buyer does not believe you can keep. If the freshness SLA draws consistent redlines, the SLA is either too aggressive or the buyer has been burned by a past incident, and you learn that for the price of a pilot rather than a full rollout.

Weeks 8-14: fix the billing system in parallel. This runs concurrently because it is the long pole. Committed tiers with metered overage require the billing system to track entitlement, consumption, and the boundary between them, then produce an invoice a controller can reconcile. Teams routinely try to run this on spreadsheets for "just the pilot" and then discover the spreadsheet has become production. Fund it properly the first time.

Weeks 12-20: enable the field, then roll out. Enablement is not a deck. It is a certification: every rep closes a committed-tier deal in a role-play against an objection script drawn from the pilot redlines, and nobody carries the new offer to a live account before passing. Comp plan changes ship with the rollout, not after — if the plan still pays more for a consumption close, reps will keep selling consumption regardless of the strategy deck. Pay a modest accelerator on committed-tier conversions for the first two quarters to buy behavior change, then normalize.

Quarter 3 onward: the vertical push. Only start this once the pricing rollout has produced two clean months of churn data. Fund it by deprecating the connector tail on a published schedule with a long deprecation window — twelve months is normal, and announcing it badly is how you manufacture the churn you were trying to prevent. Pair every deprecation notice with a named migration path and a human conversation for any account above a revenue threshold.

The RevOps operating cadence that holds it together. A weekly pipeline review that inspects committed-tier conversion separately from consumption renewals; a monthly cohort review that reads NRR by ACV band; a quarterly win-loss review that re-reads the loss taxonomy and asks whether the distribution moved. If the loss distribution has not shifted away from "price predictability" within two quarters of rollout, the repricing did not work and it is time to say so out loud rather than defend it for another year.

Downstream effects to watch. Committed tiers change forecasting mechanics — revenue becomes more predictable but expansion becomes lumpier, arriving at renewal rather than continuously. Sales capacity planning shifts, because a committed-tier motion is a longer, more consultative sale than a consumption upsell. And customer success changes shape entirely: under consumption, CS optimizes for usage growth; under committed tiers, CS optimizes for the customer staying inside their entitlement while getting more value, which is a different job with different metrics. Rewriting the CS scorecard is part of the implementation, not an afterthought, and skipping it is how a well-designed pricing change quietly fails in year two.

Related questions

Does repricing away from consumption reduce total revenue?

Usually not in year one. Committed tiers priced near trailing average with a modest premium hold revenue roughly flat while cutting churn. The risk is year two: committed accounts that outgrow their tier need a clean upgrade path, or expansion stalls where it used to happen automatically.

How do you know if open-source alternatives are actually causing losses?

Ask in win-loss interviews rather than inferring from the CRM. Self-hosted alternatives usually appear in losses where the buyer has strong internal engineering and low volume. If the losing accounts have neither, the stated reason is likely a proxy for price or reliability.

Should the connector long tail be deprecated all at once?

No. Publish a schedule, give at least a twelve-month window, and pair every notice with a migration path. Batch deprecation without migration support converts a maintenance saving into a churn event, which is a strictly worse trade.

What is the fastest signal that the fix is working?

Bill-variance dispersion in the repriced cohort, measured monthly. It moves before churn does, because churn is a lagging indicator gated by renewal dates. If variance compresses and churn does not follow within two renewal cycles, the diagnosis was wrong.

How does this apply to other consumption-priced infrastructure vendors?

The same fork appears across observability, search, and warehouse-adjacent tooling: reprice for predictability, or move up-stack into workloads that resist commoditization. The diagnostic — churn concentration, loss taxonomy, cohort NRR — transfers directly.

FAQ

Why is consumption pricing a revenue risk when it was originally a growth advantage?

Consumption pricing wins early because it removes the barrier to starting — a buyer can adopt without a large committed spend. The same mechanism becomes a liability at scale, because the bill grows from decisions made by engineers rather than approved by finance. The advantage and the risk are the same property viewed from different points in the customer lifecycle.

Can a company run committed tiers and consumption pricing at the same time?

Yes, and most eventually do — consumption for self-service entry, committed tiers above a volume threshold. The operational cost is real: two pricing models means two billing paths, two forecast methodologies, and two sets of sales motions. Run both only if the segments are genuinely distinct and the boundary between them is clearly defined.

How much of this is a RevOps problem versus a product problem?

The diagnosis is entirely a RevOps problem — nobody else has the data to tell price predictability apart from competitive displacement. The fix splits: repricing is largely RevOps, contracts, and enablement, while vertical depth is a product and engineering commitment that RevOps can only measure. Misassigning the fix to the wrong function is a common failure.

What happens to net revenue retention under committed tiers?

It typically becomes less volatile and slightly lower in the short term, because organic usage growth that previously flowed straight to revenue is now absorbed inside the entitlement. The offset is lower churn and cleaner expansion conversations at renewal. Model both effects before committing, and expect the board to ask about the NRR dip.

Is acquiring a competitor or adjacent technology a substitute for either option?

Rarely a substitute, sometimes an accelerant. Acquisition can buy time or capability, but it does not fix a pricing model that manufactures customer conflict, and integration cost typically consumes the first year. Treat an acquisition as an input to Option B, never as a replacement for the diagnosis that justifies it.

How long before the board should expect to see results?

Bill-variance compression within one quarter of rollout, churn improvement within two to three quarters depending on renewal timing, and vertical bookings no sooner than three to four quarters after that motion starts. Any plan promising vertical revenue in the same quarter it launches is describing a hope, not a forecast.

Sources

• https://www.fivetran.com/ — official product, connector, and pricing documentation
• https://www.gartner.com/en/information-technology — analyst coverage of data integration and ETL markets
• https://www.forrester.com/research/ — research on data pipeline and analytics platforms
• https://www.crunchbase.com/organization/fivetran — funding history and company profile
• https://techcrunch.com/ — reporting on data infrastructure companies and funding events
• https://www.sec.gov/edgar/searchedgar/companysearch — public filings for comparable public data-infrastructure companies
• https://docs.snowflake.com/ — documentation on native ingestion and connector capabilities
• https://docs.databricks.com/ — documentation on ingestion, Unity Catalog, and streaming
• https://aws.amazon.com/marketplace/ — cloud marketplace listing and procurement mechanics
• https://www.saastr.com/ — operator commentary on SaaS pricing, retention, and net revenue retention

Related on PULSE

• What is Fivetran and why is it a hot RevOps data movement platform for 2027?
• What is Census (Fivetran Activations) and why is it a hot RevOps data-activation platform for 2027?
• Should Snowflake acquire Fivetran?
• How'd you fix Aston Carter's revenue issues in 2026?
• How'd you fix CyberCoders's revenue issues in 2026?
• How'd you fix Illinois's NIL & athletic revenue issues in 2026?