How do we turn product integrations into shared growth?
Turn product integrations into shared growth by choosing one customer task that both products improve, agreeing each partner's delivery and distribution commitments, and measuring completed work alongside each company's commercial contribution. Start with a bounded pilot. Expand when customers repeatedly achieve the outcome and both partners can sustain their share of the work.
Turn a product integration into shared growth by making one customer task work better across both products, then agreeing how each partner will deliver, distribute and sustain that improvement. Define useful completion before building. Measure recurring customer outcomes, each company’s incremental business and each team’s operating burden separately. Helping your partner serve the shared customer is part of protecting the value of your own product.
Which product integrations should we build first?
Prioritize product integrations that remove a demonstrated obstacle for a defined customer segment, with a partner willing to maintain the workflow and help customers adopt it. A large partner audience or repeated request is a lead for investigation. Require evidence of the task, the missing capability and the customer’s willingness to change their process.
Illustrative example, used throughout this page: a proposal product and a customer relationship management system, or CRM, connect an accepted proposal to the right sales opportunity and the team responsible for delivery. The intended outcome is an accurate, acknowledged handoff. Copying the acceptance date is one technical step; delivery staff knowing what was agreed completes the customer task.
Observe how customers do that work today. Include existing connectors, spreadsheets, internal rules and deliberate human review. Ask users to show recent exceptions. Do they need a new integration, a corrected field definition or an accountable recipient? Recommend the smallest change that resolves the actual obstacle, even when neither partner gains another subscription.
Use this selection record before discussing a launch:
| Decision | Evidence to request | Reason to defer |
|---|---|---|
| Recurring customer need | The same blocked task across the intended segment | One unusual request dominates the roadmap |
| Complementary products | A capability each product contributes to useful completion | The combination adds another interface without improving the task |
| Customer readiness | A user and administrator willing to try the changed process | Interest exists only in a sales conversation |
| Reachable demand | Eligible shared accounts and a separate route to suitable new buyers | A large audience with no relevant workflow |
| Delivery feasibility | Required endpoints, permissions, field mappings and maintenance owner | Necessary access is unavailable or unstable |
| Reciprocal commitment | Named work and an accountable owner on both sides | Only a logo exchange or vague introductions |
Merge’s integration strategy guide recommends aggregating requests through surveys, CRM records and sales conversations rather than following the most vocal requester. Apply that advice by grouping requests by segment and task. Distinguish a missing integration mentioned in a lost deal from evidence that its presence would have changed the decision.
Map current shared customers separately from each partner’s additional eligible audience. The former helps assess adoption; the latter tests distribution. Count unique accounts rather than adding both customer lists together. Use only an agreed, permitted matching process. A shared account still needs a relevant problem and an operating owner.
Read also: growing through an open-source ecosystem.
Which customer task would you still improve if neither company could claim the resulting lead?

How deep should a product integration be?
Choose integration depth around the customer task, including permissions, failure recovery and the frequency of change. A referral, embedded interface, scheduled transfer and two-way synchronization are different interventions. Specify which one you are evaluating before estimating its benefits or comparing adoption.
For the proposal example, start by asking whether an accepted proposal must create a record, update an existing one or request review. Decide which product controls proposal terms and which controls opportunity status. Do not let a later synchronization silently overwrite a reviewed agreement.
| Integration form | Appropriate use | Extra responsibility |
|---|---|---|
| Referral or contextual link | The customer needs access to the complementary product | Explain the handoff and obtain appropriate permission |
| One-way transfer | One product supplies an authoritative event or field | Resolve identity, duplicates, missed updates and corrections |
| Embedded interface or action | Users need to inspect or act without changing applications | Preserve user permissions and show action status |
| Two-way synchronization | Both products legitimately change shared information | Define field ownership, conflicts, deletions and loop prevention |
| Joint offering | Customers need coordinated purchase and delivery | Agree contracting, onboarding, support and exit |
Choose the implementation route separately. Existing native functionality is a sensible first comparison. n8n’s documentation describes workflow automation, while Merge’s integration guide describes its unified API offering. Consider a workflow tool for a bounded process, a unified API for repeated integration patterns, or direct development where the required behavior warrants it. Check supported fields and operations rather than assuming an abstraction covers every exception. Each route still needs an operating owner.
Compare the downsides explicitly. For a workflow tool, include administration and recovery outside the core product. For a unified API, inspect abstraction limits, access to original records and an exit route. For direct development, allocate the maintenance and specialist capacity. Choose based on the supported task and continuing obligations rather than connector count.
An event-driven design also needs recovery. Stripe’s webhook documentation explicitly describes duplicate deliveries and events arriving out of order. For your chosen APIs, inspect retry, ordering, authentication, versioning and rate-limit rules. Require safe replay, visible failure and reconciliation appropriate to those rules.
Example acceptance exercise: deliver the proposal-accepted event twice, then deliver a correction after the opportunity has changed. Ask the operator to show why there is one handoff, which values remain authoritative and how the receiving team learns about the correction. Also exercise revoked access and a missing recipient. A successful API response alone does not satisfy those acceptance criteria.
For the underlying data design, see connecting CRM, email and marketing tools with AI. Keep AI interpretation separate from deterministic transfers when the task does not require judgment.
What should an integration partnership agreement cover?
An integration partnership agreement should allocate customer-facing responsibilities, data use, maintenance, commercial rights and exit before promotion begins. Preserve each product’s identity and each company’s ability to serve customers independently. Cooperation should make the combined workflow stronger without quietly transferring control of one business to the other.
Create a working agreement alongside the formal contract. Give every recurring responsibility a named owner and substitute on each side. Titles such as product, partnerships and customer success are insufficient when nobody has accepted the work.
| Area | Decision to record | Proposal example |
|---|---|---|
| Product scope | Supported task, prerequisites and exclusions | Accepted proposals reach an existing opportunity; historical migration is separate |
| Customer relationship | Who contracts, bills, qualifies and communicates | Each supplier retains its product relationship; a named person coordinates the joint handoff |
| Data and access | Fields, purposes, permissions, retention and deletion | Share the necessary agreement reference, not the entire account history |
| Support | First contact, escalation, incident ownership and service levels | One entry point coordinates diagnosis across both products |
| Maintenance | API changes, deprecation notice, compatibility checks and release notice | Each side names the person who approves mapping changes |
| Distribution | Specific placements, seller preparation and follow-up | Each partner demonstrates the task to eligible customers |
| Exit | Notice, customer communication, export and transition | Customers can recover their records and use a documented alternative |
Avoid treating the customer as property to divide. Allocate responsibilities and communication rights while preserving the customer’s choice. Explain which product holds which information and who handles a failed handoff. If one supplier withdraws, the customer should receive an intelligible transition rather than competing instructions.
Where the GDPR applies, Articles 5 and 6 require purpose limitation, data minimization and a lawful basis. Assess roles from the actual processing: joint determination of purposes and means engages Article 26; processing on another controller’s behalf engages Article 28. Also assess security and international transfers where relevant. A commercial partnership label does not settle those roles. Ask the responsible privacy adviser to review the actual data flow.
Example: permission to copy a proposal reference for delivery coordination should not be treated as permission to upload the customer’s contacts into the other partner’s prospecting system. Specify those purposes separately and assess the basis for each. Keep customer-facing explanations consistent with what the integration actually does.
Examine dependency explicitly. Can either partner change access terms, withdraw essential fields, compete for the same workflow or restrict contact with customers? What remains usable if that happens? Record a fallback, migration owner and decision date for unacceptable concentration. Supporting other platforms may preserve options, but assess the additional maintenance before promising portability.
Carst and Hu’s systematic ecosystem review discusses complementors’ autonomy and notes that their dependence on a platform may limit it. It supports examining coordination and value capture, not assuming every partnership creates positive network effects. A useful combination of two products is not, by itself, evidence that additional participants make either product more valuable.
Could your company leave the partnership without preventing the shared customer from continuing their work?
How do we make the economics worthwhile for both partners?
Evaluate integration economics separately for each partner and for the customer. Shared growth requires an acceptable exchange of benefits, effort and risk; identical lead counts are unnecessary. Record who receives the benefit, who performs the recurring work and what alternative that work displaces.
Use a contribution record rather than one combined pipeline headline:
| Participant | Potential benefit to verify | Work and exposure to include |
|---|---|---|
| Customer | Accurate handoffs, less re-entry and fewer unresolved exceptions | Setup, permission reviews, training and ongoing checking |
| Proposal product | Better adoption, retained customers or relevant new opportunities | Build, API maintenance, onboarding, support, distribution, security and compliance review, and seller and support enablement |
| CRM partner | Useful product activity, expansion or suitable new accounts | Field governance, support, seller preparation and incident handling |
| Hours per month per side | Proposal product: ____ | CRM partner: ____ |
These are hypotheses for the illustrative example, not reported results. Separate one-time implementation from recurring delivery and support. Compare contribution after the relevant obligations with the next-best use of the same capacity. Keep avoided churn, new business and expansion distinct; the same account should not become several independent gains in the same calculation.
A partnership may be asymmetric and still worthwhile. The proposal supplier could receive introductions while the CRM partner improves existing customers’ use. Agree how each benefit will be assessed and when it should become observable. If one side accepts a temporary imbalance for learning, name that investment and its review date rather than calling uncertain future referrals compensation.
Assess incentives at the level of the people doing the work. Ask whether a seller receives credit for introducing the other product, whether support capacity is allocated and whether a product manager can protect maintenance time. A leadership agreement needs those operating commitments before a joint campaign.
Do not pursue the partnership when the recurring burden is unacceptable, a necessary partner commitment is absent or the customer’s existing approach is adequate. A maintained connector without joint selling can remain useful. A referral-only arrangement or a reviewed manual handoff can also be the right scope.
Whose workload increases when your partnership report shows a win?
How will customers and their AI assistants discover the integration?
Make an integration discoverable at the point where a suitable customer needs the combined task, and publish enough information to evaluate it. Use the marketplace, product interface, help centre and sales conversation for their specific roles. Treat a listing as an entry point whose adoption path still needs to be designed.
HubSpot’s technology partner guide points developers to setup guides and listing checklists and describes embedded app cards and workflow actions. Explain the integration’s specific value rather than repeating general product marketing.
Example discovery path: when a proposal user selects an opportunity, show the supported CRM connection beside that task. Explain the permissions, required configuration and result before authorization. After connection, guide the user to an acknowledged handoff and show where exceptions appear. Do not remove a meaningful approval merely to describe setup as one click.
Publish a joint use-case record with these fields:
- User and trigger: who needs the connection, and what starts the task.
- Supported result: the completed workflow and its boundaries.
- Prerequisites: supported versions, account access, fields and customer preparation.
- Setup and recovery: installation, first useful action, failures and disconnection.
- Responsibility: who provides each function and where support starts.
- Evidence and review: demonstrated capability, any measured outcome, source and review date.
Write the record as accessible page text, with stable links to current documentation. Ask a colleague, and separately a buyer’s AI assistant, to extract the prerequisites and explain a non-fit situation. Inspect the source used for every consequential statement.
Before adding a separate machine-readable format, identify which consumer needs it and who will keep it consistent with the approved description. Start with making a startup offer easy for buyer agents to evaluate, including prerequisites, limitations and a clear next step.
Maintain the approved integration scope, data definitions and support boundaries for internal assistants too. The same discipline applies when giving AI agents shared company knowledge: current instructions need an owner, and uncertain compatibility needs an escalation route. An assistant should be able to explain a limitation rather than invent an answer that keeps the sale moving.
How should partners market and sell the integration together?
A joint integration launch needs a commercial route matched to the customer’s buying task, with owners for introduction, qualification and follow-up. Co-marketing supplies an explanation or audience; referrals introduce suitable buyers; co-selling coordinates an opportunity; resale changes who sells to the customer. Specify which motion you intend before counting partner activity.
Start with existing customers whose workflow already contains the problem. Forecast eligible adoption and expansion opportunities by segment, not by the full customer base. Bring sales, marketing and customer success into the same plan. When the opportunity is uncertain, propose a narrow, representative pilot rather than assuming that selling to existing customers will always be more efficient.
For discovery, publish a joint walkthrough, a use-case page or a practical template. Run a webinar only when it reaches people with the relevant task. Put the offer into an existing customer interaction where it belongs, such as a setup step or adoption review. Avoid an unrelated promotion simply because the partner has traffic. Each asset should lead to a useful action with a receiving owner.
Example seller brief: “For sales teams already using both products, show how an accepted proposal reaches the responsible delivery team. Confirm the supported opportunity structure and handoff owner. If custom approval rules are required, send the configuration for review before promising compatibility.” Give sellers a demonstration, qualification questions, exclusions and a route to a technical answer.
Agree the referral process before automating it:
- The originating partner records why the buyer fits and what information may be shared.
- The receiving partner accepts, rejects or requests clarification within an agreed window.
- One named owner coordinates the next customer conversation.
- Both sides reconcile opportunity status, outcome and the contribution being claimed.
Microsoft provides a concrete platform example: its shared-referral API process creates a partner-led deal and a linked Microsoft referral. The lesson for your operating design is to preserve connected identifiers and responsibilities.
Example: the CRM partner introduces an existing CRM customer evaluating proposal software. The proposal supplier records the introduction against that buying process; the CRM partner records an adoption opportunity in its own account. Keep the two commercial outcomes distinct. Share only necessary status, not unrestricted access to both pipelines. If both sellers contact the buyer, agree the purpose of each conversation rather than sending competing follow-ups.
Can each partner name the next customer action it will personally help complete?
Which metrics show shared growth rather than integration activity?
A shared integration scorecard should connect eligible accounts, useful activation, repeated task completion and customer outcomes to each partner’s business contribution and operating burden. Keep those layers separate. Installation and API traffic diagnose access; they do not establish that a customer completed useful work or decided to stay.
The distinction is literal in HubSpot’s partner guide: an active install is a unique customer account making at least one API call in the previous 30 days, excluding test accounts. Define your own meaningful activation around the task, and label it separately from the platform’s metric.
Example scorecard for the proposal handoff:
| Layer | Proposed measure | Denominator or boundary |
|---|---|---|
| Eligibility | Accounts meeting the agreed workflow and access requirements | Defined segment, recorded before invitation |
| Setup | Connected accounts and authorization failures | Eligible accounts invited |
| Activation | First accurate handoff acknowledged by the receiving team | Invited accounts, with setup conversion also shown |
| Time to value | Elapsed time from invitation to acknowledged handoff | Show unfinished accounts separately |
| Repeated value | Correct, acknowledged handoffs and manual corrections | Eligible handoffs during a fixed window |
| Continuing adoption | Accounts completing the task when it recurs | Activated accounts with a genuine opportunity to repeat |
| Customer outcome | Less re-entry, fewer missed handoffs or faster delivery readiness | Comparable work, including review and repair |
| Partner contribution | New customers, retention and expansion for each company | Separate start cohorts and relevant decision windows |
| Burden and risk | Support time, incidents, maintenance and customer complaints | Defined account base and observation period |
Set quality boundaries beside the main metric. Faster handoffs should not compensate for incorrect terms or unauthorized disclosure. In the example, measure completion through acknowledgment, not merely record creation. If the receiving team still chases missing scope, inspect that work before declaring the transfer a success.
For commercial reporting, agree your credit rules in advance. A practical convention is sourced for a partner introduction that originates a qualifying buying process, influenced for a documented contribution to an existing process, and assisted for help delivering or adopting the outcome. Specify timing, evidence, exclusions and dispute ownership.
Deduplicate using stable account, opportunity and referral identifiers. Do not add sourced and influenced totals together when the same opportunity qualifies for both labels. Reconcile closed outcomes with the authoritative commercial records. Credit allocation answers who receives recognition under your rules; incremental impact asks what would have happened without the intervention.
If installs increased while completed customer work stayed flat, would your report still call the partnership successful?
How can we tell whether the integration caused better retention?
A retention difference between integration users and other customers does not by itself show an integration effect. Define the intervention, eligibility, starting point and observation period before comparing accounts. Investigate whether adopters already differ in product fit, size, tenure, prior usage or available operating capacity.
When assessing a vendor’s retention claim, ask whether it describes measured customer behavior, respondents’ beliefs or a selected case. Request the cohort definition, comparison, observation period and method. Do not use a reported association or a survey of opinions as an expected causal effect for your business.
For a prospective evaluation, compare eligible accounts from a common starting point. Hansford and colleagues’ BMJ Medicine methods paper explains how misaligned eligibility, treatment assignment and follow-up create bias. Applied to active customers here, avoid grouping accounts by whether they ever activate later while counting retention from an earlier signup. Future activators had to remain customers until activation. Record that timing explicitly.
Where feasible and appropriate, randomize rollout at the account level. Microsoft Research’s experimentation guidance explains why identifiers, enterprise constraints and interactions between users affect the unit of randomization. For the proposal example, assigning colleagues within the same shared account to conflicting versions would undermine the intended workflow. Have the evaluator assess cross-account effects too.
If access cannot be withheld, consider randomly offering additional discovery or onboarding help. Innovations for Poverty Action’s randomization guidance distinguishes encouragement designs from trials of the underlying service and explains analysis by original assignment. Compare invited groups as assigned. That estimates the effect of the offer or assistance; it does not automatically isolate software usage. Interpreting encouragement as an instrument for usage requires additional assumptions, including that encouragement affects the outcome only through adoption. Onboarding help that independently improves retention would violate that condition.
Illustrative evaluation scenario: all eligible accounts retain access to the connector, but randomly selected accounts receive a guided handoff setup. Record assignment before outreach and assess both groups over the same window, including accounts that decline. A difference supports a claim about offering that setup under the evaluation conditions. Comparing only people who accepted the help would reintroduce selection.
When randomization is impractical, compare suitable contemporaneous accounts and disclose residual confounding. Matching observed characteristics does not establish that unmeasured motivation or operating maturity is equal. Preserve pre-intervention behavior and account history, and ask an evaluator whether a longitudinal or quasi-experimental design is credible.
Plan sample size, decision thresholds and observation time around the smallest effect that would change the investment decision. Include uncertainty, missing outcomes and implementation deviations in the report. A small pilot can establish feasibility and reveal failure modes while leaving retention effects unresolved. The cited ecosystem research and platform documentation do not establish a universal causal return from B2B integrations. Treat this page’s operating method as a proposal to evaluate, not a measured recipe.
When should we scale, redesign or stop an integration partnership?
Scale an integration partnership when representative customers repeatedly complete the task, both partners sustain their commitments and the evidence supports the proposed next investment. Redesign the failing part when the problem is identifiable. Stop when a necessary condition remains absent or the expected benefit no longer justifies continued work.
Use the following pilot agreement template before building or launching:
For the named customer segment, our two products will improve the specified task against the customer’s current alternative. Useful completion means the agreed observable outcome. Each partner names a delivery owner, a distribution commitment and available maintenance capacity. We record eligibility, activation, repeated outcomes, exceptions and operating effort over a defined window. We review on an agreed date and expand, revise or close using criteria recorded before the results.
Make the observation window fit the task. A customer who has no proposal to hand off during the pilot is unobserved for repeated use, not necessarily disengaged. Likewise, a renewal outcome cannot be judged before renewal is due. Continue tracking those outcomes without quietly extending every activity until something looks positive.
| Observed state | Recommended next decision |
|---|---|
| Qualified discovery, incomplete setup | Repair authorization, configuration or onboarding before adding reach |
| Completed setup, no useful handoff | Revisit task fit, field meaning and receiving-team responsibility |
| Recurring customer value, no partner acquisition | Keep the product case separate; revise or end the distribution experiment |
| Customer value, unacceptable burden for one partner | Reduce scope, change the exchange or stop expansion |
| Credible value and sustainable work for both sides | Expand one segment or distribution route, preserving the comparison |
| Unresolved data rights, critical failures or absent maintenance | Pause affected operation and resolve the condition before continuation |
For the proposal example, an initial pilot could support proceeding with the handoff while leaving the referral campaign unproven. Treat those as separate decisions. Do not require a useful product improvement to carry an unsupported acquisition story. Conversely, a productive joint campaign should not hide an integration customers cannot reliably operate.
Use what to measure in early go-to-market tests to keep the decision, evidence and uncertainty connected. Close each review with the next change, owner and date. If the partnership ends, deliver the agreed customer transition and retain permitted learning about the workflow.
How can Poldermarketing help an integration reach useful adoption?
Piet Baudoin of Poldermarketing helps post-seed startups connect integration positioning with the go-to-market work needed for adoption: a clear joint use case, qualified follow-up, CRM records and reporting that distinguishes customer outcomes from partner credit.
Piet works remotely in Dutch and English. Alongside positioning and findability in Google and AI answers, he builds and runs AI automations with care, including lead follow-up, enrichment and qualification, CRM setup, reporting, content distribution and tool integrations. A concrete starting point is one integration whose customer task is understood but whose discovery, handoff or measurement needs work. Read how Piet builds and runs the work, or request a no-cost growth scan to identify that gap. Product engineering, privacy decisions and commercial commitments stay with the people qualified and authorized to own them.
Questions people ask
Does a product integration need a commercial partnership?
No. A connector can justify its existence by improving the core experience for existing customers. Add a commercial partnership when both companies can commit to helping suitable customers discover, evaluate or adopt the combined workflow. Keep the product decision separate from the distribution decision, and avoid making useful functionality depend on an unproven referral promise.
Should an integration sync data in both directions?
Choose two-way synchronization only when the customer task requires changes from both products. Specify which system controls each field, how conflicts are resolved and how repeated updates are recognized. A one-way handoff or embedded action may complete the task with less coordination. Deeper connectivity is a design choice, not a measure of partnership quality.
What if one partner gets most of the leads?
Compare the complete exchange before demanding equal lead counts. One partner might receive new customers while the other improves retention or expands product usage. Record those benefits separately, alongside support and maintenance obligations. Continue only when both sides accept a credible contribution, with a review date for benefits that remain uncertain.
Can we claim that an integration reduces churn?
A lower churn rate among integration users is an association until the comparison addresses selection, timing and other differences. Define eligibility and a common observation window before comparing accounts. A randomized rollout can support a causal estimate where feasible. Otherwise disclose the limits and describe the observed relationship without attributing the entire difference to the integration.
How do we start without building a full marketplace app?
Begin with a permitted referral, a reviewed manual handoff or an existing connector when that can answer the customer question. Name the owner, supported workflow, data boundary and decision date. Build deeper functionality only when the smaller arrangement reveals a necessary limitation. A useful pilot should resolve uncertainty rather than imitate the eventual launch.