Designing Loans for a K‑Shaped Economy: Product Strategies Lenders Need in 2026

The U.S. credit market in 2026 is not defined by one borrower story—it is defined by divergence. In a K-shaped economy, one segment of households continues to build wealth, accumulate assets, and qualify for increasingly favorable financial products, while another segment remains exposed to inflation pressure, uneven wage growth, and brittle cash flow. For lenders, this is not just a macro headline; it is a product design problem, a pricing problem, and an underwriting problem. The winners will be the institutions that move beyond one-size-fits-all credit and build segmented offerings for both Thrivers and Strivers, using smarter risk segmentation, alternative data, and responsible expansion tactics. For a broader macro view, see our guide on the K-shaped economy in 2026 and how the divide is changing in real time.

That shift in design thinking matters because credit demand is still strong, but the needs behind it are wildly different. Thrivers are looking for speed, convenience, premium rewards, and larger ticket sizes; Strivers are looking for predictability, lower fees, flexibility, and a path to build credit without being punished by legacy scoring blind spots. The most effective lenders in 2026 will therefore treat credit product design as a portfolio strategy, not a monolith. They will pair risk segmentation with tiered underwriting, dynamic pricing, and product features that help borrowers succeed instead of simply maximizing short-term yield. If you are building or evaluating a lender stack, our related piece on integrating BNPL without increasing operational risk is a useful companion.

In practice, that means designing for the borrower’s financial context, not just their score. It means understanding which applicants are temporarily constrained versus structurally risky, which customers can handle more credit if terms are right, and which new data signals actually improve prediction instead of just creating complexity. It also means avoiding the trap of “responsible lending” being interpreted as “do nothing.” Financial inclusion requires disciplined expansion, not blanket denial. And in 2026, lenders that fail to adapt will lose good borrowers to fintech competitors that can underwrite faster, price more precisely, and offer better borrower journeys.

1. What a K‑Shaped Economy Means for Lending in 2026

Thrivers and Strivers are different markets, not just different risk bands

In a K-shaped economy, Thrivers are households benefiting from asset appreciation, steady employment, and stronger liquidity. They may already have strong credit profiles, but they are also more likely to shop for better terms and switch providers if the value proposition is weak. Strivers, by contrast, often have volatile income, thinner savings, and higher sensitivity to fees, late charges, and payment timing. Treating both groups as the same “mass market” customer is one of the fastest ways to misprice risk or miss growth opportunities. The smartest lenders will separate these groups at the product level, not only at the risk score level.

Equifax’s recent analysis suggests the financial-health divide remains real, but the rate of divergence may be slowing, with some lower-score consumers stabilizing and Gen Z improving faster as they enter the workforce. That nuance matters: slowing divergence is not the same as convergence. Lenders should not assume broad healing; instead, they should identify sub-segments showing upward momentum and create pathways that reward progress. For example, a borrower moving from irregular gig income to a stable salaried role may deserve a different offer structure within months, not years.

The old credit decision tree is too blunt for divergent economies

Traditional underwriting was built for a relatively linear consumer market: score, income, debt, approve or decline. That approach still works for certain portfolio segments, but it increasingly fails to capture modern household dynamics. Two borrowers with the same score can have dramatically different cash-flow resilience, employment volatility, or savings buffers. One may be a Thriver with concentrated stock wealth and a high salary, while the other may be a Striver with thin-file credit but consistent deposits and low discretionary spending.

This is why lenders need more granular segmentation. The point is not to replace FICO or traditional bureau data; it is to supplement it with signals that better predict repayment behavior and product fit. The goal is a more accurate match between borrower reality and loan structure. For deeper context on how businesses adapt to changing consumer realities, review Equifax’s K-shaped economy overview and compare it with the practical product decisions discussed below.

Why 2026 is a product design inflection point

In 2026, lending is increasingly software-defined. Underwriting engines can ingest bank transaction data, payroll feeds, cash-flow analytics, device signals, and account history almost in real time. That creates a new competitive edge: lenders can build product ladders that start small, validate behavior, then expand responsibly. The old model of “approve big or decline” is giving way to staged access, where limit increases, term extensions, and pricing improvements are earned through performance.

This is especially important in credit card, personal loan, point-of-sale, and small-business lending. These products have different loss dynamics, but they share one challenge: many applicants are not static. Their risk profile can improve quickly, especially when better cash-flow data or payment history becomes available. Lenders that build adaptive offerings will be better positioned to serve both sides of the K without overextending capital.

2. Segmentation Framework: How to Separate Thrivers from Strivers

Use behavior-based segmentation, not just bureau bands

The first step is to move beyond score buckets. A borrower with a 640 score who maintains strong deposit inflows, low overdraft incidence, and stable payroll patterns may be a lower-risk customer than a borrower with a 720 score but heavy revolving utilization and unstable income. In a K-shaped economy, that distinction is not academic—it drives pricing, line assignment, and collections strategy. Lenders should create segmentation layers that combine bureau data with cash-flow, income stability, and transactional behavior.

A practical model might include four dimensions: capacity, stability, resilience, and growth trajectory. Capacity asks whether the customer can afford the payment today. Stability asks whether income and spending patterns are predictable. Resilience asks how much buffer exists if something breaks. Growth trajectory asks whether the customer is moving up or down financially. That framework lets lenders design offers that align with borrower reality, instead of forcing every applicant into the same risk box.

Alternative data should answer specific underwriting questions

Alternative data is only useful if it improves a defined decision. For example, bank transaction data can help verify recurring income and detect cash-flow seasonality. Payroll data can improve employment verification and reduce fraud. Rent and utility payment data can support thin-file borrowers. Device and identity data can help detect synthetic identity risk or application anomalies. But lenders should not add data indiscriminately; every signal needs a clear role in prediction, fraud reduction, or product qualification.

This is where governance matters. If your model cannot explain why a signal improves loss performance, it can become a compliance liability. Responsible use of alternative data means validating lift, checking for proxy discrimination, and ensuring adverse-action explanations remain clear. For a useful adjacent lesson on disclosures and reporting, read what platform risk disclosures mean for tax and compliance reporting. The broader principle is the same: transparency protects both trust and scale.

Borrower segment archetypes should map to product intent

Thrivers and Strivers should not simply receive different price points; they should receive different product architectures. Thrivers may want premium rewards, higher limits, instant funding, and smoother digital servicing. Strivers may benefit more from small starter limits, predictable installments, grace periods, and credit-building rewards tied to repayment behavior. If lenders do not intentionally separate those intents, they risk offering the wrong product to the right borrower or the right product to the wrong stage of financial life.

That is why segmentation should feed the entire lifecycle: acquisition, underwriting, pricing, servicing, collections, and limit management. A product that looks profitable at origination can become expensive if it ignores borrower volatility. Conversely, a product that seems conservative can become a high-LTV franchise if it is designed to help customers graduate to better terms over time.

3. Product Design Tweaks That Work in a Split Economy

Build tiered loan ladders instead of single-shot approvals

One of the most effective 2026 playbooks is the loan ladder. Instead of offering one static approval amount, lenders can create three or four product tiers that expand as the borrower proves reliability. A starter tier might cap exposure at a modest amount with shorter terms, autopay incentives, and clear milestone-based increases. A mid-tier could extend more credit after several successful cycles. A premium tier could offer lower APRs, larger limits, and flexible draw options for customers who consistently perform.

This approach is especially useful for Strivers and near-prime borrowers. It allows lenders to gather performance data while giving the customer a visible path forward. That path improves retention because borrowers can see a future inside the product rather than needing to shop elsewhere. The design principle resembles other staged product experiences, such as how financial health updates show that momentum matters as much as static score.

Offer split-product architecture for different use cases

Not every borrower needs the same loan structure. Thrivers often prefer convenience products: fast personal loans, premium card lines, or large-ticket installment financing. Strivers often need utility products: emergency credit, income-smoothing loans, debt-consolidation tools, or credit-builder products. A lender serving both groups should consider separate product families rather than one universal loan with a single price and term. That reduces cross-subsidy confusion and improves marketing precision.

For example, a lender could offer a “build” product with lower initial exposure and credit reporting designed to help thin-file borrowers establish history, alongside a “flex” product for customers with stronger profiles who want revolving access or larger installment capacity. If the borrower graduates from build to flex, the economics improve on both sides: lower losses early, higher retention later. This is the same logic behind successful membership-style financial products, similar to the pacing used in membership design that scales through tiers rather than one-time purchases.

Design features that lower regret and improve repayment

Product success is not just about approval rates; it is about payment behavior. The most lender-friendly design tweaks are often boring but powerful: due-date flexibility, payment reminders, partial-payment options where appropriate, autopay discounts, and transparent fee caps. Borrowers in the Striver segment benefit disproportionately from these features because they reduce the likelihood of avoidable delinquency. For Thrivers, convenience features like instant funding and self-service line increases create satisfaction and reduce churn.

A useful benchmark is to ask whether the feature helps the borrower either plan better, pay earlier, or recover faster from a temporary cash-flow shock. If the answer is no, it is probably a marketing embellishment rather than a risk-control improvement. For digital workflow inspiration, lenders modernizing their servicing stack may also find value in our guide to automation patterns that replace manual workflows, because the operational principle is identical: reduce friction where human process adds delay, not value.

4. Pricing Strategies That Match Risk Without Exploitation

Price for loss, not for desperation

Dynamic pricing is essential in a K-shaped economy, but it must be implemented with restraint. A high-risk borrower should not be charged a punitive rate just because they have limited options. Instead, lenders should price to expected loss, funding cost, servicing cost, and a reasonable margin, then use product structure to manage remaining risk. That might mean lower limits, shorter terms, or installment schedules rather than simply escalating APRs.

Overpricing vulnerable borrowers can backfire. It increases payment shock, raises delinquency, and creates reputational risk that can outlast the deal itself. The better approach is to use underwriting to sort risk more accurately and use pricing to reflect incremental differences. If your model is doing all the work, your pricing should not need to become predatory to compensate.

Use tiered APRs, fees, and rewards to influence behavior

Pricing architecture can shape borrower behavior. For example, lenders might offer lower APRs for autopay enrollment, on-time streaks, or verified income stability. They might remove origination fees for borrowers who complete a financial education module or who accept lower starting limits. On the Thriver side, premium tiers can monetize convenience via faster disbursement, concierge servicing, or rewards that offset the higher rate.

The key is alignment: each pricing lever should encourage a behavior that improves either repayment or engagement. Pricing should not be a blunt extraction tool. In a segmented market, thoughtful pricing can create a win-win by helping borrowers self-sort into the level of product they can truly sustain.

Benchmark against transparency and total cost of credit

Borrowers compare offers on more than APR, especially in digital lending. They look at funding speed, payment flexibility, fees, and service quality. That means lenders need to present total cost of credit clearly, including all material fees and likely behaviors that change cost over time. A product that appears cheap on the front end but layers on late fees, insufficient-fund charges, or rollover risk will lose trust quickly.

As an operational discipline, lenders should pair price testing with consumer clarity testing. Can a borrower understand what this loan costs in one minute? Can they see the consequences of missed payments without reading legal jargon? If not, the product may be compliant on paper and confusing in practice. For a related example of how to evaluate value versus hype, see how macro trends are reshaping credit behavior in real time.

5. Underwriting with Alternative Data: What Actually Helps

Cash-flow underwriting is the most practical upgrade

Among alternative data options, cash-flow underwriting is one of the most immediately useful. It can reveal recurring income, rent burden, debt service patterns, and volatility that traditional scores miss. A borrower with modest income but strong monthly surplus and disciplined spending may be a better risk than a higher-scoring borrower with erratic outflows. Cash-flow data also helps lenders identify temporary stress versus structural weakness, which improves both origination and workout decisions.

To deploy it responsibly, lenders should establish consistent lookback windows, define recurring transactions carefully, and distinguish between non-recurring inflows and durable income. The model should be tested across cohorts to ensure it is not simply favoring wealthier customers with more digitally visible financial lives. The goal is inclusion with precision, not inclusion with hidden bias.

Use alternative data as a supplement, not a replacement

Alternative data works best when it adds signal to a strong underwriting stack. It should not be used to override obvious red flags or replace common-sense debt capacity checks. Likewise, lenders should avoid assuming that more data automatically means better decisions. Extra data can increase false confidence, especially when correlations are weak or unstable across time.

That is why model governance matters. Every alternative-data field should have an assigned purpose, an approved storage policy, and a documented explanation for adverse actions. If an applicant is declined because of patterns inferred from bank transactions or device intelligence, the lender must be prepared to explain the decision in compliant language. The cleaner your logic, the easier it is to scale responsibly.

Build inclusion pathways for thin-file and emerging consumers

Financial inclusion in 2026 is not just about opening the door; it is about creating a credible first product. Thin-file consumers, Gen Z workers, gig earners, and recent immigrants often have payment behavior that looks strong in the real world but weak in a legacy bureau system. Lenders that build starter products with verified income, small initial limits, and easy graduation rules can unlock profitable growth while expanding access.

This is especially relevant as younger consumers begin building histories. Equifax notes that Gen Z’s financial health is improving faster than older cohorts in some datasets, which suggests a window to acquire them early with the right structures. A responsible lender can use that momentum to create a long-term relationship, rather than forcing a rejection that sends the customer to a less disciplined competitor.

6. Responsible Credit Expansion Playbooks

Start small, prove performance, then scale

The safest expansion playbook in a split economy is staged growth. Approve modest initial exposure, monitor payment behavior and cash-flow changes, then expand only after proof points are met. This reduces loss while giving borrowers a path to higher limits or better rates. It is also easier to communicate to regulators and customer advocates because the decision logic is progressive rather than arbitrary.

Use clear graduation triggers: on-time payment streaks, improved deposit stability, reduced utilization, or verified income growth. Then automate the review process so customers are not stuck waiting months for manual reassessment. The best programs make progress visible and attainable. That visibility increases trust and reduces attrition.

Design hardship pathways before the borrower needs them

Responsible expansion requires responsible exit ramps. If a borrower encounters a shock, the product should offer hardship options that preserve dignity and reduce long-term damage: skip-a-pay controls, due date shifting, temporary minimum-payment reductions, or fee waivers for documented hardship. These features can prevent a small disruption from becoming a default event.

Hardship design is not charity; it is portfolio protection. Borrowers who recover after a temporary shock often become better long-term customers than those who are pushed into collections or churn. For lenders serving volatile segments, this is one of the highest-ROI product investments they can make.

Pair growth with guardrails and monitoring

Expansion should always come with portfolio monitoring. Watch for early signals such as payment drift, rising utilization, overdraft frequency, and missed autopay attempts. If risk rises, the response should be proportional: pause limit increases, reduce promotional offers, or trigger support outreach. The goal is to prevent an optimistic growth strategy from becoming a hidden loss spiral.

One useful mindset is borrowed from operational planning in other sectors: scale only as fast as your controls can absorb change. That same discipline appears in articles on building analytics pipelines and maintaining offline workflow libraries—the principle is stable, durable process beats reactive improvisation.

7. Channel, UX, and Collections: The Hidden Product Layer

Borrowers judge the loan through the interface

In digital lending, the user experience is part of the product. A borrower who cannot understand the offer, locate payment information, or contact support will behave as if the product itself is broken. This is especially important for Strivers, who have less tolerance for ambiguity and fewer resources to absorb mistakes. Clear dashboards, payment reminders, transparent payoff paths, and plain-language notices are not nice-to-haves—they are risk controls.

For Thrivers, the UX challenge is different: speed, control, and self-service matter more than handholding. They want instant decisions, seamless funding, and easy account management across devices. A good design system should make both segments feel that the product is built for them, even though the underlying risk rules are different.

Collections should be segmented, human, and data-driven

Collections is often where lenders lose the trust they worked so hard to build. In a K-shaped economy, a one-size-fits-all delinquency strategy can damage healthy borrowers and push strained borrowers further into distress. Instead, collections should be segmented by reason for delinquency, not just days past due. A borrower who missed because of timing may need a reminder or date shift, while a borrower in structural trouble may need a structured workout.

Modern collections teams should use behavioral data, channel preferences, and hardship indicators to match the right intervention to the right borrower. That improves recovery and reduces complaint risk. It also creates a more humane experience, which can pay off in long-term loyalty and better word-of-mouth acquisition.

Operational design should support speed without losing control

Because lending products are increasingly digital, backend workflows matter as much as front-end offers. If underwriting, servicing, and collections systems are fragmented, the borrower experience becomes inconsistent and operational risk rises. Institutions modernizing their stack can borrow from other automation-heavy domains, including manual workflow replacement strategies and other cloud-native operations playbooks. The lesson is simple: product design is only real if operations can execute it consistently.

8. The 2026 Lender Playbook: How to Build for Both Sides of the K

A practical blueprint for Thrivers

For Thrivers, focus on premium, frictionless, and relationship-enhancing products. Offer larger limits, faster funding, premium support, and value-added features such as rewards or cash-back offsets. Underwrite with strong traditional and alternative data, but keep the experience simple and fast. These borrowers will often accept a higher rate if the convenience and utility are obvious, but they will leave immediately if the value equation breaks.

Cross-sell opportunities matter more in this segment. If a borrower is financially healthy today, the lender should build a multi-product relationship through cards, installment options, savings-linked features, or embedded payment tools. The objective is to become the customer’s primary financial utility, not just a single loan provider.

A practical blueprint for Strivers

For Strivers, focus on stability, progress, and trust. Start with small, clear, affordable products. Use alternative data to improve approval precision, but keep pricing transparent and guardrails strong. Reward positive behavior with limit increases, rate reductions, or access to better terms. Make every step feel like progress, because progress is what sustains retention and reduces default.

This segment is where financial inclusion is won or lost. If lenders only see risk, they will decline borrowers who could have become loyal customers. If they see potential but ignore discipline, they will create losses. The winning formula is controlled access plus visible advancement.

A measurement framework lenders can use immediately

Lenders should track metrics that reflect both economics and inclusion. On the economics side: approval rate, loss rate, charge-off timing, return on risk-adjusted capital, and retention. On the inclusion side: approval lift for thin-file borrowers, graduation rate, payment success after hardship, and share of customers who move from starter to standard products. These metrics should be reviewed by segment, not only at the portfolio level.

When a segment underperforms, the question should be whether the underwriting is too loose, the pricing is too high, or the product is misaligned. That diagnostic discipline prevents knee-jerk policy changes that hurt growth. If you want more on how segmentation can reshape decisions, the logic parallels sector-based tailoring strategies: the more precisely you match the offer to the audience, the better the outcomes.

9. Risks, Compliance, and Governance in a Segmented Lending Model

Alternative data can improve inclusion—or increase bias

The same alternative data that helps a lender approve more responsibly can also introduce fairness problems if it proxies for protected traits or overweights digital visibility. Governance must therefore include bias testing, model monitoring, and documented rationale for each feature used. Lenders should also ensure adverse-action notices remain understandable and meaningful. If a borrower cannot tell what they could have done differently, the process is not truly transparent.

Regulatory expectations are moving toward clearer explainability and stronger consumer protections. This means lenders need cross-functional review among credit risk, legal, compliance, and product teams before launching new segmentation logic. A strong governance program is not a brake on innovation; it is what lets innovation survive scrutiny.

Stress-test the portfolio for K-shaped shocks

Portfolio management in 2026 must account for asymmetric stress. A K-shaped downturn can affect Strivers first and hardest, while Thrivers may remain resilient or even benefit from asset gains. Lenders should stress-test not only unemployment and rate shocks, but also inflation persistence, rent pressure, and income volatility across segments. Scenario planning should examine how a slowdown would affect approval quality, utilization, and delinquency timing.

This matters because many lenders are still optimized for average conditions. In a divergent economy, the average is misleading. The correct response is segment-specific shock modeling that tells you where losses will emerge first and where growth may still be available.

Governance should align product, risk, and customer outcomes

To stay responsible, lenders should define “good outcomes” at launch. For example, a starter loan could be successful if it helps a borrower build history, stay current, and graduate to a better product within a set timeframe. A premium product could be successful if it delivers retention and profitable usage without increasing complaint rates. Those definitions keep teams aligned and reduce the chance that sales incentives overpower customer wellbeing.

For a complementary view of how platforms should manage disclosure and risk language, see platform risk disclosures and reporting. The underlying message is consistent: trust is built when product truth matches customer expectation.

10. Comparison Table: Loan Design Choices by Segment

The table below summarizes how lenders can differentiate loan products, underwriting, and pricing for Thrivers and Strivers without compromising discipline.

11. A Step-by-Step Launch Plan for 2026

Step 1: Segment the market by financial trajectory

Start by building a segmentation model that incorporates score, cash flow, income stability, and volatility. Do not rely on one variable to decide product fit. Identify which borrowers are Thrivers, which are Strivers, and which are moving between the two. This trajectory view is more useful than static labels because it reflects where the borrower is headed, not just where they are today.

Step 2: Define a product ladder for each segment

Create tiered offerings with explicit rules for progression. For Strivers, start with low exposure and clear graduation criteria. For Thrivers, create premium tiers that monetize convenience and deepen relationships. Every tier should have a reason to exist and a clear economic model.

Step 3: Validate alternative data and explainability

Test each alternative data source for lift, fairness, and operational utility. Use only the signals that materially improve decisions. Ensure your adverse-action and customer communications can explain decisions in plain language. This step determines whether your expansion is sustainable or merely experimental.

Step 4: Align pricing with product purpose

Make sure rates and fees reinforce behavior you want, such as autopay, on-time repayment, or lower initial risk. Pricing should not be your only defense against loss. A well-structured product with good underwriting should reduce the need for aggressive pricing.

Step 5: Monitor cohorts and adjust quickly

Track each cohort’s delinquency, usage, and graduation performance. If a product is attracting the wrong borrowers, tighten qualification rules. If a segment is outperforming, consider cautious expansion. The important thing is to operate a living portfolio, not a static one.

12. Conclusion: The Best Lenders Will Design for Reality, Not Averages

The central lesson of 2026 is that lending cannot be built for a mythical average consumer. The market is split, and the split shows up in income volatility, asset accumulation, cash-flow resilience, and borrowing behavior. Thrivers and Strivers need different loan designs, different pricing logic, and different pathways to success. Lenders that adapt will gain share, improve risk-adjusted returns, and expand access responsibly. Those that cling to blunt underwriting and one-size-fits-all products will lose the best borrowers and overexpose themselves to the worst ones.

The opportunity is substantial: use alternative data to underwrite more accurately, build tiered products that reward performance, and create pricing and servicing structures that reflect actual borrower needs. That is how lenders can grow in a K-shaped economy without abandoning discipline. If you are building your next product roadmap, start with a segmentation model, map it to a laddered offer, and design the borrower journey with transparency at every step. For additional operational inspiration, review our guides on workflow automation, resilient document systems, and the latest K-shaped economy signals.

The main risk is misalignment. Thrivers may find the product too restrictive or too expensive relative to convenience, while Strivers may find it too fragile, too costly, or too easy to fall behind on. That leads to lower retention, worse delinquency, and missed growth opportunities.

Which alternative data source is most useful for underwriting in 2026?

Cash-flow data is often the most immediately useful because it helps measure income stability, recurring obligations, and payment capacity. Payroll, rent, and utility data can also be valuable, especially for thin-file borrowers. The best choice depends on the decision you are trying to improve.

How can lenders expand credit responsibly without increasing losses?

Use staged exposure, low initial limits, performance-based graduation, and hardship pathways. Expand only after borrowers prove repayment behavior. Monitor cohort performance closely and adjust pricing or qualification rules when risk changes.

Should lenders lower APRs for Strivers to improve inclusion?

Not automatically. APR should reflect expected loss and cost structure, but lenders can use smaller limits, shorter terms, fee discipline, and behavior-based discounts to improve affordability without underpricing risk. Inclusion should come from better product design, not hidden subsidy.

How do lenders keep alternative-data underwriting compliant?

They should document why each data source is used, test for predictive lift and bias, maintain explainability, and ensure adverse-action notices are meaningful. Compliance teams should be involved before launch, not after.

What is the best first step for a lender entering this strategy?

Start with segmentation. If you cannot clearly identify Thrivers, Strivers, and transition borrowers, product design and pricing will remain too blunt to be effective. Once segmentation is solid, build the tiered offer and underwriting rules around it.

Related Reading

• How to Integrate Buy Now, Pay Later (BNPL) Without Increasing Operational Risk - A practical guide to adding flexible credit features without weakening controls.
• What Platform Risk Disclosures Mean for Your Tax and Compliance Reporting - Learn how disclosure design affects trust, reporting, and customer understanding.
• Rewiring Ad Ops: Automation Patterns to Replace Manual IO Workflows - Useful patterns for lenders modernizing slow, manual back-office processes.
• Offline Workflow Libraries for Air-Gapped Teams: What to Store and Why - A resilience playbook that translates well to regulated financial operations.
• Build Your Own Training Analytics Pipeline: A Beginner’s Guide for Coaches and Enthusiasts - A helpful analogy for lenders building durable analytics and cohort monitoring systems.