Meet the Expert: Farah Sidiqui, Head of Analytics-Supply Chain, InsuraLogic

With two decades in analytics-driven logistics for insurers, Farah Sidiqui has scaled platforms through hurricanes, regulatory overhauls, and two major M&A integrations. She’s seen insurance supply-chains move from asset-heavy models to networks that “curate and orchestrate” risk with third parties—while compliance, fraud, and margin pressure are all moving targets. Her team’s focus: make real-world decisions anchored in data, not dogma.

Q1: Where do senior supply-chain teams in insurance get risk assessment wrong, especially when analytics should be front and center?

Farah: The classic misstep is treating risk assessment as a one-time compliance hurdle, not a dynamic hypothesis continually tested by data. In legacy insurance, risk frameworks were about document boxes—underwriting checklists, audit trails, supplier certificates. But in analytics-first teams, you need live feedback loops.

Take catastrophe modeling. Old-school: buy a flood model, get an Excel, set deductibles, move on. New-school: ingest new rainfall sensor APIs daily, monitor claim lags, A/B test vendor dispatch protocols, and adapt. If your risk model doesn’t change after a major hail event or a tweak in claims triage logic, you are not data-driven; you’re gambling.

A Forrester report from March 2024 found that only 28% of insurance supply-chain execs had risk models updated within a week of significant external events. That gap between event and model update is where losses multiply.

Q2: You mentioned the shift from “ownership” to “experience” in supply-chains. Can you unpack how that changes risk frameworks?

Farah: Absolutely. “Ownership” meant you procured, warehoused, and managed every asset—vehicles, contractors, FNOL call centers. With “experience,” you orchestrate third-party providers, on-demand assessors, even AI triage bots.

Risk gets more distributed—and sneakier. For example, when relying on gig-work adjusters, your risk exposure isn’t only fraud or SLA misses. It’s also model drift as claim severity mixes shift, or soft fraud rising when your data signals lag behind reality.

The risk framework now needs to ingest not just what you control, but what your partners’ data shows in near real-time. That means API integrations, event streaming, and robust anomaly detection on fields like “outlier claim duration per provider.” If your data only covers assets you own, you’ll miss systemic exposure building in partner networks. The move from ownership to experience doesn’t just redistribute risk—it obfuscates it, unless your analytics platform is built for external telemetry.

Gotcha

Don’t treat third-party risk feeds as “nice-to-have” extras. I’ve seen a claims vendor with a 98% SLA suddenly dip below 85% for a specific region, flagged only because we piped their operational logs directly into our risk dashboard. Without that, we would have missed a 3-day claims delay after a local server outage, costing six figures in regulatory penalties.

Q3: What’s the most underutilized data source in insurance supply-chain risk today?

Farah: Feedback loops from field operations. Everyone captures transactional data—delivery times, adjuster assignments, claim close rates. But very few senior teams systematically collect and analyze agent or customer feedback at the event level, correlated with supply-chain decisions.

For instance, after Hurricane Ian, we used Zigpoll and Medallia to pulse field adjusters and insureds after big claims. One insight: adjusters flagged certain restoration partners as “often unreachable post-dispatch,” a data point never flagged by our default vendor scorecards. Quantifying that (18% unreachable rate in the first 72 hours) and tying it to payout delays let us reroute $5M of restoration work and improved net promoter score by 2.9 points quarter-on-quarter.

Table: Comparing Feedback Data Sources

Edge Case

If you roll out feedback tools after major events, you’ll get survivor bias: only the least-impacted (or most frustrated) respond. To avoid this, embed feedback requests directly in claim closure or supply dispatch workflows, not as one-off surveys.

Q4: How do you balance quant-driven experimentation with regulatory risk in insurance supply-chains?

Farah: Carefully—and with audit-ready documentation at every step. Insurers, especially in catastrophe response, want to A/B test supply allocation: which vendor for which peril, what sequence of actions, which digital triage trigger.

But regulators want explainability, not just effectiveness. Our approach: any experiment gets a “risk protocol” doc—what’s being changed, what’s the fallback if KPIs go off-track, and what’s the audit trail.

For example, when we piloted dynamic vendor pricing after storm surges, every price change event was logged, and regulators (in two states) got weekly reports summarizing the experiment. If claim costs spiked abnormally, we could roll back the change and trace every impacted policyholder.

Optimization Tip

Automate flagging of non-conforming events (e.g., price changes outside regulatory bands) using real-time rules in your analytics stack. Manual reviews after the fact leave you exposed. We use Azure Sentinel with custom logic for this.

Limitation

This approach slows iteration speed. If you’re used to pure tech A/B testing, get ready for paperwork. In regulated insurance, compliance is part of your experimentation toolkit—it’s not optional.

Q5: How do you quantify and monitor model drift in risk assessment frameworks, especially given constantly changing supply chain data?

Farah: Model drift is a silent killer—especially as claim types, vendor networks, or fraud vectors morph post-event. First, baseline your models with reference periods (e.g., Q1 2023 hurricane claims), then set up real-time statistical drift detectors on both input data and top-level risk outputs.

A practical tactic: threshold-based triggers. For instance, if average claim closure time shifts by >15% vs rolling 60-day norm and the distribution of assigned vendors changes by >10%, trigger a mandatory review.

Example

After we moved to app-based adjuster dispatch in Texas, our “days to first contact” dropped from 5.2 to 2.4 days—but denial rates tripled for water claims. Only by monitoring both outcome and input drift did we spot a bug in the new app’s question tree that filtered out valid claims. Fixing that required both tech and workflow changes.

Edge Case

Drift isn’t always bad. Sometimes, an abrupt shift is an intended business outcome (like a new triage protocol reducing cycle times). The trick: separating desired drift from undesired drift, and documenting each change's intent.

Q6: What’s overlooked in managing risk when integrating with new analytics vendors or SaaS platforms?

Farah: Integration risk is always underestimated. Senior leaders often assume an API contract means “data will flow and be correct.” But real-world supply-chains see schema mismatches, laggy updates, edge-case failures.

Always build a “gray-box” testing environment. Don’t just test the happy path. Simulate delayed webhook responses, missing mandatory fields, or out-of-band data types. We once lost visibility into a week’s worth of claims because a vendor sent “null” as a string instead of an empty field—it bypassed our validation, leading to a reporting black hole and a $60K reserve error.

Table: Integration Risks and Mitigations

Gotcha

Don’t assume that “successful” API calls mean correct data. Build downstream reconciliation checks—compare claim counts, or total payouts per day, between systems.

Q7: Actionable advice—what should senior supply-chain teams do differently next quarter, as they rethink risk frameworks?

Farah: Three things:

1. Shorten feedback cycles: Push for weekly or even daily model updates after major events—don’t wait for month-end close.
2. Instrument partner data: If your analytics stack isn’t pulling in third-party operational logs, fix that gap now.
3. Automate drift detection: Don’t wait for a spike in bad outcomes; configure triggers and alerts for input and outcome drift.

Not every framework or tool will work for every insurer. If your supply-chain is still asset-heavy (own fleets, in-house staff), you’ll need a different emphasis than if you're orchestrating networks of partners and SaaS tools. One framework won’t fit all. But any senior team can get sharper by making risk frameworks not just a compliance checkbox—but a living, data-driven experiment.

Farah’s final word: The fundamental shift is this: risk in insurance supply-chains is now an analytics challenge, not just an operational obligation. Experience over ownership means you don't control every variable—but you can instrument every variable. Senior leaders who make that mindset and tooling shift will stop guessing, and start managing risk with eyes wide open.