Every BES (Bireysel Emeklilik Sistemi) provider in Turkey has the same dirty secret in their database: a long tail of participants nobody can actually contact anymore. The phone number is dead. The address points to a building that was demolished in 2014. The email bounces. The TCKN is still valid, the contributions are still sitting there, and the regulator still expects you to do something specific with that account at specific times.
This is not an edge case. After a portfolio runs for ten or fifteen years, the dormant tail becomes structurally significant — sometimes 8-12% of accounts depending on cohort. And the data infrastructure most companies stood up between 2013 and 2018 was never designed for participants who stopped being participants in any practical sense but didn't legally cease to be one.
What "Dormant" Actually Means in the Pipeline
Regulatory dormancy and operational dormancy are not the same thing, and this is where the first cracks appear.
- Regulatory dormancy: defined by contribution behavior, account age, and specific SEDDK/EGM thresholds. Triggers reporting obligations and eventual disposition workflows.
- Operational dormancy: the participant is unreachable through any channel you have on file. No bounce-back loop closes.
- Legal dormancy: the participant may be deceased, legally incapacitated, or have undergone a name change (marriage, court order) that your KYC data hasn't caught.
Most pipelines were built around regulatory dormancy because that's what produces reports. The other two get discovered accidentally, usually when a campaign goes out and 14% of SMS messages fail, or when the call center spends three weeks trying to reach someone who died in 2019.
Where the Pipeline Actually Breaks
The failure modes are not bugs. They are consequences of assumptions baked into the original schema.
Contact information stored as current-state, not as history. When a participant updates their phone number, the old one gets overwritten. So when the new number stops working two years later, you have no fallback. The MERNIS integration helps for addresses, but only if you actually run the reconciliation — and most teams run it quarterly at best, monthly only under audit pressure.
Name changes break joins downstream. A woman marries, updates her name in MERNIS, your KYC refresh picks it up, but the document management system still has files indexed under the old name. The CRM has campaign history under the old name. The complaint system has tickets under the old name. The participant ID holds it all together until somebody runs a manual query joining on name, which still happens more than anyone wants to admit.
Soft-delete patterns that aren't actually soft. Some legacy systems mark accounts as inactive in ways that exclude them from default queries. The dormant account exists, the regulator expects it in the report, but the standard dashboard query filters it out. Discovery happens during reconciliation against the EGM transfer files, which is the worst possible time.
Heir and beneficiary data is almost always worse than participant data. If the participant is deceased and you need to contact heirs, the contact data for heirs was captured once, at enrollment, and never refreshed. It's typically 5-15 years stale.
The Compliance Layer Sitting On Top
Under Turkish regulation, you cannot simply let dormant accounts sit. There are notification obligations, transfer obligations to the state pension fund after defined periods, and reporting that distinguishes between contactable and uncontactable participants. The regulator increasingly expects you to demonstrate that you tried — and "tried" means an auditable trail of contact attempts across multiple channels with documented outcomes.
This is where the data quality problem becomes a legal liability problem. If your pipeline cannot produce, on demand, the full contact history for a specific participant including which channels were attempted, when, and what the result was, you cannot defend the disposition decision. I have seen companies discover during audit prep that their SMS gateway logs were retained for 90 days, their email logs for 180, and their call center notes for one year — meaning the audit trail for an account that went dormant in 2021 was structurally incomplete by 2023.
What Actually Has to Change
This is not solved by buying a data quality tool. The fixes are architectural.
- Treat contact information as event-sourced, not state-based. Every change is an append. The current phone number is a projection, not the truth. Old numbers remain queryable, because sometimes the old one works when the new one doesn't.
- Run MERNIS and KPS reconciliation on a defined cadence and log the deltas. Monthly minimum. The delta log is itself an audit artifact.
- Separate participant identity from participant contactability in the schema. A participant can exist and be uncontactable. That should be a first-class state, not an inference from failed campaigns.
- Retain contact attempt outcomes for the full account lifetime plus the regulatory retention period. Not the SMS gateway's default. Pull the outcomes into your own warehouse and keep them.
- Build the dormancy classification as a derived attribute computed nightly, with versioned logic. When the regulator changes the definition (and they will), you need to be able to recompute history under the new rule without losing the old classification.
The Uncomfortable Part
Most of this work is invisible until it isn't. Nobody gets promoted for building a robust dormant account pipeline. The project gets funded after an audit finding, a regulator letter, or a court case where a beneficiary's lawyer demonstrates that the company couldn't prove it tried to contact anyone.
The participants who disappear don't disappear from the balance sheet. They sit there, accumulating returns, accumulating regulatory obligations, accumulating risk. The data infrastructure either knows where they went or it doesn't — and if it doesn't, that's not an operations problem to be solved next quarter. It's a structural debt that compounds at exactly the rate the portfolio grows.