Two years after the two-pot retirement system took effect on 1 September 2024, the headline numbers offer a measure of comfort: two-thirds of new contributions are locked into the retirement pot until later life, and 54% of Discovery retirement fund members have not touched their savings pot at all since launch. Strip away the aggregate, though, and the trajectory is uncomfortable. Discovery data presented by Guy Chennells, Chief Commercial Officer for Corporate and Employee Benefits at Discovery, at the Institute of Retirement Funds Africa (Irfa) annual conference shows that 43% of eligible members made a withdrawal in 2025, climbing to 46% in 2026, while the average amount available for withdrawal dropped from 36% to 27% over the same period. The savings pot is being drawn down faster than it refills, and the drivers reach well beyond ordinary financial need.

That gap between what the legislation preserves on paper and what members actually retain at retirement is now the central challenge for anyone advising on, designing, or receiving a retirement benefit in South Africa. The cause is not a simple failure of self-discipline. Financial stress, mental-health strain, and a measurable rise in gambling behaviour are all part of a picture the system was never designed to address on its own.

What the two-pot system actually does

The two-pot structure was enabled through amendments to the Pension Funds Act 24 of 1956 and aligned changes to the Income Tax Act 58 of 1962, taking effect on 1 September 2024. Under those rules, two-thirds of new contributions flow into a retirement pot that remains inaccessible until retirement. The remaining third goes into a savings pot that members can access once per tax year, with the withdrawal taxed at the individual's marginal rate by the South African Revenue Service (SARS).

The design addressed a structural flaw that had eroded retirement adequacy for decades. Under the previous regime, leaving an employer typically forced a binary choice: preserve the full fund value or take the cash. Nic Whittles, senior employee benefits consultant at BDO Wealth, described the old model this way:

The previous system placed the responsibility for preservation almost entirely on the individual. Every time someone changed jobs, they had to decide whether to preserve their retirement savings or take the money. Many people cashed in and effectively started their retirement savings journey again.

The two-pot structure removes that either-or moment at every job change. The subtler risk it introduces is different: once members know the savings pot exists and is legally accessible, the pull to treat it as a permanent emergency reserve is difficult to contain, particularly when household budgets are already stretched.

Withdrawals are climbing, not stabilising

Since the September 2024 launch there have been three annual withdrawal windows, and a small but consistent segment of Discovery members has exercised all three. The Irfa conference data shows withdrawal rates rising year on year while the available pool shrinks, a combination that points toward accelerating depletion rather than the measured, occasional use the architects of the system anticipated.

The 2025 Sanlam Benchmark Survey provides the household-level context. It found 44% of retirement-fund members had drawn on emergency savings to meet regular expenses, and 43% had no emergency fund at all. For those households, the savings pot is not a top-up or a buffer: it is the emergency fund, by default, and it is being used accordingly.

There is a second layer that the financial-literacy framing does not reach. Discovery data indicates that individuals dealing with poor mental health, including anxiety, depression, and sleep disturbances, are demonstrably more likely to access their retirement savings via the savings pot. Alongside that finding, gambling activity across South Africa has risen 226% since 2019, with online gambling spend now comparable to levels recorded in countries with significantly higher per-capita incomes. A portion of the money leaving retirement pots is not covering utility arrears or school fees. It is being lost to gambling platforms, a dynamic that no pension-fund communication strategy was built to intercept.

For employees: what a withdrawal really costs

If a withdrawal feels like a practical solution in the moment, two cost layers are worth separating before committing. The first is the SARS tax charge, visible and immediate: the withdrawn amount is added to taxable income for that year and taxed at the member's marginal rate under the Income Tax Act 58 of 1962. For members in higher brackets, SARS can reclaim 30% to 45% of the withdrawal before it reaches a bank account.

The second cost is invisible in the moment but compounds over decades. BDO Wealth ran a worked illustration on a member earning R35,000 per month contributing 10% of salary. Left fully invested, the savings pot reaches roughly R717,000 after 10 years and R2.58 million after 20 years. Withdrawing R21,000 from that pot every two years pulls the 10-year balance down to approximately R585,000 and the 20-year balance to around R2.1 million. That is a R132,000 shortfall at the 10-year mark and a R481,000 shortfall at 20 years, produced by what feels, at the time, like a modest top-up. Whittles frames the arithmetic plainly:

When money remains invested over 20, 30 or 40 years, compound growth can have a dramatic impact on retirement outcomes. Every withdrawal reduces not only the amount invested, but also the future growth that money could have generated.

Before drawing on the savings pot, three practical checks are worth running:

• Model the SARS tax hit first. Use SARS's marginal-rate tables for the current tax year to calculate what arrives in your account after tax. For many members the net amount is substantially lower than the gross withdrawal, which changes whether the withdrawal actually covers the need.
• Check whether a lower-cost route exists. A short-term credit facility, a negotiated payment plan with a creditor, or deferring a non-essential purchase often carries a lower 20-year cost than a retirement withdrawal. The comparison is worth doing explicitly.
• Ask whether the pressure is financial or emotional. If the impulse to withdraw coincides with sleep problems, persistent anxiety, or gambling losses, the withdrawal will not address the root cause. A conversation with a financial planner or a medical professional is a more productive first step than a transaction.

The savings pot is a legitimate safety net built into the Pension Funds Act 24 of 1956 for exactly this reason. It works best when used rarely, deliberately, and with a clear view of what it removes from a future retirement income.

For HR and benefits leaders: the communication gap is the retention gap

The 2025 Sanlam Benchmark Survey contains one figure that reframes the employer's responsibility: 68% of retirement-fund members said they would contribute more toward retirement savings if they could afford to do so. This is not a workforce that is indifferent to its own financial future. It is a workforce under sustained financial pressure that is reaching for the only pool it can legally access without leaving its employer.

On awareness, the picture is genuinely positive. Sanlam recorded a jump from 59% in 2024 to 92% in 2025 on system awareness, and more than 90% of members indicated that retirement-benefit education and communication matter to them. Raazia Khan, head of benefits at NMG, has noted that the reform has meaningfully improved dialogue between retirement funds and their members and helped surface unclaimed benefits. Nancy Andrews, chair of Irfa, has credited the legislation with preserving substantially more retirement capital than the previous regime allowed.

Awareness, then, is not the gap. Decision support at the point of withdrawal is. Whittles puts it directly:

Members need support when they are making these decisions. That conversation should consider their immediate needs, the tax implications and the long-term impact on retirement savings. That's where employers, benefit consultants, and financial planners can make a meaningful difference.

Where the current benefits playbook falls short

Most retirement communication, including what the South African Board for People Practices (SABPP) has long recommended under its retirement benefit communication standards, treats the withdrawal decision primarily as a financial-literacy question. The Discovery mental-health correlation suggests that framing misses a substantial share of the members most likely to withdraw. Annual pension statements and occasional webinars cannot reach a worker whose withdrawal decision is driven by a depressive episode, chronic insomnia, or a gambling loss that needs covering before month-end. The Financial Sector Conduct Authority (FSCA), which oversees retirement fund governance and member communication standards, has not yet issued specific guidance on integrating mental-health pathways into two-pot member communication, leaving that design choice to individual funds and their employer clients.

What to bundle into benefits communication now

• Place a decision-support option at the withdrawal touchpoint. When a member initiates a savings-pot withdrawal request, the fund portal should surface an offered 15-minute call with a benefits consultant before the transaction completes. Not a mandatory gate, but a visible, friction-reducing second opinion that costs nothing to take.
• Route mental-health support through the same channel as financial wellness. Members most at risk of withdrawing are also those least likely to self-refer to an employee assistance programme. A benefits message that mentions both in the same communication removes the stigma of navigating them separately and reduces the number of steps between distress and help.
• Personalise the compound-growth calculation. The R132,000 and R481,000 figures from the BDO Wealth illustration land very differently when the calculator uses the member's actual salary and contribution rate rather than a generic example. Generic numbers invite disengagement; personalised projections prompt reconsideration.
• Track the right outcome metrics. Withdrawal rate is a lagging indicator. More informative leading indicators include: how many members used a decision-support option before completing a withdrawal, how many revised the withdrawal amount after a consultation, and how many accessed mental-health support in the same quarter as a withdrawal request. These tell you whether the programme is functioning or merely existing.

For HR teams, the shift carries a strategic dimension beyond compliance. A benefits programme that actively helps members preserve their savings pot becomes a credible element of an employer value proposition and a retention lever. A programme that records rising withdrawal rates quarter on quarter without an intervention response becomes a reputational liability, particularly as employee review platforms make benefits quality visible to candidates before they apply.

What year three should look like

The first two years of the two-pot system demonstrated that the structural design works. Capital is being preserved that would previously have leaked out at every job change, and Irfa's Nancy Andrews and Sanlam's own data both confirm that preservation levels exceed those recorded under the old single-pot regime. The question for the next phase is whether employers, benefit consultants, retirement funds, and the FSCA can translate that structural gain into individual retirement adequacy at scale.

Whittles offers a calibration that benefits professionals across South Africa should carry into their 2026-27 planning cycles:

Two years in, we shouldn't look only at what has been withdrawn. We should also look at what is now being preserved. That is where the long-term success of the two-pot system will ultimately be measured.

One structural caveat deserves naming. The Discovery and Sanlam datasets both skew heavily toward formally employed, middle-income members. Workers on or near the National Minimum Wage (currently R27.58 per hour under the National Minimum Wage Act 9 of 2018), workers in precarious or informal employment, and union members in COSATU-affiliated funds whose benefit structures may differ, are largely absent from these figures. Their withdrawal patterns are almost certainly harder than the middle-income picture, and the data to confirm or challenge that assessment is not yet publicly available from SARS or the FSCA. For employers who do have line of sight into their own workforce demographics, acting on what is visible is more productive than waiting for aggregate industry figures that may not arrive for another two reporting cycles.