The approval came through on a Tuesday and it felt like a win. Six engineering roles, unfrozen after nine months, with the original requisitions still sitting in the system ready to repost. We reposted four of them that week. Within a month, two had produced no shortlist at all, one had lost its best candidate at the offer stage over a salary band written the previous year, and the fourth had turned into an argument about whether the role was the same role. The approval was never the hard part. I had treated a restart as resuming something paused, and it is not that at all.
A freeze does not suspend an organisation. It reorganises one. Work redistributes, people absorb things quietly, interviewers get out of practice, salary bands drift away from the market, and the team forms private theories about what the freeze meant for their own prospects. Nine months later you are not resuming a hiring plan. You are launching one inside a company that has adapted to not hiring, and that is a different project with a different first step.
Why Restarts Fail at the Stages Nobody Watches
The failures are predictable and they cluster. Almost none of them are sourcing failures, which is where everyone looks first.
They happen at the offer stage, because the band is stale. They happen at the panel stage, because interviewers who have not scored a candidate in nine months disagree about the bar without realising they disagree. They happen internally, when an engineer who covered two roles for three quarters learns about new headcount from a job advertisement. And they happen at the specification stage, where the frozen requisition describes a role that the surviving team has already partially absorbed, so the new hire arrives into a job that is two thirds done and nobody can say which third is left.
That last one is the most expensive and the least discussed. A frozen requisition is a photograph of an organisation that no longer exists. Reposting it does not get you the engineer you needed in February. It gets you an engineer shaped like a hole that has partly closed.
The 7 Steps, in Order
Step 1: Re-derive the headcount from the work, not the old plan
Close the old requisitions. Not archive, close, so nobody reposts them while you are thinking. Then rebuild the list from a single question asked of each team lead: what is the work in front of you now, and which part of it has no owner.
The answers will not match the frozen plan, and the mismatches are the useful output. Some roles have genuinely closed because someone grew into them, and promoting that person is cheaper and faster than hiring. Some have grown, because a frozen role becomes two jobs when a team covers it badly for three quarters. And a few will be new, usually operational roles created by the workarounds the team invented to survive.
Expect this to reduce your number. Ours went from six to four plus two internal moves, which felt like a loss at the time and was the single best decision in the sequence. Filling four roles well beats half-filling six. The arithmetic for this sits in our method for engineering capacity planning in Singapore.
Step 2: Reopen one role, not all of them
Run one search from posting to offer before approving the rest. Pick a role that is representative rather than the easiest or the most urgent, because you are using it as an instrument.
What that single search tells you is worth more than the time it costs. It tells you what your employer brand looks like after a public hiring gap, measured in response rate rather than in opinion. It tells you how long your loop actually takes now, which will be longer than you remember. It tells you which stage is broken, and it does so once rather than six times in parallel with six hiring managers each forming a different theory.
The objection to this is always urgency: we cannot afford to hire sequentially. In my experience the opposite is true after a freeze. Six simultaneous searches on an unrepaired process produce six slow searches and a demoralised panel, and the calendar cost of fixing things mid-flight exceeds the cost of one diagnostic search up front.
Step 3: Rebuild the panel before the pipeline
Interviewers decay faster than job descriptions and much more quietly. Someone who interviewed confidently nine months ago has lost the calibration, forgotten the debrief discipline and, in a third of cases, no longer wants to be on the panel at all but has not been asked.
Three things to do before a candidate reaches the loop. Ask every interviewer whether they still want the duty, and accept the answer without negotiation, because a reluctant interviewer produces slow debriefs and polite non-signals. Re-run calibration against decisions you have already closed, where the outcome is known, which surfaces drift far faster than discussing hypothetical candidates. And gate anyone who has not scored in six months behind one shadow session, which is cheap and which nobody resents when it applies to everybody.
The levelling conversation belongs here too, before offers rather than during them. If your bands or levels shifted during the freeze, the panel needs to know what a given level means today. That ladder work sits in our guide to engineering career ladders and promotion levels in Singapore.
Step 4: Tell the internal story before the external one
Announce the restart to the team before the first advertisement is live. This is the step that gets skipped because it feels like communications rather than hiring, and it is the one with the sharpest downside.
Consider how it looks from inside. An engineer has spent three quarters covering work that belonged to a role that was frozen. They have been told, implicitly, that there is no budget. Then a job advertisement appears for something adjacent to what they have been doing. The inference is immediate and unflattering: there was budget, and it was not for me.
Say four things, in this order. Hiring is restarting. Here is what changed to allow it. Here is what it means for the people who absorbed the extra work, specifically, including any internal moves from step one. And here is what we are not doing, which matters because people fill silence with the worst available interpretation. Retention risk peaks in the fortnight after a restart becomes visible, not during the freeze, and the mechanics of that are in our developer retention method.
Our expert take
If I could only keep one step from this list it would be this one, and it is the step with no recruiting work in it at all. The most expensive thing that happens during a hiring restart is not a failed search. It is a senior engineer who held the team together through the freeze deciding, in the week your careers page refreshes, that the loyalty was not mutual. That resignation costs more than the three hires you were trying to make, it arrives two months later so nobody connects it to the restart, and the entire prevention is one honest conversation held before the advertisement rather than after it.
Restarting engineering hiring in Singapore
Send us the roles you have just had approved and when the freeze started. We will tell you which requisitions no longer describe a real job, and which one to run first as a diagnostic.
Get startedStep 5: Re-price against today, not against the band you froze
This is the step that cost us a finalist, so I will be direct about it. A salary band written before a freeze is not a conservative starting point. It is a wrong number that feels authoritative because it was approved.
Rebuild the range from three sources, in descending order of reliability. First, offers you have actually closed or lost in the last few months, including the ones that got away and what they went for, which is the only hard evidence you own. Second, what candidates tell you unprompted in first conversations, which is softer but current. Third, published benchmarks, which lag and which flatten exactly the specialist roles you are most likely to be hiring.
Then do the unglamorous part: get the revised range approved before the first candidate reaches the final stage, not when an offer is being drafted. A band that has to be re-approved during an offer conversation adds a week, and a week at that point is where candidates accept other offers. The approach to the offer itself is in our method for improving developer offer acceptance rate in Singapore.
Step 6: Sequence so each hire makes the next one easier
Once the diagnostic search has closed and the process is repaired, order the remaining roles deliberately. The intuitive order is hardest first, on the theory that the difficult search needs the most runway. I think that is usually wrong after a freeze.
Order by how much each hire reduces load on the people running the loop and unlocks work the later hires depend on. A hard senior search consumes enormous panel capacity for months, and if it is first it starves everything behind it at exactly the moment your interviewers are least practised. Put one or two roles ahead of it that close faster, relieve the people who have been covering, and give the panel repetitions on lower-stakes decisions.
Write the sequence down with rough dates and share it with the hiring managers, because the thing it prevents is the quiet reopening of role five while role two is still running. Someone will try. The written sequence is what makes declining that an administrative matter rather than a political one.
Step 7: Name the condition that would freeze you again
Write down the specific metric and threshold that would pause hiring, and share it with the hiring managers and, in substance, with candidates who ask.
This feels like inviting bad luck and does the opposite. A restart with no stated stopping condition produces hedging at every level: managers who do not fully commit a panel because they suspect the plan is provisional, recruiters who hedge with candidates, candidates who sense the hedge and keep other processes warm. Naming the condition converts an ambient fear into a known boundary, and everyone commits inside that boundary.
Make it a number, not a mood. A revenue threshold, a runway figure in months, a specific contract outcome. If nobody can articulate what would stop hiring again, that is worth discovering in a planning meeting rather than in week five of three live searches.
The Two Things I Got Wrong
The first was reposting. Four requisitions went back up in week one because they existed and because reposting felt like momentum. Two of them described roles that had partly closed, and the resulting conversations with candidates were genuinely embarrassing: we could not answer clearly what the person would own, because internally we had not agreed. If you take one thing from this, close the old requisitions before you think about any of the rest.
The second was treating the stale salary band as the safe option. I went to market with last year numbers because they were approved and because asking for more felt like bad timing after a freeze. We lost a finalist in week five, re-approved a higher band in week seven, and filled the role in week fourteen at slightly above what the revised band would have been. The delay cost more than the increase, which is the general shape of this mistake.
When Not to Restart Yet
Three situations where I would wait, and waiting deliberately is different from being frozen.
- The stopping condition from step seven is already close. If the metric that would pause hiring is within a quarter of triggering, you are about to run searches you may have to abandon. Withdrawing an offer or pausing a candidate at final stage does lasting damage to a reputation in a market as small as this one. Wait for clearance, or hire one role rather than four.
- The team structure is still being decided. Hiring into an organisation whose reporting lines change next month produces roles that nobody owns on arrival. Settle the structure, then specify the roles against it.
- Nobody has the capacity to interview. If the people who would staff the panel are the same people covering the gaps the freeze created, you do not have a hiring process, you have a plan to overload your most loaded people. Relieve them first, with an internal move or a contractor, then open the search.
If what you are rebuilding is the platform alongside the team, the scoping side of that is in our guide to custom software development in Singapore.
Our expert take
The reason the one-role rule in step two meets so much resistance is that a restart approval arrives with political energy behind it, and spending that energy on a single search feels wasteful. I would argue the opposite: the approval is the scarcest thing you have, and the fastest way to lose it is three visibly stalled searches in month two. One search that closes well produces a second approval conversation you walk into with evidence. Four searches that limp produce a review, and reviews after a freeze tend to end the way the freeze did.
Frequently Asked Questions
How long does it take to restart hiring after a long freeze?
Should we rehire the people we let go or who left during the freeze?
Do we need to tell candidates there was a freeze?
What if the freeze was never formally announced and just happened?
Opening engineering roles again in Singapore
We screen developers for Singapore employers and we run the diagnostic search with you, including the part that is no fun: telling you which of your newly approved roles is not a real job yet.
Get startedRelated reading: Engineering backfill plan · Improving offer acceptance rate · Developer retention
