Is this you
You are about to post the same seat again, and you have already had this argument with yourself once.
There are two versions of you doing that, and from the outside they are identical. In one, the first search did not fill, or it filled and did not hold, and the advertisement is going back up because nothing else has changed. In the other, the first hire worked, the volume moved, and the seat is coming back because the business needs a second one.
If this is your tenth hire and not your first, the trap runs the other way to a novice's. You have a job advertisement that has worked before. It is the fastest thing on your desk, it takes almost no time to put back up, and reposting it feels like action when it is a decision.
The calls are full of the first version. One principal had hired out of desperation, in those words, and put it down to rushing. Another watched a settlement officer seat run most of a year and take more than one advertisement before it filled. She is the one quoted in the next section. The second version turns up in the advertisements rather than in the calls, which is where section 05 goes to find it.
The uncomfortable one is the principal who was asked what the empty seat was costing him and answered that it cost roughly nothing. The work still got done. Nobody was waiting. He wanted the person in place before the volume arrived, and until it did, nothing in the business forced the search to close. That is why so many second searches run long, and it has nothing to do with the advertisement.
FinTalent client call records, 2025 to 2026
Well, the latest settlement officer we hired ... turned out perfect, which is an absolute catch. But it took eight months to get there. It took two, three, two people we hired and didn't and fell through to finally getting through the third advert ...
One signal, two opposite businesses
A repeat advertisement is one fact with two readings, and the fact does not carry the reading. That is the whole problem. Everyone who sees it is guessing at which one you are, including us, and including you when you look at somebody else's.
The tell most principals expect is money, and it runs the wrong way. Across the whole window, 524 of 991 advertisements name a dollar figure, a little over half. Among the 144 that are a seat coming back it is 53, closer to two in five. So a second posting tells a reader less about the pay than a first one does, not more, and 34 of the 65 seats never named a salary at either attempt. If the plan for the second attempt is to compete on money, most of the field you are competing with is not showing theirs either.
What the rewriting actually moves is the pool. One advisory ran a broker seat repeatedly and, across the versions, the experience ask widens from an experienced mortgage broker to an experienced mortgage broker or credit assessor, a band appears where none had been named, and a line about a flexible culture becomes flexible working options. One of those postings runs from a second office interstate and not from the Parramatta one, which is a different answer again. What none of it records is whether the seat then filled. The advertisement shows you what the employer changed their mind about, and stops there.
And then the part nobody can sell you. The frozen set carries no posting date, so nothing in it knows which version came first. Every pairing in this guide is written as one version and the other. You are the only person who knows the order you posted yours in, which makes this diagnosis yours and not something a signal can do for you.
BD Machine ad index and FinTalent client call records, 2025 to 2026
What re-advertising looks like from outside
- 144 of 991
- advertisements were a seat coming back
- 65 seats, keyed on the employer plus the job title with both folded for case and spacing. The filter admits a pairing seen two to five times, and in practice no seat in this set went past four postings. Recruitment agencies and franchise brands are excluded, and both exclusions carry weight: an agency advertises other people's seats, and one franchise brand advertised the same title at a different band each time, which is several franchisee offices and not one seat coming back. Before those exclusions the raw figure is 97 repeat pairs, so 32 of them were noise.
- BD Machine ad index, 991 ads, 23 Feb to 20 Aug 2026, re-advertising derivation
- 26 of 65
- ran the advertisement again word for word
- The same string, once case and spacing are folded. No judgement call and no similarity threshold sits inside that figure, which is why it is the one comparison here worth printing. They are not the vague ones and they are not the thin ones. They name a salary at about the rate the rest of the population does, and some are precise about what the seat pays. Whatever the second posting was, for these it was not a revision.
- BD Machine ad index, re-advertising derivation. Comparison on the advertisement body, not on the salary field
- 53 of 144
- of the repeat advertisements named a figure
- Against 524 of 991 across the whole window, so a seat coming back is less likely to tell you what it pays than a seat being advertised for the first time. That is a fact about how much the reader is told, not about whether the employer is cheap, and this guide does not have the second one. The comparison is advertisement against advertisement on a single salary test, because a seat count and an advertisement count are not the same measure and reading one against the other is how you get the answer backwards.
- BD Machine ad index, 144 re-advertised ads against 991 in the window, one salary test applied to both
- No dates
- so the gap between the two attempts is unmeasurable
- The figure this guide would most like to give you. The frozen set carries no posting date, so nothing here records how long there was between postings or even which came first, and the rows sort by identifier. Anyone quoting you a number of days for a re-advertised seat is estimating, and an earlier median that circulated came from a different population and does not survive into this one.
- BD Machine ad index. date_posted is not among the frozen columns
The fork
The seat did not fill
- Then the market has already answered this advertisement. Posting it again unchanged asks the same question of the same people, and two in five of the seats in this population did exactly that.
- Read what the advertisement says the job is before you touch the money. One North Sydney brokerage ran a senior credit analyst seat two ways. One version sells it as a senior person's right hand, saving them time and streamlining communication. The other says outright that it is not an admin support role, and leads on lending policy and real autonomy on deal structuring. Neither version names a salary, so nothing about that pair is about pay.
- The pool is the fix more often than the price is. The Parramatta advisory's experience ask widens from a mortgage broker to a mortgage broker or credit assessor between versions. Same seat, a wider pool. If a seat is not filling, that line is cheaper to change than your salary band and it changes who ever sees the advertisement.
- Change one thing you can name. A challenger SME bank ran a senior commercial seat three ways, and the only difference across them is a single clause about experience on very large transactions, carried in one version and absent from the others. One variable, so whatever happens next, they can read it. Rewrite everything at once and the third attempt teaches you nothing either.
The seat filled and the business needs another
- Then say so in the opening line, because it changes who applies. One Brisbane brokerage opens by saying it wants more brokers than it has, and that it has more leads than brokers to service them. It ran twice, unchanged, in one city. Nothing about that reads as a failed search, and a candidate can tell.
- Treat a new market as a new search and not as a repeat. A national broking group ran the same advertisement in three cities, and its own opening line names most of the mainland capitals it is hiring in. The local pool, the pay and the competition all differ between them, and the advertisement that carried one is not automatically the right one for the next.
- Check which seat the words were written for. One regional Victorian group posted the same job title in two towns at opposite ends of the state. The two advertisements are not reposts of each other, they are genuinely different documents, and only one is built as an entry pathway with training and mentoring provided. That is the right way round. The failure is the other way, when a second office inherits the words written for the first and quietly inherits the seniority with them.
- Name who onboards them before you post. A second chair costs a desk, a licence, a system seat and somebody's attention for a quarter. If the answer to who trains them is the person who trained the first one, and that person is already at capacity, the advertisement is not the piece of work in front of you.
BD Machine ad index, re-advertising derivation, and FinTalent placement experience
Work out your own answer
Twenty minutes with both advertisements open, and you start with the one thing no dataset has: you know which of them you posted first.
-
Put both versions side by side and mark what moved
Not what you meant to change. What is actually different on the page: the experience ask, the title, the location, the benefits list, the salary line, the length.
Most employers who re-advertise find less has moved than they remember. Two in five of this population changed nothing at all. Among the ones that changed a little, a corrected spelling sits next to a deleted flexible-working line and next to a job that quietly moved to another city, so read the diff and do not trust your memory of it.
-
Check which seat these words were written for
This is the step for the principal who has hired ten times and not once. A job advertisement that worked is the most reusable document in the business and the easiest one to reuse wrongly.
Two employers in this set posted one title in two places and meant two different jobs by it, one for somebody already broking and one built as an entry pathway. If your words came off a search that filled, check that the seniority, the location and the pay they promise still describe the seat you are filling now.
-
Write the sentence that says why you are back
One line, for you, before anything is posted. Either the first search did not fill and here is the thing I am changing, or the first search worked and this is a second seat.
Everything downstream follows from which sentence you write, including whether the work in front of you is the advertisement at all. Note where you are posting it while you are there. Back into the same market, the people who could take it have already seen your advertisement and said no by saying nothing.
-
If it did not fill, name the change and name what stays
One variable, deliberately, and write it down beside the date. The employer in this set who changed a single clause can tell you what they learned from it. The ones who rebuilt everything at once cannot.
Then name what you are not changing and why, because that is the thing you will be tempted to trade away in week six when the seat is still open.
What people do instead
- Posting it again unchanged and calling it persistence. Two in five of these seats did, and the clearest is a Victorian boutique that ran a senior lending associate seat repeatedly, in one state, with the band named every time, and never touched a word. Not vague, not hiding the money, just the same shot fired again.
- Reaching for the money, because it is the only lever that can be pulled in an afternoon. For most of this population it is not even the visible variable, because 34 of the 65 never named a salary at either attempt. One Gold Coast brokerage sits inside that 34 and moved the money language anyway. Two versions of one advertisement, otherwise almost identical, and one of them promises a six-figure package in the headline and again in the body. The other deletes it from the headline and softens it to an adjective in the body. Nobody can say which came first, which is exactly the point. One seat, two different promises, and no actual figure in either.
- Forgetting that the second advertisement is read differently to the first. One broker said that before she applies anywhere now she wants to know how long everyone stayed in the business, because you do not always see that from the outside. Your repeat posting is public and so is your tenure record, and the experienced candidates you want are the ones most likely to read both.
- Deciding the answer before the diagnosis. Reading somebody else's repeat as desperation is the common one, and several employers in this set state their own growth in the opening sentence. Reading your own as growth when the seat simply did not fill is the expensive one, because it buys another six weeks of the same advertisement.
Do this today
The one thing to do today
Write one sentence at the top of the draft, for yourself and not for the page. It starts with I am back in the market because, and you have to finish it.
If the sentence is that the first search did not fill, then what you are writing is a revision, and it needs one named change and not a fresh coat of paint. If the sentence is that the business grew, the advertisement is probably fine and the work is somewhere else, in who onboards them and what the second desk costs.
The posting cannot tell a reader which of the two you are. The sentence tells you, and you are the only person whose decision it changes.
Where this comes from
The market figures are counted from a frozen advertisement set and re-derived from it, so they cannot move underneath the guide. Everything else is what principals and brokers said in recorded conversations, checked against the transcripts and not the summaries, and anonymised. Where a measure turned out weaker than it first looked it was cut rather than softened, and this section says which ones.
- BD Machine ad index. 991 Australian mortgage and finance advertisements posted between 23 February and 20 August 2026, frozen so the counts cannot move. 524 of them name a dollar figure, which is the rate the re-advertising population is compared against, on the same test applied to both.
- The re-advertising derivation. Run against that frozen set. A seat is an employer plus a job title, both case and whitespace folded, appearing between two and five times. Six or more is a standing advertisement rather than a seat being filled, and in practice no seat here went past four. Recruitment agencies and franchise brands are excluded, which takes 97 raw repeat pairs down to 65 seats and 144 advertisements. Both exclusions are load-bearing and 32 pairs fell to them.
- One comparison is published and the rest are described, on purpose. The only similarity figure here is the 26 that reposted word for word, which is exact string equality after folding case and spacing. Everything past that needs a threshold, and the thresholds move. Two honest implementations of the same comparison disagree about where a light edit stops and a rewrite starts, and they carry seats across the line in both directions. On the fairer of the two the rewrite group also falls below the publish gate. So the guide counts the word-for-word repeats and describes everything else, which is the treatment the split guide gave its own weakest figure.
- Two geographic patterns were cut, not softened. Earlier drafts claimed that just over half of the unchanged repeats ran in more than one city, and that about two thirds of the rewrites came back to the same city. About a third of the 65 seats are missing a city on at least one advertisement, and both figures only hold if a blank counts as a city. Read across the seats where every advertisement records a city, the first claim falls under half and the second rests on too few seats to state at all. Neither is in this guide, and there is no geographic tell in it.
- Where the population is not clean, stated rather than hidden. The exclusion list is exact company names, so at least one Sydney recruitment agency survived it, another row reads like an agency posting with the employer never named, and one is a global law firm's business development seat, which is not a mortgage or finance employer at all. Separately, a branch network staffing two branches looks exactly like one seat coming back. Several mutual bank rows are one title posted in two towns with two different contact people, which is the same problem the franchise exclusion catches and the name list does not reach. And 65 seats are not 65 employers: there are 58 distinct company names across them, a handful of businesses hold more than one seat, and one business appears under two spellings of its name. Seats is the honest unit and it is the one used throughout.
- What is below the publish gate. This series does not publish a count below twenty-one. The group that changed a little, the group that rewrote, the same-city and new-city splits, the state and role-family breakdowns, the businesses holding more than one seat, and every count of salary movement inside a group all sit under it, so each is described in words and never given as a figure. Salary movement is also soft on its own terms, because the salary field is free text and some apparent moves are one band written two ways, base salary plus bonuses against base salary and bonuses.
- FinTalent client and candidate call records, 2025 to 2026. Every quote verbatim from a recorded conversation and verified against the transcript and not the summary. Each was checked to confirm the speaker raised the point themselves rather than agreeing with something we had said. Nothing said by anyone at FinTalent is published here as a principal's testimony. Where more than one passage comes from the same conversation the guide names the link, which it does in section 01 for the principal quoted in section 02. Some principals in this series appear in more than one guide in it, each time for a different point and each time from a different part of the same conversation, so a reader of two guides is reading one source twice rather than two sources agreeing. Nothing in this guide is corroborated that way and nothing counts anyone twice. Businesses and people are anonymised, including in the pull quote, whose only cuts remove a first name.
Download this guide as a PDF. No form, no sign-up.
