The real cost of chasing the same requirement twice

Why is duplicated work more than duplicated triage?

Because triage is the cheapest stage it touches. If one role exists as twelve records, you do not just read it twelve times. You build a shortlist against each, you make twelve separate decisions about which consultants fit, you track twelve submission threads, and you carry twelve chances of a collision.

The cost is multiplicative rather than additive, and it lands on the stages that are already the most expensive. Shortlisting is the largest single block of idle time on most desks; multiplying it by twelve is not a rounding error.

It also degrades quality in a way that is hard to see. A recruiter working the fourth copy of a role at 4pm is not making the same decision they made on the first copy at 9am.

What does deleting duplicates cost commercially?

Price discovery, first. The spread between the best and worst rate on the same role routinely runs 15–20%, and that spread only exists as information while all the copies do. Delete eleven of twelve and you have thrown away the comparison that tells you which route to submit on.

Optionality, second. If the vendor you chose goes quiet — and vendors go quiet constantly, for reasons that have nothing to do with your candidate — the alternative routes are gone from your system. The desk then starts over on a role it has already qualified for, and the consultant waits.

Both losses are invisible in the moment. Nobody records the rate they did not negotiate or the route they could not fall back to, which is why deletion feels clean and costs money.

How do duplicate records cause RTR collisions?

A right-to-represent authorises one vendor to submit one consultant to one requirement. Enforcing that is a lock, and a lock needs something to attach to. If twelve copies of a role exist as twelve unrelated records, there is no shared object holding the fact that this consultant is already represented on it.

So two recruiters — or one recruiter on two different days — submit the same person through two chains to the same end client. The client does not adjudicate who was first. They disqualify the candidate, and frequently remember the agency that caused it.

This is the failure that turns a bookkeeping problem into a commercial one, and it is entirely preventable by collapsing first.

What does canonical-plus-routes actually look like?

One canonical job holds the role: skills, work authorization, location, end client, duration, and the shortlist. Every inbound copy attaches to it as a route, carrying its own vendor, rate, tax term, and layer count.

The desk then works one role and picks a route at submission time. Nothing is chased twice, nothing is thrown away, and the RTR lock has a single object to attach to.

• The shortlist is built once, against the canonical requirement.
• Rate comparison becomes a column rather than an inbox search.
• Right-to-represent is enforced per route, so no candidate goes out twice on one role.
• If a route stalls, the next-best one is already qualified and ready.
• Layer count sits next to rate, so the margin-versus-proximity trade is visible.

How do you tell two postings are the same role?

Not by exact text. Vendors rewrite titles, trim or pad the skills list, and adjust rates at every hop, so string matching fails on the first realistic example.

The workable approach is two-stage: compare the parsed requirements in vector space to find candidates for a match, then confirm on the structured fields that do not drift — end client, location, work-authorization requirement, and duration. Semantic similarity alone produces false merges between genuinely different roles at the same client; structured confirmation is what makes it safe.

Anything ambiguous should surface for a human rather than merge silently. A wrong merge hides a real role, and a hidden role is worse than a duplicated one.

How much is this worth on a real desk?

Take the triage cost first, because it is the easiest to measure. If a morning triage session runs 90 minutes and collapsing duplicates removes two-thirds of it, that is an hour a day of recruiter time — roughly 250 hours a year on a single desk.

Then the idle days, which dwarf it. If duplicated shortlisting adds even two days to the median time-to-redeploy, a 40-consultant bench on nine-month assignments loses about 106 consultant-days a year. At $65 an hour and 22% gross margin, that is roughly $12,000 of margin — from one stage, from one cause.

Then the collisions, which are lumpy and unbudgeted. One disqualified candidate on a role you would have filled is a full placement fee plus the residual damage to the vendor relationship.

How do you test this without buying anything?

Spend a week collapsing by hand. Take every inbound requirement, group the copies, and keep a single sheet per real role with a row per vendor showing rate, layer count, and contact. Nothing gets deleted.

Measure two things: minutes spent on triage per day, and how often the rate you eventually submitted at was not the first rate you saw. The second number is the one that surprises people — it is the value of the price discovery that deletion was destroying.

If the week is convincing, the argument for automating it is that the manual version does not survive contact with a busy Monday. The discipline is the hard part, not the concept.

What changes on the desk once duplicates are collapsed?

The morning changes first and most visibly. An inbox of ninety postings becomes a dozen real roles with a vendor list attached to each, and the triage session stops being a reading exercise and becomes a prioritisation one. Recruiters notice this within days and it is the change they comment on.

The commercial change takes a quarter to show up and is larger. Every role now carries its own price ladder, so the question at submission time shifts from "which vendor emailed me about this" to "which route gets this consultant in front of the decision maker at the best rate". Those are different questions and they produce different margins.

The third change is defensive. Right-to-represent enforcement becomes possible because there is a single object to lock against, which removes a category of self-inflicted disqualification that most desks have simply learned to live with.

What does not change is requirement flow. Collapsing duplicates makes the roles you receive visible and workable; it does not create new ones. A desk whose real problem is that not enough is arriving will see a tidier inbox and an unchanged median, and that is the correct diagnosis rather than a failure of the approach.