The previous piece described the fake reference economy. This one is about what replaces the process it has broken.
The short version: stop verifying the referee and start verifying the employment. Then, and only then, talk to a human about how the candidate performed.
That reversal sounds small. In practice it changes the order of every step, the data an employer asks for, the consent it needs, and what the reference call is for.
The principle: ground truth before testimony
A reference call is testimony. Someone says the candidate worked there. A reference house exists because testimony can be bought.
Ground truth is a record that exists independently of the candidate and the referee, in a system neither controls. Provident fund contributions. Tax records. Payroll data. Government labour registries. Bank salary credits. Company registration databases.
The rebuilt model puts ground truth first. The employer establishes, from independent records, that the candidate was employed by a real organisation for the period claimed. Only after that is established does the employer go looking for a human who can speak to performance.
This does not make the reference call worthless. It makes it a different tool. It stops being a check on whether the job existed and becomes a conversation about how well the candidate did it. That is what references were always supposed to be.
Layer one: independent employment records
What counts as ground truth depends on the jurisdiction, and a global verification programme has to map it country by country.
In India, the primary source is the Employees’ Provident Fund Organisation record, accessed through the candidate’s Universal Account Number. A UAN history shows which establishments made PF contributions for the candidate and over what dates. It exposes employers the candidate did not disclose, overlapping employment, and gaps that the CV papered over. It does not show title or salary, and candidates who worked below the PF threshold or at exempt establishments may have legitimate gaps. But as a test of whether a claimed employer actually employed the candidate, it is close to definitive, and no reference house can alter it.
The second Indian source is the Ministry of Corporate Affairs registry. A claimed employer that has no registration, or whose registration post-dates the claimed employment, or whose registered address is a residential flat, is a signal worth following.
In the UK, the sources the Better Hiring Institute recommends are HMRC employment history, which the candidate can obtain and share, payroll data, and open-banking salary evidence. Each confirms that a specific employer paid the candidate over specific dates. Companies House plays the role MCA plays in India.
In the Gulf, the equivalents are ministry labour records and work permit histories. In the UAE, MOHRE’s employment process creates a record tied to the work permit; in Saudi Arabia, Qiwa documents private-sector contracts. Both can be checked with candidate consent, and both are beyond a reference house’s reach.
In the US, The Work Number and similar payroll-derived databases cover a large share of employers, and for the rest, the candidate’s W-2 forms serve the same function HMRC history serves in the UK.
The consent point matters. Every one of these sources involves the candidate’s personal data, and in India specifically, the DPDP framework requires an itemised notice and consent for each. That is not an obstacle. A candidate who refuses consent to a UAN check, while offering a referee’s mobile number, has told the employer something.
Layer two: verify the referee, not just the reference
Once the employment is established from records, the employer still wants to talk to someone. The second layer is making sure that someone is real.
The tactics here are well established and routinely skipped.
Find the referee independently. Not through the number on the CV, but through the company’s published switchboard, its LinkedIn presence, or a direct search for the named individual. If the referee cannot be found anywhere except the candidate’s CV, treat the reference as unverified.
Check the domain. A reference house email domain is often registered recently, or is a close variant of a real company’s domain. A WHOIS lookup takes thirty seconds.
Check the timeline. Did the referee hold the claimed role during the claimed dates? A “line manager” whose own LinkedIn shows them joining the company after the candidate left is a reference house tell.
Listen for scripts. Reference houses answer common questions well. They answer unexpected ones badly. A question about a specific project, a specific office, or a colleague the candidate did not mention is the fastest way to find the edge of the script.
Cross-reference against known operators. The UK now maintains lists of identified reference houses, and screening firms that build reference-house checks into their process can flag a match before the call is made. Rosser’s practical advice to employers is to insist that their screening company does this.
Layer three: the reference call, repositioned
With the employment established and the referee verified, the reference call becomes what it should be: a structured conversation about performance, conduct and fit.
Standardise the questions so that answers can be compared across referees and across candidates. Ask about specifics: a project, a decision, a difficult period. Ask what the referee would want to know if they were hiring the candidate again.
And consider the one tactic reference houses cannot defend against.
The backdoor reference
A reference house controls the referees the candidate supplies. It has no control over the ones the candidate did not.
Zety’s professional references survey found that 37% of jobseekers had experienced an employer contacting a former manager or colleague who was not on their reference list. The same survey found that one in ten jobseekers had asked a friend or family member to pose as a former coworker, and that candidates regularly coach their referees.
The backdoor reference, contacting someone at the claimed employer whom the candidate did not nominate, is the single most effective counter to reference fraud, because it reaches a person the fraud has not reached. It also reaches a person with no incentive to flatter.
It is also the tactic most employers avoid, for two reasons that deserve a straight answer.
The first is legal. In most jurisdictions, contacting a former employer about a candidate is lawful with the candidate’s consent, and the consent form can be written to cover the employer rather than named individuals. Where a jurisdiction restricts this, the restriction should be respected. Where it does not, the hesitation is habit, not law.
The second is the candidate’s current employer. Nobody should contact a candidate’s present employer without explicit, specific consent, because the cost to the candidate of exposure is high and the fairness problem is real. But the restriction belongs to the current employer only. For previous employers, where the candidate has already left, the risk to the candidate is low and the verification value is high.
A verification programme that reserves backdoor references for roles above a defined risk tier, with consent language that covers it, will catch a category of fraud that no other step reaches.
The sequence, written out
For a mid-to-senior hire in a role with access to money, data or customers, the rebuilt employment verification runs in this order.
First, with itemised consent, pull the independent employment record: UAN history in India, HMRC and payroll data in the UK, ministry records in the Gulf, payroll databases in the US. Compare it line by line against the CV. Every undisclosed employer, overlap and gap gets a question.
Second, for each claimed employer, confirm the organisation exists and existed during the claimed dates, through the company registry and its public footprint.
Third, for each nominated referee, verify the person independently of the CV: find them, confirm their role and tenure, check the domain.
Fourth, screen the referee contact details against known reference-house indicators.
Fifth, conduct the structured reference conversation, focused on performance.
Sixth, for roles above the defined risk tier, make one backdoor reference contact per previous employer, under the consent already obtained.
The first four steps are where reference fraud is caught. The fifth is where the employer learns something useful. The sixth is insurance.
What this costs, and what it saves
The honest objection to this model is time. A reference call takes ten minutes. A source-record check takes consent, a data request, and a comparison.
Two responses.
The first is that the ten-minute call was never producing verification. It was producing the feeling of verification. The cost of a fraudulent hire, in the Cifas framing, is not a bad employee; it is an insider with a blackmailer. Measured against that, a day’s additional turnaround is cheap.
The second is that the source-record check is mostly automatable. UAN histories, HMRC records and payroll databases are queried, not phoned. A verification partner with the right integrations can return the independent employment record faster than a referee returns a voicemail. The slow part of the old process was the human. The fast part of the new one is the data.
The reference call is not dead. It has been demoted.
Employers will keep calling referees, and they should. The conversation about how someone works is valuable, and no database provides it.
What has to end is the reference call as the evidence that a job existed. The fake reference economy has made that use of it unreliable, at scale, in every market this firm operates in. The record that proves employment sits in a provident fund ledger, a tax database or a ministry registry, and the candidate’s consent unlocks it.
Check the ledger first. Then make the call.
