RATES CURRENT 8/2026 The wages employers in Pakistan post, role by role See the rates
PAKISTAN GUIDE

Methodology: how SourceGrowth's data works

Updated 8/2026

Authority you can check. Here's how every number on this site gets made.

Every claim on this site falls into one of two categories: public-data indexes we've labeled and organized, or ground verification our own team in Pakistan has physically confirmed. This page explains the difference, states plainly what "verified" will mean when that label starts appearing, and covers our rules for aggregating quote data, refreshing numbers, and disclosing financial relationships. Read this page once and every other page on the site should make sense in context.

Two kinds of data, never blended

Our index. Public job-board postings, published EOR pricing, government trade and tariff data, and other information anyone could in principle collect. We normalize it, label its source and pull date, and present it clearly as index data, not as something our team has physically confirmed. An index page tells you where its numbers came from and when they were pulled.

Three index datasets are live today, each dated on its own page: 913 posted-salary observations behind the hiring cost guide and the rate indexes, five years of monthly US customs values across 93 HS chapters on the trade flows page, and 899 Pakistani software firms in the provider registry. Every published dataset and its pull date is listed on the indexes page.

Our verification. Facts our own Pakistan operations team has physically confirmed: an entity exists at a stated address, a facility matches its claimed capabilities, certifications were sighted in person rather than taken on the supplier's word. This is the layer that makes us different from a site that only aggregates what's already public, and it's the layer we're most careful never to imply before it actually exists.

As of this page's publish date, nothing on this site is labeled "verified" yet. Ground verification requires a physical visit our team logs with a date, and that verification program is standing up alongside the rest of the site. Any page describing providers, factories, or shortlists before that program has run will say so plainly rather than implying a verification status it doesn't have.

What "verified" will mean

Once verification entries start publishing, the word "verified" on this site means a specific, checkable thing: a member of our Pakistan operations team physically visited the location, confirmed the facts stated, and logged the visit with a date. Every verified record will carry a last_verified date visible on the page. Verification expires: we re-confirm every twelve months or immediately on any adverse signal, and a record whose verification has lapsed will be labeled accordingly rather than left looking current. A fact we haven't personally confirmed is labeled "reported," not "verified," even when we have good reason to believe it. We hold that line even when it makes a page less impressive to read.

Verification isn't an audit. When a buyer needs formal compliance or quality certification, we refer that work to licensed third-party auditors, firms in the class of SGS, QIMA, or V-Trust, rather than implying our team's visit substitutes for one. We don't perform audits ourselves and we never will.

How quote and benchmark data gets aggregated

Two different collection methods feed the number layer, and we label which one produced a given figure rather than blending them.

Observed postings. The wage index is built from public job postings where the employer stated an explicit figure. Nothing is inferred, modeled, or taken from an aggregator's estimated-salary page. A cell is a role family and seniority combination, and it publishes only at three or more observations.

Solicited quotes. MOQ, lead-time, and buyer-facing cost benchmarks come from real quotes gathered on a standing cadence, never from a single source presented as representative, with a cell being a defined combination like sector, product category, and order-volume band. Same floor: three or more independent quotes before anything publishes.

Under either method, a cell below three observations shows "insufficient data" rather than a number built on a sample too small to mean anything, and the sample size prints on the page so you can weigh it yourself. Where a published figure is derived, a buyer-facing rate built from an observed wage plus a modeled vendor markup, for example, we state the derivation directly: what was observed, what was modeled, and the assumption behind the model, rather than presenting a derived number as if it were an observed one.

The cost model: what they earn, your cost to employ, typical agency quote

Every role page, rate index, and comparison on this site carries three numbers side by side: what the person earns on paper, what it actually costs you to employ them, and what a typical agency quotes for the same work. Only the first is observed. The other two are computed from a disclosed formula, and this section is the formula.

Your cost to employ is the posted wage plus Pakistan's employer-side statutory adds. It does not include the Employer of Record fee, which we disclose separately (below) rather than blend into the per-role figure. A flat monthly fee, applied inside every row, would distort the ratio badly on lower-wage roles (a $100/mo data-entry wage next to a flat $599-699 fee reads nothing like the real economics of that hire).

  • Statutory adds. Pakistan's employer-side statutory categories are EOBI (the national old-age benefits scheme), provincial social security (PESSI in Punjab, SESSI in Sindh, and the KP and Balochistan equivalents), and gratuity that accrues per completed year of service. We attempted to source current, primary rates for all three in the 8/2026 pull behind this page. EOBI's published split, 5% employer and 1% employee of the government-notified minimum wage, came through a compiler source (Wikipedia, citing the EOBI Act 1976) after the EOBI government site itself refused every connection attempt (SSL and DNS failures). PESSI Punjab, SESSI Sindh, KP-ESSI, the ILO's NATLEX database and country profile, and the ISSA country profile were all unreachable or returned a blocked response in that same pull, as were every EOR platform's own published country page we checked (Deel, Multiplier, Papaya Global, Skuad, Remote, Horizons, Playroll, Rippling, Oyster, Ecovis). We could not assemble a clean, current, primary rate for the combined provincial-social-security-plus-gratuity layer, and EOBI's rate applies to a different wage base (the minimum wage, not the posted salary) than the other two, so we do not fold it into a single blended percentage either. Per our own rule against inventing a precise number we cannot source, the statutory-adds line on this site is a single disclosed band: 12-18% of gross wage, per statute mix, labeled our estimate pending primary confirmation. We re-attempt sourcing on every dataset refresh and will replace this band with itemized, dated rates the moment we can confirm them from a primary source.
  • EOR fee (separate line, not blended per role). Hiring through an Employer of Record instead of a contractor or your own entity adds a flat $599 to $699 per person per month, on top of your cost to employ above, per Deel and Multiplier's public Pakistan pricing pages, already cited elsewhere on this site. We show this once per page as a flat add-on, not folded into every row, so it does not distort the ratio for lower-wage roles.

Typical agency quote is not observed either. It is a disclosed working model: 1.8x to 2.6x the fully loaded cost above (wage plus statutory adds, before the EOR fee), informed by the wage-versus-billed-rate pattern documented in our rate articles (agencies and BPO firms price in recruiting, training, supervision, seat infrastructure, and absence coverage on top of the wage they pay). We have not collected a representative sample of actual vendor invoices, so we do not present this as an observed billed rate, and we say so everywhere it renders. If a quote you receive falls outside this band, that gap is worth asking your vendor about directly, and it is exactly the conversation our connect link exists to start.

Every rendered instance of these figures links back to this section and carries the word "modeled" or "estimate" once, small, not as a disclaimer wall. The underlying config (config/cost_model.php in our codebase) carries every rate with its source name, URL, and the date we checked it, so the provenance behind a given figure is never more than one file away.

The "% saved" figure

Where a role has a defensible US occupational match, rate pages also show what hiring in Pakistan saves against paying that same role in the US. The comparison is deliberately conservative: the US side is salary alone, no employer-side costs added, sourced to the US Bureau of Labor Statistics' Occupational Employment and Wage Statistics program (median annual wage, national estimate). The Pakistan side is fully loaded (wage plus the statutory-adds band above), so the US figure gets no help and our own figure gets no discount. We take the midpoint of our loaded band against the flat US monthly figure: % saved = 1 - (our loaded midpoint / US monthly wage).

Not every role gets this figure. Where the role is a genuinely US-occupation-spanning catch-all, or where more than one US occupation could plausibly apply, we publish no percentage rather than force a comparison we cannot stand behind. The table shows a blank mark instead. See database/data/us/METHOD.md in our codebase for the full occupation mapping and the roles left out.

Refresh dates on everything

Every index, benchmark, and conditions page on this site shows when its data was last refreshed. Conditions on the ground move faster than most content publishing cycles, and a number without a visible date is functionally a number nobody can trust. We'd rather show an older date honestly than quietly refresh the page's copy while leaving stale numbers looking current. Cadence varies by data type: wage index data refreshes monthly with a quarterly published update, quote benchmarks refresh quarterly (faster on any US tariff change), and operating-conditions entries refresh monthly.

When an index number and a real quote disagree

They will, and the disagreement is usually information rather than error. An index number is a distribution: posted wages across many employers, or customs values across many shipments. A quote is one counterparty pricing one job at one moment, including everything specific to it that the index cannot see.

Our rule is that a quote never silently overwrites an index cell. A single quote outside a published band is recorded, not published as a correction, because one observation does not move a distribution. If quotes keep landing outside the band in the same direction, that is a signal the index is stale or the cell is mis-specified, and the fix is to re-pull or re-cut the cell and say so on the page, not to quietly widen the range until the inconvenient number fits inside it.

For a reader, the practical version: use a published band to judge whether a quote is in the right neighborhood and to ask better questions about the gap. Do not use it to argue a vendor down to a number that describes a different transaction than the one you are buying.

Corrections

When we get something wrong, we fix it on the page and say what changed. A corrected page carries a dated note stating what the previous version said and why it was wrong, rather than a silent edit that leaves anyone who acted on the old number with no way to know. A figure that turns out to be unsupportable gets removed rather than softened into a vaguer version of the same claim.

This applies to numbers we sourced in good faith from someone else, too. If a cited source is corrected or withdrawn, the claim resting on it comes down here as well, on the same terms, whether or not anyone noticed.

Disclosure policy

We intend to build referral relationships with vetted providers and suppliers over time, and when that happens, the relationship will be disclosed directly on the page where the referral appears, not buried in a separate policy document a reader has to go find. Any shortlist or "vetted" page will carry its inclusion criteria on the same page, so a reader can see exactly what earned a provider a spot before deciding whether to trust the list. We monetize referrals openly rather than presenting the site as neutral when it isn't.

What we never do

We don't perform our own compliance or quality audits and present them as independent verification. That work goes to licensed third-party auditors, always disclosed as a referral, never claimed as our own capability. We don't accept payment to rank a provider higher in a shortlist or index. We don't publish a number as "verified" without a logged physical visit behind it. We don't extrapolate one zone's or one supplier's conditions into a country-wide claim. And we don't let a page's copy get more polished while its underlying data goes stale; if we can't refresh a number on schedule, the page says so.

This is what backs every take on this site. Check the number, check the date, check who's behind it. That's the offer.

Sources: this page describes SourceGrowth's own operating rules and is not built from external citations; it implements the dataset-collection and disclosure commitments stated in the property's brand platform and ground-truth dataset protocol.

Data status: wages-v0, last pulled 8/2026 CURRENT (refresh cadence: every 31 days)
Public job-posting wage layer (posted salaries only, PKR). Wages are what employers pay locally - not billed outsourcing rates.

See if Pakistan fits what you're building.

The guide above is the market. Our people on the ground turn it into a plan for your team or your product, and tell you straight if the fit is wrong.

Get a straight fit-check