Methodology

How these numbers are produced, and what they do not mean. If a figure here looks wrong, this page should tell you why.

What this measures

Advertised salary — the pay range an employer put in a public job posting. It is not actual pay, not a wage survey, and not a cost-of-living estimate. Employers who advertise a range tend to differ from those who do not, so the figures describe the advertised market, not the labour market.

Where the data comes from

Public job postings from Philippine job boards, collected in a real browser at a human pace. Only the fields needed for statistics are kept: title, employer name, location, employment type, posting date and the salary text. Job descriptions are not stored or republished, and no applicant or contact information is collected.

How a posting becomes a number

• Salary parsing. The salary text is read into a currency, a period (hour / day / week / month / year) and a range. Fixed project fees and postings whose salary field contains contact details are excluded.
• Conversion. Everything is converted to PHP per month using the exchange rate on the first day of the month and fixed period factors: hour × 173.33, day × 21.75 (261 working days a year ÷ 12), week × 4.33, year ÷ 12. The basis of every conversion is stored with the row. Remote work for overseas clients is mostly advertised in US dollars, but not all of it: some of those ads are in pesos, and they are counted as written.
• Role classification. Titles are normalised (promotional noise removed) and matched to a role family by dictionary. Unmatched titles are kept but excluded from statistics.
• De-duplication. Postings sharing employer, title, location and salary are collapsed to one — including repeats from the same site. Staffing agencies re-list the same role dozens of times a month.

What is excluded

• Segments with fewer than 10 salary-disclosing postings, to avoid identifying individual employers.
• Postings that state no salary — which is most of them, and not at random.
• Government vacancies, whose pay is set by statute rather than the market.
• Amounts that cannot be a monthly wage (implausibly low or outside a plausible range), flagged and dropped.
• Internship, OJT and practicum ads, whose figure is an allowance rather than a wage; and ads whose stated period we cannot trust — an annual figure below the monthly minimum wage, or a board printing yearly pay under a “per month” label. We drop those rather than guess what the advertiser meant.
• Postings advertising below the lowest statutory rate that could legally apply are flagged (they remain visible in counts but are marked).
• We say so when a figure leans on one source or one employer. A staffing agency can post the same role across fifteen cities under fifty different headlines; those are separate listings and we count them separately, but they carry only a handful of distinct salaries between them. In August 2026 one agency accounted for 48% of the customer-service ads that state pay. Collapsing them to one row per employer and amount left the median unchanged at ₱24,000 — so we do not adjust the number, we tell you the concentration, because the risk is not that today's figure is wrong but that next month's moves when one advertiser changes its mind.

The minimum wage line

Role pages for Metro Manila (NCR, ₱755 a day under Wage Order NCR-28) and Metro Cebu (Central Visayas, ₱540 a day under Wage Order ROVII-26) draw the regional minimum wage as a dashed line and give the median as a multiple of it. The daily rate is turned into a monthly figure at 21.75 working days. Other metros and the national pages show no line yet: we enter each region's wage order by hand from the wage board, and a single national floor would hold regions with much lower wage orders to Metro Manila's rate.

Monthly insights

Each report carries one short analysis drawn from the same postings as the index, limited to Philippines-based jobs. Two things are read from the title of the ad, never from its body: whether it says no experience is needed (phrases such as “no experience”, “without experience”, “fresh grads”, “freshers”, “career shifters”) and, for BPO roles, which client account it names (travel, healthcare, telco, financial, insurance, retail, gaming). An employer is counted as a staffing agency when its name or stated industry says so; employers we cannot classify are left out of that comparison. That includes every ad from OnlineJobs.ph, a board that does not show the employer's name: with no name there is no way to tell an agency from a direct employer, so those ads are dropped from any comparison by employer type rather than counted as a third group. Every figure shows its count, and groups under 10 postings are not shown.

Pay by employer

Role pages and the employer list show employers and staffing agencies with at least 10 ads that state pay for that role and city. One month is too thin for most employers, so these figures pool the last 12 months (the same ads as the index, one count per ad even when several boards carry it) and take the median once over the pooled ads. Names are matched across boards, but a company can still appear twice when boards spell it differently; we leave that rather than guess. Ads without an employer name are not listed. When one employer supplied a large share of a month's ads, the role page names it in the list. We never show owners, officers, recruiters or contact details.

Each board is reported on its own — there is no single national figure

A plain median across all postings moves whenever the mix of postings changes — a month with more overseas-remote listings looks like a pay rise that never happened. Until September 2026 the headline was a fixed-weight composite meant to hold that mix still: role × work-arrangement medians weighted by the base month's composition.

We withdrew it on 13 September 2026. Between April and August 2026 the composite rose 37.1%, from ₱27,740 to ₱38,019. Over the same five months Bossjob's own median — the one board present in all five — did not move at all: ₱22,500 every month. Recomputing on a fixed basket, following only the 18 segments that had a published median in every one of those months, gives +2.4%. The rest is arrival: the number of segments with a median went from 61 in April to 99 in August as our collection widened, and each new segment entered the basket at whatever it happened to pay. Fixed weights cancel a change in the share of each segment; they do not cancel segments arriving and leaving. On top of that, in every one of those months between 79% and 100% of the basket's weight sat in segments where a single board supplied more than 60% of the ads. The number was tracking how much of each board we had managed to collect, not what employers pay.

What replaces it is narrower and true: the median advertised pay on each board, for Philippines-based jobs, published separately. A board's median is comparable with itself from month to month. Boards are not comparable with each other — a board carrying mostly overseas-client remote work will sit far above one carrying government vacancies, and that gap is the work, not the pay. Ads for overseas clients are counted separately from Philippines-based ads on every board, and the two are never added together. Role × city medians, which were never affected by this, are unchanged.

A board needs 50 ads with a stated salary in that scope before it gets a median — five times the ten-posting floor used everywhere else, because a board median is a headline figure rather than one cell in a table. A board also has to be having a normal month: if a month holds less than a tenth of what that board usually carries, the figure is withheld. It would be measuring how much we managed to collect, not what the board pays — the same reason the combined index was withdrawn. In July 2026 that applies to JobStreet, which had 81 ads against 6,519 the following month, and read ₱9,000 higher as a result. Every board is shown with its count. The figures are downloadable as board-medians.csv.

What the medians leave out

Of the ads that state a salary in August 2026, 73% reach a published median. The rest are dropped for two reasons, and neither drop is random: 19% because we could not place the job in a role (those ads have a median of ₱30,000 against ₱27,500 for the ones we keep — roughly 9% higher, as unusual and senior titles are the hardest to classify), and 8% because the pay was flagged as unreliable — an outlier, a guessed period, a currency we had to infer (median ₱49,004, far above the rest, because odd formats cluster at the top end).

This pushes the published figure down, so it is worth knowing how far. On the largest board in August, the median is ₱23,750 as published, ₱24,500 if the ads we cannot classify are added back, and ₱24,000 if every ad that states a salary is counted regardless of flags — a spread of about 3%. The exclusions change the number by a little; they do not change what it says.

Months collected retrospectively are not published

Historical months were reconstructed from postings still listed on the source sites. Postings that were filled or expired are gone, and what remains is not a random sample of what was advertised at the time. Those months are computed but withheld, and marked as such.

Revisions: preliminary, then final

Job ads keep surfacing after the month they were posted in, so a month's figure is not settled the day the month ends. We handle this the way statistical agencies do. A month is published first as a preliminary estimate; late-listed ads can revise it until the following month's report is released, at which point it becomes final and late-listed ads no longer move it. Each report states whether it is preliminary or final and the date it was first published. The CSV carries arelease column with the same status.

A change to the parser or the aggregation rules can still restate a month that has closed; each row records the versions that produced it. The largest such restatement to date is the withdrawal of the composite index described above, on 13 September 2026.

On 21 September 2026 we restated three things. Pay that is plainly in the wrong unit — what can only be a yearly figure listed by a board as “per month”, or a “per year” amount below the monthly minimum wage — no longer enters a median; we drop it rather than guess the intended unit. Internship and OJT allowances are no longer counted as advertised pay. And a language ad now has to ask for the language in its title; a title that only contains the word (a Japanese-owned plant, a Japanese restaurant) no longer counts, which had put the Japanese figure too low. July 2026 had already been declared final; it was restated as well, so that one series does not mix two definitions.

On 27 October 2026 we restated every month again, final ones included: more job-title wordings are recognised, three metro areas were added, and ads on Philippine boards for jobs located abroad were taken out of the figures. A final month changes only through a restatement like this one — never because late ads arrived.

Contact

Corrections and takedown requests are welcome — the underlying postings can be removed from the statistics on request.