A weekly series becomes a finding when a reader can see what it counts, over whom and as of when. This guide is for research, analyst-relations and media teams who turn the Fokals weekly Market Series into charts, indexes and statements: how to pick a series, what to test before a number goes into print, how to chart and cite it, and how to word a claim so that it says no more than the data does.
What one row holds
A series is one metric over one dimension and one window. A row is that series at one as-of date, a Sunday, written once. Six groups of columns decide what you can say.
| Column | What it is | Use in a finding |
|---|---|---|
as_of, window_days | The Sunday the window ends, and its length: 7, 30 or 180 days | State both in every caption |
count | Events, postings or companies in the window; for a median metric, the median itself | The numerator |
cohort | What the rate is over: companies or websites covered since before the window began, or postings for share and pay metrics | The base: quote it beside the rate |
rate | The count in the metric's unit: per 100 companies or websites, a share, or a median in US dollars | The headline figure |
previous, growth | The same count over the window before, and the change between the two | The direction of the count; for a share metric, not of the share |
index | 100 at the series' first as-of date, then the rate relative to it | Comparing series that start at different levels |
Rates for event metrics are taken on a same-store cohort: companies covered since before the window began and still covered at its end, so growth in the Fokals index is not mistaken for growth in the market. The blog post on how the cohorts stay comparable sets out the arithmetic.
An illustrative row, with invented values, shows the reading:
metric hiring_new, dimension_kind industry, window_days 30
count 960, cohort 600, rate 160, previous 870, growth 0.1034, index 112Read it as three sentences, each with its own test. In the 30 days to the as-of date, the 600 companies in the cohort opened 160 postings per 100 companies. That is 10 percent more postings than in the 30 days before, counted over the same companies. The rate is 12 percent above its level at the first as-of date of the series. Each sentence needs the window, the as-of date and the cohort beside it.
Choosing a series
Start from the question, then pick the family. The catalogue at GET /api/v1/series/metrics lists all 28 families with their units, windows, dimensions and subjects. The table maps common research questions to families.
| Question | Family | Unit |
|---|---|---|
| Is hiring speeding up in an industry? | hiring_new | Postings opened per 100 companies |
| Are advertised wages moving? | pay_median | Median yearly US dollars |
| Which technologies are spreading? | tech_added:salesforce, tech_prevalence:salesforce | Websites per 100 |
| Is more of the news about funding or restructuring? | news_event:funding_round, news_event:layoffs_restructuring | Companies per 100 |
| Are companies changing platform or entering markets? | replatformed, market_added | Websites per 100 |
Pulling and keeping the data
The REST API returns series rows newest first. The first call below asks for the 30-day hiring rate of every industry from the as-of date you choose, and the second lists the series that moved most at the latest as-of. Both carry your key as a bearer token.
GET /api/v1/series?family=hiring_new&dimension_kind=industry&window=30&since=2026-10-04&limit=1000
GET /api/v1/series/movers?dimension_kind=industry&window=30&family=hiring_new&limit=10The movers call is a way to find a story. Treat what it returns as a lead, not a finding: the largest growth is often on a small count, so read count and cohort before you pick it up.
Rows are written once and not revised, so a pull stored with its pull date reproduces a chart on any later day. Key what you store on as_of, metric, dimension_kind, dimension and window_days. A new as-of is written each week for the Sunday that has just closed, so pull once a week and append the rows to your store without overwriting earlier ones. The export endpoints return the same rows as files in CSV or JSON Lines.
A claim that rests on two series needs them to match. Pair hiring_new with ai_role_share for the same industry, window and as-of date to show whether a rise in postings comes with a rise in the share of AI roles, and cite both. Two series that move together show an association, not a cause. Do not set a 7-day row beside a 30-day row, or two dimensions side by side, as if they were one comparison.
Tests before a number is printed
- Read the cohort. Quote it beside every rate. Every row stands on at least 20 companies, or 20 postings for pay and share series, and you can set a higher floor of your own for a headline figure.
- Name the unit. A rate per 100 companies is not a share of postings. Say which in words, because a reader will take a bare 12 for a percentage.
- Do not chart growth alone. For a share metric,
growthis the change in the count of one kind of posting, so it rises when all postings rise even if the share falls. Chartrateorindex, and showcountandcohortbeside any growth. - Use windows that do not overlap. Weekly rows of a 30-day window share 23 days of postings. Compare 7-day rows week on week, or 30-day rows five weeks apart.
- Treat a gap as unknown. A week with no row for a series means the cohort was under the floor, not that the value was zero. Leave the gap in the chart.
- Check where each series starts. Take the earliest
as_ofof every series you chart, because each series sets its own index base at its first as-of, and compare indexes only over the dates they share. - Name the population. The series cover the companies of the Fokals index, and the hiring series cover those that publish a job board. Size bands are built on stated headcount.
- Check the dimension.
countryis the headquarters country, not where the work is done, andindustryis a label produced under a named, frozen version.
Charting and citing
Draw one point per as-of date and leave a gap where no row was written. Start an index chart at 100 and say that the base is the first as-of. Put the unit on the axis and the window in the subtitle. A footnote that carries the metric, dimension, window, as-of date and cohort lets a reader trace the figure:
Source: Fokals Market Series, metric hiring_new, dimension industry, 30-day window, as of <date>.
Rate per 100 companies in a same-store cohort of <cohort> companies with a job board.For a share metric, replace the second line with the postings counted, for example the share of the cohort's postings opened in the window that state remote work. The wording matters as much as the chart. Say what was counted, not what you infer from it.
| Instead of | Write |
|---|---|
| Hiring rose 12 percent | Postings opened by the cohort rose 12 percent against the previous 30 days |
| Remote work fell | The share of new postings that state remote work fell |
| Layoffs are rising | More companies per 100 announced layoffs or restructuring |
| The market for a tool is growing | The number of websites per 100 running the tool rose |
What the series measure
The series describe the companies in the Fokals index, and they record what companies publish and do: postings opened, announcements of restructuring, tools detected on a website. Read them as measures of that activity, in the words of the table above: postings opened, not jobs created; announcements of restructuring, not jobs lost; tools detected on a website, not licences held. Every row is dated by its as-of Sunday and written once, so a growth figure compares like with like: the window against the window before it, over the same companies.
How Fokals delivers it
The Market Series dataset holds 28 families of weekly series, written once for each Sunday, with a count, the cohort, a rate, the previous window, growth and an index. They reach you through the REST API, described in the API reference, or as bulk files. Licensing is by written agreement for internal use, embedding in a product or redistribution, and publishing a chart or a finding is a use to name in that agreement; the data licensing page describes the three uses. For two worked readings of the series, see measuring remote and hybrid work trends and sector hiring trends for macro and thematic research.
Frequently asked questions
How do I cite the Fokals market series?
Name the metric, the dimension, the window length and the as-of date, and quote the cohort. For example: Fokals Market Series, hiring metric, industry dimension, 30-day window, as of a stated Sunday, rate per 100 companies in a cohort of a stated size. Add that the series covers the companies of the Fokals index.
How large is the cohort behind each row?
Every row stands on at least 20 companies, or 20 postings for pay and share metrics, so a figure always rests on a base large enough to read. The row carries its cohort, and you quote it beside the rate. Where a week has no row for a series, the cohort was under that floor, which says nothing about the value, so leave a gap in the chart or use a broader dimension.
Can the numbers be revised after I publish?
No. Each row is written once for its as-of date and not rewritten, so a figure you cite can be reproduced on any later day. A change to the definition of a metric is released as a new metric, never as a rewrite. Store your pull with its date anyway, so that a chart can be rebuilt without a new request.
What is the difference between growth and index?
growth compares the count in a window with the count in the window before it. For an event metric the cohort is the same in both, so it is also the change in the rate; for a share metric it is the change in the count of postings, not in the share. index compares the rate with the rate at the series' first as-of date, which is set to 100, so it suits comparing industries that start at different levels.
Can I publish charts built from Fokals data?
Licensing is by written agreement for internal use, embedding in a product or redistribution. Publishing a chart or a finding built from the series is a use to settle in that agreement before you do it, together with how the source is to be named.
The queries and code on this page are examples to adapt. Test them in your own environment before you rely on them.