Buying offices repeat themselves. That is a forecast.
An office that has bought the same class of thing every spring for six years will buy it again next spring. Nobody publishes that, and it is steadier than most published forecasts, because it is what the office did rather than what it planned last October.
money not obligated by 30 September goes back, and nobody gets credit for handing it back
When, who, what, how much, and how sure
Every row is one office and one class of thing: the month it is expected, the window either side, the typical size it goes for, how many buys the pattern rests on, the gap between them, who won it last time, and a confidence out of a hundred.
Filter by code, PSC, office, agency, place of performance, set-aside, value and horizon. How far ahead you can look is set by your plan: 12 months on One Code and Shop, 24 on Pro, 60 on Capture and Enterprise. A row more than two years out says so, because a date that far ahead rests on fewer and longer gaps, and it sits beside what agencies have published for the same years. Filter by confidence to see only what is worth acting on. Sort by date to plan a quarter, or by confidence to plan a year.
- Expected month with an honest window, never a bare date
- The typical award size, from the median of what was actually paid
- The incumbent, so you know who you would be taking it from
- A confidence built from the rhythm, the buy count and the competition
How a prediction is made
Award history is grouped by buying office and class of thing. Where that group has at least three separate buys, the gaps between them are measured. If the gaps are steady enough to mean something, the next buy is projected from the last one plus the typical gap, and a window is drawn from how much the gaps vary.
Actions inside forty-five days of each other are treated as one buy, because that is one requirement being paid in parts rather than a supplier opportunity that comes round twice in six weeks.
Confidence rises with a steady rhythm, more buys on record, an annual cadence, and more than one company having won it. It falls when the timing is irregular, when the same vendor wins every time, and when the office is already overdue against its own pattern.
Where an agency has put out a forecast, its lines sit in the same list
Some agencies publish what they intend to buy, once or twice a year, as a spreadsheet on their own site. Those documents say things no award record can: a requirement that has never been bought before, a named person to talk to, a set-aside decision not yet made, an estimated value on work with no history. Where one has been loaded, its lines appear in this list beside the projected ones.
They are never blended. A projected row is a rhythm read out of what an office did; a published line is what an office says it intends, which is a different kind of claim and is labelled as one. A published line carries no confidence score, because a plan has no rhythm in it to measure, and inventing a number there would be manufacturing evidence for somebody else's intention. One switch shows either kind on its own.
- The requirement in the agency's own words, which is the part no award record carries
- The quarter or fiscal year turned into a real window, never a bare date
- Marked as published rather than projected, on every row
- No confidence score, because there is nothing there to measure
UEI, CAGE, then normalized name
What this is not
A projected row is not an agency procurement forecast document. It is a pattern read out of the federal award record, and a pattern can break: budgets get cut, requirements get bundled into a larger vehicle, an option gets exercised instead. A published line is the opposite problem: agencies file their forecasts late, loosely, and sometimes not at all, and a line on one is a statement of intent that nobody is bound by. Treat the window as the answer, not the date, treat a low confidence as the warning it is, and treat a published line as what somebody planned rather than what will happen. Every projected row shows the buys it rests on so you can check the working yourself.
Where it is in the app
Forecast, under Intelligence in the menu. Intelligence is the part of LastAwarded that finds the market, one of the eight parts.
Related
- Recompete radarContracts with a known end date, rather than a predicted one.
- How to win federal contractsWhy the months before the notice are the ones that matter.
- Contracting peopleWho to talk to once you know what is coming.
See who was last awarded in your NAICS
Every code, every buying office, every vendor and every recompete, organized from the federal record and read nightly. Open it with a plan.