Methodology

Where the data comes from

Kenyan public bodies publish their tenders and contract awards on the Public Procurement Information Portal (PPIP), run by the Public Procurement Regulatory Authority (PPRA), at tenders.go.ke. The Open Contracting Partnership mirrors that data monthly as a free public download in the Open Contracting Data Registry. This site works from that mirror. Every article states the date of the snapshot it uses; the mirror can lag PPIP by up to a month.

Source: Public Procurement Regulatory Authority (PPRA), via PPIP / Open Contracting Data Registry.

We publish analysis, totals and facts about individual public awards. We do not republish the dataset itself.

Which awards are counted

• Awards in Kenyan shillings with a value of at least KES 1,000. Smaller values are treated as data-entry errors.
• Records reviewed and found not to be purchases are left out, for example a financing arrangement recorded as a contract. Each one is listed with its reason and the date of the decision, and nothing is left out for its size alone. An article that leaves a record out says so in a note, with the effect on its figures.
• So an article's award count is lower than the number of contracts in the dataset. The difference is contracts with a value of zero, a value below KES 1,000, no value at all, or a currency other than shillings, and any records left out after review. Each article states how many were left out and why. Most are awards published with a value of zero, which usually means the value was never entered.
• An award's date is the contract signing date. PPIP publishes no separate award date.
• Articles give the period most awards were signed in: the last three calendar years of signing dates. A few awards carry much older dates, sometimes years before the portal's data begins, which are likely mis-keyed; some have no date. Both kinds are still counted. Each article says how many there are.
• These are published awards, not payments and not total public spending. Some buyers publish more completely than others.
• Awards reach PPIP late, so the most recent months always look low. They are marked as incomplete wherever they appear.

The biggest 1% of awards

A small number of very large awards hold most of the published value. A single award can move a total by billions of shillings. So every total on this site is also given without the largest 1% of awards, alongside award counts and medians. The article says where that cut-off falls.

Buyer types

PPIP does not classify buyers, so buyer types come from the buyer's name, using these rules in order:

1. County commissioners: the name contains "county commissioner". These are national government offices, so they are counted with ministries and state departments (group 5).
2. County governments and assemblies: the name contains "county" together with "government" or "assembly"; or "municipality"; or "county" together with "authority" or "corporation" (county-owned bodies).
3. Constituency development funds: the name contains "NG-CDF", "CDF", "constituency development" or "constituencies development".
4. Universities, colleges and schools: the name contains "university", "college", "institute", "polytechnic", "school", "academy" or "TVC".
5. Ministries and state departments: the name contains "ministry" or "state department", or is exactly the name of a state department that publishes without either word: "The National Treasury", "Office of the Prime Cabinet Secretary", "Public Health and Professional Standards", "Early Learning & Basic Education" or "Tourism".
6. Water and sanitation companies: the name contains "water" together with "sanitation", "sewerage", "services" or "company". Most are county-owned companies, so they are kept apart from national bodies.
7. Other national bodies: everything else, mostly state corporations, authorities, commissions and agencies.

"National bodies" in articles means groups 1, 5 and 7 together. Name rules can misclassify a buyer. If you see one that is wrong, please tell us.

County and sector: derived, not published by PPIP

PPIP's records do not say which county a buyer is in, and they put every purchase into one of three broad categories: goods, works or services. We add two fields. Both are clearly labelled wherever they appear, and neither is presented as PPIP data.

County. A county is given only when the buyer is plainly a county government or county body: its name pairs one of Kenya's 47 county names with the word "County". Examples are "Kilifi County Government", "County Assembly of Kisumu" and "Kakamega County Water and Sanitation Company". National government offices named after a county are labelled National. These include county commissioners, deputy county commissioners, the National Government Affirmative Action Fund, sub-county offices and service-delivery units. So are all other buyers: ministries, state corporations, schools, hospitals and water companies whose name does not include "County". "National" therefore means "not identified as a county body", not "located in Nairobi". The rules are versioned (county-rules/1).

Sector (estimate). A sector is estimated from keywords in the tender or contract title. Examples are "hospital" or "drug" for Health, "borehole" or "sewer" for Water and sanitation, "culvert" or "murram" for Roads and transport, "laptop" or "software" for ICT, and "beef" or "catering" for Food and catering. If the title matches nothing, keywords in the buyer's name are used, but only sectors that describe what a buyer does, such as health or education. The first matching sector wins, in a fixed order. A title that mentions a school, hall, office or market is not counted as Roads and transport, because "road" is often part of a place name. Titles with no keyword are labelled Unclassified. Keyword estimates can be wrong. Treat them as a guide, and check the title. The rules are versioned (sector-rules/1).

Suppliers and personal data

• Contact details (email addresses, phone numbers and contact names) are removed when the data is imported and are never published.
• Some suppliers are individuals trading under their own name. Suppliers are named only after their records have been reviewed, and individuals are never named.
• The same business can appear under several spellings. Spellings that differ only in capitalisation, punctuation or a legal form such as "Ltd" or "Limited" are grouped. Other differences are reviewed by a person before any supplier is named.
• Repeat awards to the same supplier are often routine purchases under quotations or framework contracts. They are not evidence of wrongdoing.

How articles are checked

Before an article is published:

• every number in it is recomputed from the data, and it is held back if the data has changed since it was written;
• the text is checked for contact details and for the names of suppliers that have not been reviewed;
• the data source's licence terms are checked, and the source is credited;
• a person reads and approves it.

If an article turns out to be wrong, it is corrected with a visible note, or retracted with the reason shown at the same address.

The home page figures (computed, not written)

The home page is computed from the PPIP records each time the site is built. No person writes or edits these figures, so they say "Computed from PPIP" instead of "Human reviewed". Before each production update, the owner reviews a list of every figure that changed since the last update and confirms it.

• Data updated, data through. "Data updated" is when we last downloaded or checked the PPIP data, not when the page

was built. "Data through" is the date of the PPIP snapshot the figures come from. Tenders published after that date are not included yet.

• Open tenders and closing dates. A tender is open when its closing time, as published on PPIP, is after the page

was built. Closing times more than a year ahead, or before the tender was published, are treated as data-entry errors and left out. Buyers can change closing dates by addendum, so check each tender on tenders.go.ke before you bid. How recent they are depends on the newest PPIP snapshot: PPRA publishes its data about once a month, so tenders published after the snapshot date are not included yet, and the page says so. When the newest snapshot is more than 14 days old, the tender sections carry a warning that newer tenders are missing; after 45 days they are hidden until new data arrives.

• Recent months. Buyers publish awards late, so award figures leave out the 3 most recent months. A month of tender

publications counts only once the snapshot covers the whole month.

• Signals and sector pulse. Tender publishing is seasonal (the financial year starts in July), so a signal compares

the last 3 complete months with the same 3 months a year earlier. Only buyers that published tenders in both periods are counted, so a rise is not just more buyers starting to use PPIP. A change is shown only when at least 30 buyers published in both periods, each period has at least 30 tenders, and no single buyer published more than half of them. Confidence is High for counts read directly from PPIP with at least 100 tenders in each period, and Medium when the signal uses a sector estimate or fewer tenders.

• Similar awards (historical range). For an open tender, the range is the middle half (from the 25th to the 75th

percentile) of awards with the same sector estimate and the same PPIP category (goods, works or services), signed in the 24 months before the snapshot. It is shown only when at least 20 such awards exist, rounded to two significant figures, with the number of awards it is based on. It is an estimate of what similar contracts were worth, not a forecast of this tender's value.

• Recent awards against similar awards. The largest awards signed in the last complete month, each compared with

the median of similar awards (same sector estimate and category, 24 months, at least 20 awards).

• County activity. Tenders published in the last 12 complete months by buyers whose names identify them as a

county government or county body (the county rule above). National bodies are not shown. The 47 squares are placed roughly where each county is; they are not a map, and the ranked list beside them gives the numbers.

• Who is buying. Buyers ranked by the total value of awards signed in the last complete quarter, with how many

awards that is: a single very large award can put a buyer at the top on its own.

• People and suppliers. Suppliers are not named on the home page or the Open tenders pages. A tender or award

whose text contains an email address, a phone number or a supplier's name is left out of the lists.

• Synthetic test records are never counted.