Guide · 8 min read
Salary benchmarking for tech roles in Kenya
Most salary data for Kenyan tech roles comes from self-reported surveys, which are slow, sparse and biased toward whoever felt like answering. Published salary bands from live roles are a better instrument. They are also a smaller sample, and knowing when a number is too thin to use matters more than having a number at all.
Published
Self-reported surveys versus published bands
A survey asks people what they earn. Responses skew toward the engaged and the satisfied, arrive months after the fact, and cannot distinguish base from total compensation reliably.
A published band is what an employer committed to pay for a role that is open right now. It is forward-looking, dated, and verifiable. The trade-off is volume: far fewer roles publish a band than there are people willing to fill in a survey.
For setting an offer, the published band is the better instrument, because it tells you what you are actually competing against this quarter.
Read the sample size before you read the median
A median drawn from five roles and one drawn from fifty are different objects wearing the same label. Any benchmark that does not show you its sample size is asking you to trust it blindly.
As a working rule: fewer than about ten observations tells you the rough order of magnitude and nothing more. Twenty or more starts to be usable for setting a specific band. Below five, a figure should not be published at all, because with a small enough cell you can reverse-engineer an individual person's pay.
That last point is why our own salary pages suppress any group under five roles entirely, and why they widen the geography — from city to country to region — rather than lowering the threshold.
Nairobi is not the same market as Kenya
Nairobi concentrates most senior technical hiring in the country, and its bands run above the national picture. Benchmarking a Nairobi role against a Kenya-wide median will under-price it.
The reverse error is also common: applying Nairobi bands to a fully remote role open to candidates anywhere in the country, which over-prices it relative to the competition for that specific candidate.
Decide which market you are actually competing in before you pick a comparison set. The question is not where your company is, it is where your candidate has alternatives.
Currency, review cadence and the total number
Shilling-denominated bands need an explicit review cadence to stay competitive over a multi-year tenure. A band that is right today and unreviewed for three years is a resignation with a delay.
Equity is rarely a substitute at this stage of most local markets. Candidates discount it heavily, and usually correctly. If equity is a large share of your offer, expect to explain the maths and expect scepticism.
Benchmark base against base. Comparing your total package against someone else's base salary is the most common way teams convince themselves they are paying well when they are not.
Setting a band you can defend
Pick your comparison market, pull the median and the quartiles, and decide deliberately where in that range you intend to sit and why.
Write the band down before you meet anyone. Bands invented during a negotiation are how identical roles end up paid differently, which is a fairness problem and eventually a retention one.
Publish it. The reply rate difference between roles that show a number and roles that say 'competitive' is large, and it costs nothing to capture.
Questions
- What is a reliable sample size for a salary benchmark?
- Ten or more observations gives you a usable order of magnitude; twenty or more supports setting a specific band. Below five, a figure should not be published at all — small cells can expose an individual's pay.
- Should I benchmark against Nairobi or against Kenya as a whole?
- Against whichever market your candidate has alternatives in. For an in-office Nairobi role, use Nairobi. For a fully remote role open nationally, the wider figure is closer to what you are competing against.
- Are published salary bands better data than surveys?
- For setting an offer, generally yes: they are current, dated and represent real employer commitments rather than recollections. The trade-off is a smaller sample, so sample size should always be visible.
- How often should a salary band be reviewed?
- At least annually for shilling-denominated bands, and in writing. An unreviewed band loses competitiveness quietly, and the first signal you get is usually a resignation.