In the chat we discuss that Kolesa company does not pull their weight when speaking about developer community. I agree with that. One community ask (maybe pointless) is to have raw data for yearly survey which they produce and this one of pillars of their public weight. Maybe not strongest, since they do yearly conferences, but still it’s visible.

But I believe this is pointeless ask, since Kolesa already made lot of results public by giving detailed results. Let’s see what we can deduce from this numbers.

Here the list of report

We have total sample size. It’s in the report.

Опрос проводился среди 643 специалистов разного уровня и направлений — от junior до senior, а также среди IT-руководителей, что позволило получить объективную картину текущих тенденций

Let’s create variable total_developers with value 643.

Initially let’s find statements which give us 100% in total.

74% разработчиков работают в казахстанских IT-компаниях, но растёт число специалистов в международных компаниях (14%) и на удалёнке (12%).

I will assume that this question Where you working? have 3 options

  • In Kazakhstan company
  • In international company
  • Remotely

so what this mean.

kazakhstan = 0.74 * total_developers
international_office = 0.14 * total_developers
international_remote = 0.12 * total_developers

this is slightly incorrect formula, since if it’s true then kazakhstan = 475.82 which obviosuly not true. Since even 0.99 of person is impossible to have real world. It is human or not as a whole. So let’s modify eqations

kazakhstan = total_developers * 74 / 100
international_office = total_developers * 14 / 100
international_remote = total_developers * 12 / 100

And now we can state that all variables kazakhstan, international_office and international_remote is ℕ.

kazakhstan = 475
international_office = 90
international_remote = 77

Hmmm. Something does not adds up. It’s because somewhere rounding happens. Let’s assume that rounding to nearest number happens. Mathematically we can define that like this.

kazakhstan = (total_developers * 74 + 50) / 100
international_office = (total_developers * 14 + 50) / 100
international_remote = (total_developers * 12 + 50) / 100

The reason for +50 is essentially we add .5 to number and then take whole of it. This is how Math.Round works mathematically.

kazakhstan = 476
international_office = 90
international_remote = 77

Now let’s look at company profile distribution.

type

If you notice, due to rouding if you addsup all values you will have 101% - great for publicity. So let’s modify formula a bit to account for that.

product = (total_developers * 38 + 50) / 101
fintech = (total_developers * 29 + 50) / 101
outsource = (total_developers * 13 + 50) / 101
startup = (total_developers * 8 + 50) / 101
other = (total_developers * 5 + 50) / 101
telecom = (total_developers * 5 + 50) / 101
commerce = (total_developers * 3 + 50) / 101

Let’s solve it.

product = 242
fintech = 185
outsource = 83
startup = 51
other = 32
telecom = 32
commerce = 19

Hmm. Again, total is 644. We have one additional person. Sigh. Actually if we have statment: Developers with property was a% out of total_developers that means we there roundining that also mean, that somewhere less developers then then strict percentage.

formula for the range [property_min, property_max] approximately following

property_min = floor((total_developers * (a - 0.5)) / 100)
property_max = floor((total_developers * (a + 0.5)) / 100)

Let’s recalculate values

kazakhstan = [472, 479]
international_office = [86, 93]
international_remote = [73, 80]

and

product = [241, 245]
fintech = [181, 187]
outsource = [79, 85]
startup = [47, 54]
other = [28, 35]
telecom = [28, 35]
commerce = [15, 22]

Speciality

Now let’s look at the structure by speciality.

type

This means that we have

backend = [318, 324]
frontend = [118, 125]
android = [93, 99]
ios = [73, 80]
cross_platform = [16, 22] // [22] after subgroup analysis
fullstack = [3, 9] // [4, 8] after subgroup analysis

And because of that let’s take a look at what how many backend developers by language we have

go = [103, 105]
php = [93, 95]
java = [87, 89]
csharp = [27]

Now frontend developers

javascript = [110, 116]
typescript = [6]
html = [2, 3]

and iOS developers

swift = [71, 78]
objective_c = [2]

and Android developers

kotlin = [90, 96]
cpp = [1, 2]
java = [1]

cross-platform developers

dart = [17]
javascript = [2]
java = [1]
php = [1]
typescript = [1]

and full stack developers

java = [2, 4]
javascript = [1, 2]
python = [1, 2]

Grades

Now we can analyze grades

middle = [241, 247]
senior = [241, 247]
junior = [67, 73]
lead = [67, 73]
director = [3, 9]
intern = [3, 9]
executive = [1, 3]

Work type

If you compare work type from this chart

type

where remote work is 26% and compare with original chart in the beginning of the report, I see discrepancies between remote international work and whole remote work. So it’s likely see presense of Kazakhstan remote work. Let’s calculate how remote work and office work distributed in Kazakhstan.

So we can see that 12% is percentage of the remote work in KZ.

kazakhstan_office = [392, 406]
kazakhstan_remote = [73, 80]

So in combined we will have

kazakhstan_office = [392, 406]
kazakhstan_remote = [73, 80]
international_office = [86, 93]
international_remote = [73, 80]

I believe I can reverse a bit of information from salary. But not today probably.

UPD: Let’s project these numbers on total developers population in Kazakhstan. According to the Bureau of National Statistics there 19 500 developers. Let’s project this, by multiplying by coefficient 30,32 (approx. 19500/643). The actual coeefficient should be smaller, since 19 500 number includes 1C developers, AI developer which does not counted in the Kolese report.

kazakhstan_office = [11 888, 12 313]
kazakhstan_remote = [2 214, 2 426]
international_office = [2 608, 2 820]
international_remote = [2 214, 2 426]

I don’t feel comfortable project other numbers, since likely the Kolesa Group did not reach some nich audience or questionary is build in such way, as to make other groups under represented. Like 1 Java developer across all Android developers seems fishy. Does there no legacy projects at all in Kazakhstan? Does any mobile developers in the outsource companies never touch Java on Android? Questionable. Also number of interns really small in comparing to what actually on the market. Yeah, they are maybe not interested enough, but still. Same for executive positions (not the target audience). Number of full-stack developers is surprisingly small, given that lot of developers actually somewhat of full-stack developers, sicne they have to touch lot of things.