The lion’s share of responses was from software companies working across various industries - this was not unexpected given that my network is heavily tech-inclined.

Nearly half of the responses were from well-established companies (10 years or more). Nearly 25% of all responses were from early startups (less than 5 years).

A similar bias is seen towards larger companies (by employee count) in the responses.

Almost 60% of responses were from funded companies.

However, an even larger number was from profitable companies. I’ll admit that this caught me by surprise.

Frankly, I didn’t expect the red (“No it doesn’t”) to be that huge.

Basically all but two companies follow annual reviews and appraisals. Seems like a pretty established convention.

The reds are people who are managers, blue are individual contributors. Most of the responses have come from the 3 to 8 years range.

Also, though this is a ridiculously small sample set to base any inference on, the proportion of managers is higher with more experienced people (duh 🤷).

Note: I don’t know if the 4% “Bachelors” were engineers or not, so I left them as independent category.

Two thirds of the responses were Individual Contributors.

This is an interesting data point, because it tells me a couple of things:

• It is more a representation of what employees have experienced (or think is fair) versus what employers have given out (or consider fair)
• Consequently, (and I’m making a big assumption here), I expect the min-max percentage range performance increment data to be slightly more biased towards right.

For designations, over half of the respondents were software engineer ICs. I have also combined some more designations into the same grouping where appropriate. That said, there were a few which didn’t directly fall into any group, so I kept them as is.

Smaller proportion of responses in the less than 5L and more than 50L category (unsurprisingly), the rest seems pretty evenly distributed.

I tried to come up with a good visualization for showing this information coherently – and since it was a while I’d used python+seaborn for data viz, I got carried away and tried a bunch of different things. I’m sharing all the charts below - they all say the same thing (give or take), you can read the ones you are most comfortable with.

• Fig 1: Histogram

  The simplest one to understand is a histogram. However, there’s way too much overlap, which makes it harder to read. X axis is the percentage increment, Y axis is the counts (number of occurrences of each % number).

  As you might see, 5% and 10% are clear favorites (which makes sense given our tendency and liking to have nice round numbers).

  

• Fig 2: Line chart (KDE function)

  Basically the same histogram, programmatically converted to a smooth line chart.

  Note that in this case, the Y axis is not the count, just the relative occurrence.

  Other than the lack of counts, this chart might be the simplest to read. We can clearly see that the red (Low Performer) peak is left-most at ~5%, followed by Average at ~10%, High at ~12% and Outstanding just a bit higher at 16%.

  Note that these are modes; the actual range of numbers was super wide. For example there were actually a couple of entries for 50-60% increments (‼️) for outstanding performers, so just looking at the peaks might not paint the correct picture.

  (As a side note, I personally believe 50-60% increments might actually be signs of a pay and/or role correction, not an actual “annual increment”, but that’s just my opinion)

  

• Fig 3: Violin Plot

  This is one of my favorite charts, for multiple reasons.

  • It looks nice
  • I was able to separate out the “min” and “max” for each performance category so they can be looked at separately
  • It provides the 25th, 50th and 75th percentiles for each range at a glance
  • The variation range / outliers for each category
  • Shows the comparison between performance categories just as effectively as the KDE chart.

  If you look closely, there is much more information to be drawn out here:

  • The median for low and average performers is about 5% and 10% respectively. This matches what we saw earlier.
  • For high performers, however, the median (thick dashed line) for the min is different from the max. I don’t know if it’s the mathematically accurate way of saying it, but the median range seems to be 12-15%
  • For outstanding performers, the median increment is 20%. Also, the long length of the violin kinda shows the variation in the numbers (though you could technically infer this from the long-tail of the KDE as well). More importantly, if you look at the 75th percentile, you can infer that 25% of the people said that outstanding employees make 30% increments or more.

  

This was surprising for me. I expected significant deviation here, but it seems from the KDE plot that (at least in our sample set), the employee’s current salary does not significantly affect % changes.

(histogram and violin plots weren’t very legible, so we’ll stick with this plot here)

It would seem from the histogram below that ICs score the most of the high increments, but do remember that two-thirds of our respondents were ICs, so the data is likely biased for that (esp. if you were comparing counts across categories).

It might be more useful to consider each independently, so we use the KDE plot (below). Here also, surprisingly, it seems that the peaks are pretty much aligned around 10%. Do note that this is not further segregated by performance category.

The natural next question to ask is: do the increments change with time (years of experience) ? In my head, they would taper down (because 20-30% increases every year would probably compound to a hell lot of money, especially given tech salaries these days.

Let’s see what the data says…

We plot the percentage increment on the X axis, as previously done. Additionally, we plot the Experience (years) on the Y axis. The dots are color coded by performance category, as shown in the legend.

Observations:

• The chart becomes sparser as you go top-right, which matches my expectation above
• Except for a couple of stray data points, 30% seems to be an solid barrier for anyone over 6 years of experience. Put another way, only people who are ≤ 6 years have any increment over 30%. This leads me to further strengthen my assumption that the higher percentages are more likely salary/role corrections instead of standard increments.

Now that we have this chart, we could use it to study other characteristics… See next section.

Once again using the scatter plot, with X as percentage, Y as experience, and using colors and shapes for role and performance respectively.

Personally, I found it too overwhelming, so I filtered out the ICs, and the chart becomes relatively readable. Two things I notice are:

• ICs were indeed the bulk of the dots over 30%
• If you look at the blue plus/square symbols, it looks like Founders like giving themselves a higher raise once in a while 😅 No judgements though, they might be correcting their own paltry pay, who knows.