app.py

from altair.vegalite.v4.schema import channels
from pandas.core.tools.datetimes import Scalar
import streamlit as st
import pandas as pd
import numpy as np
import altair as alt
from altair import datum

st.title('What should B2B SaaS startup do to grow the business?')

def load_data():
    data = pd.read_csv('B2B_SaaS_data.csv')

    data.rename(columns={
        'employees_sales_dec_2021': 'sales employees (dec 2021)', 
        'employees_sales_dec_2020': 'sales employees (dec 2020)',
        'employees_sales_growth_1y': 'YoY sales employees (%, dec)',
        'employees_engineering_dec_2021': 'engineering employees (dec 2021)', 
        'employees_engineering_dec_2020': 'engineering employees (dec 2020)',
        'employees_engineering_growth_1y': 'YoY engineering employees (%, dec)',
        'employees_total_dec_2021': 'employees (dec 2021)',
        'employees_total_dec_2020': 'employees (dec 2020)',
        'employees_bd_dec_2021': 'business dev employees (dec 2021)',
        'employees_bd_dec_2020': 'business dev employees (dec 2020)',
        'employees_bd_growth_1y': 'YoY business dev employees (%, dec)',
        'employees_support_dec_2021': 'support employees (dec 2021)',
        'employees_support_dec_2020': 'support employees (dec 2020)',
        'employees_support_growth_1y': 'YoY support employees (%, dec)',
    }, inplace=True)

    cols_to_keep = [
        'company',
        'ticker',
        'sales employees (dec 2021)', 
        'sales employees (dec 2020)',
        'YoY sales employees (%, dec)',
        'engineering employees (dec 2021)', 
        'engineering employees (dec 2020)',
        'YoY engineering employees (%, dec)',
        'employees (dec 2021)',
        'employees (dec 2020)',
        'business dev employees (dec 2021)',
        'business dev employees (dec 2020)',
        'YoY business dev employees (%, dec)',
        'support employees (dec 2021)',
        'support employees (dec 2020)',
        'YoY support employees (%, dec)',
    ]
    data['MRR ($1M, nov 2021)'] = round(data['MRR_1K_USD_nov_2021'] / 1000)
    data['MRR ($1M, nov 2020)'] = round(data['MRR_1K_USD_nov_2020'] / 1000)
    data['YoY MRR (%, nov)'] = (data['MRR ($1M, nov 2021)'] - data['MRR ($1M, nov 2020)'])/data['MRR ($1M, nov 2020)']
    cols_to_keep.extend(['MRR ($1M, nov 2021)', 'MRR ($1M, nov 2020)', 'YoY MRR (%, nov)'])

    # get last year number of employees from growth
    for type in ['sales', 'business dev', 'support', 'engineering']:
        col_dec_2020 = f'{type} employees (dec 2020)'
        col_dec_2021 = f'{type} employees (dec 2021)'
        col_growth = f'YoY {type} employees (%, dec)'

        data[col_dec_2020] = data[col_dec_2021] / (1 + data[col_growth])
        data[col_dec_2020] = data[col_dec_2020].round()

    # customer success and related ratios
    for year in ['2020', '2021']:
        data[f'customer success employees (dec {year})'] = data[f'business dev employees (dec {year})'] + data[f'support employees (dec {year})']
        cols_to_keep.append(f'customer success employees (dec {year})')

        data[f'customer success:sales employees (dec {year})'] = data[f'customer success employees (dec {year})'] / data[f'sales employees (dec {year})']
        cols_to_keep.append(f'customer success:sales employees (dec {year})')

        data[f'support:sales employees (dec {year})'] = data[f'support employees (dec {year})'] / data[f'sales employees (dec {year})']
        cols_to_keep.append(f'support:sales employees (dec {year})')

        data[f'support:engineering employees (dec {year})'] = data[f'support employees (dec {year})'] / data[f'engineering employees (dec {year})']
        cols_to_keep.append(f'support:engineering employees (dec {year})')

        data[f'sales:engineering employees (dec {year})'] =data[f'sales employees (dec {year})'] / data[f'engineering employees (dec {year})']
        cols_to_keep.append(f'sales:engineering employees (dec {year})')

        data[f'customer success:engineering employees (dec {year})'] =data[f'customer success employees (dec {year})'] / data[f'engineering employees (dec {year})']
        cols_to_keep.append(f'customer success:engineering employees (dec {year})')

    # YoY growth in customer success
    data['YoY customer success employees (%, dec)'] = (data['customer success employees (dec 2021)'] - data['customer success employees (dec 2020)']) / data['customer success employees (dec 2020)']
    cols_to_keep.extend(['YoY customer success employees (%, dec)'])

    data['YoY customer success:engineering (%, dec)'] = (data[f'customer success:engineering employees (dec 2021)'] - data[f'customer success:engineering employees (dec 2020)']) / data[f'customer success:engineering employees (dec 2020)']
    data['YoY customer success:sales (%, dec)'] = (data[f'customer success:sales employees (dec 2021)'] - data[f'customer success:sales employees (dec 2020)']) / data[f'customer success:sales employees (dec 2020)']
    cols_to_keep.extend(['YoY customer success:engineering (%, dec)', 'YoY customer success:sales (%, dec)'])

    # additional filters
    data['growth source'] = np.where(data['sales:engineering employees (dec 2020)'] < 0.6, 'product-led growth?', 'sales-led growth?')
    data['run-rate (estimate)'] = np.where(data['MRR ($1M, nov 2021)'] > 70, '$1B+', '<$1B')
    cols_to_keep.extend(['growth source', 'run-rate (estimate)'])
    return data.loc[data['MRR ($1M, nov 2021)'] < 160, cols_to_keep] 

data = load_data()

"""
Through analysis of public **B2B SaaS companies**, I am aiming to answer the following question. 

>**What should B2B SaaS startup do to grow the business?**
> - [ ]  Hire more engineers to build new features?
> - [ ]  Hire more salespeople?
> - [ ]  Hire more support and customer success managers to take care of existing customers?

**My hypothesis is that correct mix of employees (customer success, sales, engineering) may fuel mid-term and long-term growth of B2B SaaS startup.** 
If this hypothesis is true, then the information about current mix of employees may enhance other financial and non-financial information (`not available to general public, incl. me`) 
to make a better prediction of future revenues, while startups can make better hiring decisions. 

---------------

##### Why do I believe this hypothesis might be true? 

This hypothesis is developed after reflecting upon my past work experience.

Before becoming **Data Scientist**, I worked 2.5 years at **Microsoft** as a Sales Engineer / Customer Success Manager, 
driving new revenue and usage of **SaaS** products in the region of 9 countries. 

During my time there (2016-2018) MSFT stock grew from $50 to $110, 
which was driven by successful transformation into cloud leader and switch towards **subscription business model**.
Due to small size of the subsidiary, in addition to being responsible for driving usage of Office 365 cloud services and Windows 10 deployment,
I ended up helping to drive new revenue and usage of almost all subscription products in the enterprise segment. 

As a result, I learned few things about **B2B SaaS** business and how correct hiring decisions and right focus may impact the growth:

> 1. If you have few customers, you need to add more (quite obvious, I know).
> 2. If you forget about existing customers, you will limit the mid-term and long-term growth, because renewal will be a challenge, while upsell/cross-sell wil be impossible.
> 3. It is easier to upsell/cross-sell than to add new customer.
> 4. When you sell to companies, you sell to specific people inside the company. These people switch jobs and if they use and like your product and company, 
they will try to bring your product at their new job and will recommend to friends and colleagues (organic growth!).

With above and other learnings in mind and without access to confidential information (financials, product usage, customers, etc),
I believe just by looking at who work in the company we can get few insights, helpful for growing B2B SaaS startups.

--------------
##### Product-led growth vs sales-led growth? Less than $1B run-rate vs more than $1B run-rate?
"""
col_strategy_size_text, col_strategy_size_charts = st.columns(2)

with col_strategy_size_text:
    st.write(
    """
My choice of dataset for analysis fell on public B2B SaaS companies mostly because their financial performance (**MRR**) is available. 
The drawback of that choice is that their run-rate is likely higher than run-rate of a B2B SaaS startup that would be interested in **raising more money from VCs or through alternative funding** 
and therefore derived insights may not be that applicable.

There are **26** B2B SaaS public companies in my dataset. Besides estimate of MRR (quarterly revenue divided by 3) in November 2021 and November 2020, I collected from **LinkedIn** information such as 
number of employees, what they do (in general terms according to **LinkedIn**), YoY change in employees count, and number of followers on **LinkedIn** (main social network for B2B sales).

With the information above, I can segment my dataset further that will help in understanding employee mix better. 

> 1. From the ratio of number of salespeople to engineers (`sales:engineering`) we can guess if the company pursues **product-led growth** or **sales-led growth strategy** (or it is just slow in hiring sales).

> 2. Although all these companies are public, we still can segment them by run-rate: those with run-rate of roughly $1B+ and those who didn't achieve this milestone yet. 
The bigger company will have different challenges in growing. 
    """
    )

growth_source = alt.Chart(data).mark_circle().encode(
    x=alt.X('sales:engineering employees (dec 2021)', title='sales:engineering (dec 2021)'),
    y=alt.Y('YoY MRR (%, nov)'),
    color=alt.Color('growth source', legend=alt.Legend(orient='top'), scale=alt.Scale(scheme='set1')),
    tooltip=['company', 'ticker'],
    size=alt.Size('MRR ($1M, nov 2021)', legend=alt.Legend(orient='bottom', symbolFillColor='grey'), bin=alt.Bin(step=30))
)

small_big_split = alt.Chart(data).mark_circle().encode(
    x=alt.X('MRR ($1M, nov 2021)',),
    y=alt.Y('YoY MRR (%, nov)'),
    color=alt.Color('run-rate (estimate)',legend=alt.Legend(orient='top'), scale=alt.Scale(scheme='set2')),
    tooltip=['company', 'ticker'],
    size=alt.Size('employees (dec 2021)', legend=alt.Legend(orient='bottom', symbolFillColor='grey'), bin=alt.Bin(step=2000))
)

with col_strategy_size_charts:
    st.altair_chart(growth_source, use_container_width=True)
    st.altair_chart(small_big_split, use_container_width=True)

"""
-------------------------------------------------
##### How does revenue growth correlate with employee count growth?

For bigger companies (with $1B+ run-rate) we can see the **positive correlation between MRR growth rate and 
growth of hiring salespeople, customer success managers, and engineers.** 
There is one company in this category that seems to pursue product-led growth strategy as discussed before, 
but since this company is Cloudera (enterprise data cloud), sales-led growth strategy would fit better in my opinion.

For smaller companies, however, the picture is different.
"""

col_hiring_in_small, col_hiring_in_big = st.columns(2)

def get_growth_scatter_plot(df, col_x, col_size, size_title, bin_step, col_x_title=None, height=250, max_y=0.7, clip=True):
    x_title = col_x_title if col_x_title else col_x
    chart = alt.Chart(df).mark_circle(clip=clip).encode(
        x=alt.X(
            col_x,
            title=x_title,
        ),
        y=alt.Y(
            'YoY MRR (%, nov)', 
            scale=alt.Scale(domain=(0,max_y), clamp=False),
            axis=alt.Axis(values=[0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1]),
        ),
        color=alt.Color(
            'growth source', 
            legend=alt.Legend(orient='top'), 
            scale=alt.Scale(scheme='set1')
        ),
        tooltip=['company', 'ticker', 'YoY MRR (%, nov)', col_x, col_size],
        size=alt.Size(
            col_size, 
            legend=alt.Legend(
                orient='bottom', 
                symbolFillColor='grey', 
                title=size_title
            ),
            bin=alt.Bin(step=bin_step)
        )
    ).properties(
        height=height,
    )
    return chart

engineering_hiring_in_small = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']!='$1B+'],
    col_x='YoY engineering employees (%, dec)',
    col_size='engineering employees (dec 2021)',
    size_title='employees in engineering',
    bin_step=200,
    max_y=0.6
)

sales_hiring_in_small = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']!='$1B+'],
    col_x='YoY sales employees (%, dec)',
    col_size='sales employees (dec 2021)',
    size_title='employees in sales',
    bin_step=200,
    max_y=0.6
)

support_hiring_in_small = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']!='$1B+'],
    col_x='YoY customer success employees (%, dec)',
    col_size='customer success employees (dec 2021)',
    size_title='employees in cust success',
    bin_step=200,
    max_y=0.6
)

engineering_hiring_in_big = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']=='$1B+'],
    col_x='YoY engineering employees (%, dec)',
    col_size='engineering employees (dec 2021)',
    size_title='employees in engineering',
    bin_step=400
)

sales_hiring_in_big = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']=='$1B+'],
    col_x='YoY sales employees (%, dec)',
    col_size='sales employees (dec 2021)',
    size_title='employees in sales',
    bin_step=400
)

support_hiring_in_big = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']=='$1B+'],
    col_x='YoY customer success employees (%, dec)',
    col_size='customer success employees (dec 2021)',
    size_title='employees in cust success',
    bin_step=600
)

with col_hiring_in_small: 
    st.caption('Companies with less than $1B run-rate')
    st.altair_chart(engineering_hiring_in_small, use_container_width=True)
    st.caption('Companies with less than $1B run-rate')
    st.altair_chart(sales_hiring_in_small, use_container_width=True)
    st.caption('Companies with less than $1B run-rate')
    st.altair_chart(support_hiring_in_small, use_container_width=True)

with col_hiring_in_big:
    st.caption('Companies with roughly $1B+ run-rate')
    st.altair_chart(engineering_hiring_in_big, use_container_width=True)
    st.caption('Companies with roughly $1B+ run-rate')
    st.altair_chart(sales_hiring_in_big, use_container_width=True)
    st.caption('Companies with roughly $1B+ run-rate')
    st.altair_chart(support_hiring_in_big, use_container_width=True)


"""
---------------------------------
##### How does revenue growth correlate with employee mix?
"""

col_ratio_text, col_ratio_charts = st.columns(2)

sales_engineering_ratio_in_small = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']!='$1B+'],
    col_x='sales:engineering employees (dec 2020)',
    col_size='engineering employees (dec 2020)',
    size_title='employees in engineering',
    bin_step=150,
    col_x_title='sales:engineering (dec 2020)',
    height=300,
    max_y=0.5
)

success_engineering_ratio_in_small = get_growth_scatter_plot(
    df=data[data['run-rate (estimate)']!='$1B+'],
    col_x='customer success:engineering employees (dec 2020)',
    col_size='engineering employees (dec 2020)',
    size_title='employees in engineering',
    bin_step=150,
    col_x_title='customer success:engineering (dec 2020)',
    height=300,
    max_y=0.5
)

with col_ratio_text: 
    st.write(
    """
Data shows that simply hiring more salespeople, engineers or customer success managers/support doesn't 
positively correlate with MRR growth for smaller companies as opposed to those who reached milestone of $1B run-rate. 

However, what seems to be relevant at this stage of growth are 
* the ratio of salespeople to engineers (`sales:engineering`). 
* the ratio of customer success managers to engineers (`customer success:engineering`)

Sales-led strategy means by definition that you need to have strong sales organization to grow and 
it makes sense that already public companies who found product-market fit are scaling by having more salespeople.

Regardless of growth strategy choice, it seems like that having more customer success employees who will **support existing customers 
with deployment, usage, and troubleshooting** correlates with revenue growth.
    """
    )

with col_ratio_charts:
    st.caption('Companies with less than $1B run-rate')
    st.altair_chart(sales_engineering_ratio_in_small, use_container_width=True)
    st.caption('Companies with less than $1B run-rate')
    st.altair_chart(success_engineering_ratio_in_small, use_container_width=True)  

"""
---------------------------------------------
##### So, is my hypothesis true? What is the recommendation then?

Any company who wants to be successful needs to be obsessed about its customers. 

In B2B SaaS space it means that, in addition to having a product that solves customers' problems, 
a B2B SaaS company should have **enough employees to support the customers** in their interactions with the product. 
My data seems to show how many such employees is enough to have in comparison to engineers.

To make it even more relevant and actionable for private B2B SaaS startups, 
it would be interesting to analyze **employee mix in comparison to number of customers and contract size**, 
so that early-stage B2B SaaS startups can plan their hiring decisions accordingly.
"""

st.caption('analyzed and written by Adilet Gaparov')
st.caption('https://adiletgaparov.com')