Optimising Sales - A Data Science and Analytics Challenge
By James Lea
Cash is king, and sales drives an organisation. But how can we win most efficiently? In this piece I set out a data science and analytics challenge using UK public sector contract data.
Sales
Winning work is central to business. If we don’t win, we have no revenue, and – ultimately - no pay cheque at the end of the month.
Selling projects is as important as delivering projects. One cannot exist without the other.
In the project data analytics community we look at project delivery data, and use this to understand our projects better, so that we can deliver them more effectively.
How can we extend and apply those techniques to improve our ability to win work?
This challenge deliberately spans the gap between sales and delivery.
- How can we exploit the data that’s now available to maximise the payback from sales?
- What can we do to identify opportunities that are worth pursuing
- How can we ‘qualify out’ (“no bid”) opportunities that are likely to be unsuccessful, to focus our limited resources elsewhere? “Cost of sales” – we cannot win everything.
The challenge is to see how we can use past sales data to determine how we should respond to contracts being tendered. Should we bid or qualify out (no bid)? We will use a combination of predictive and prescriptive analytics to work out the best solution.
Scenario / Use Case
In this challenge we’ll imagine we’re a sales professional working for a Digital Outcome Specialist. We’re seeking contracts with the UK public sector bodies.
We’ll look at live opportunity data – for real contracts being offered to the market today – and see how we can generate insights that maximise our wins over cost of sales.
Analysis Techniques
This challenge requires us to think about:
- business: how this relates to the goals of the organisation
- optimisation: how to maximise gain while minimising cost
- how to interpret the data: this is real-world data with gaps and ambiguities– process mining techniques will help infer the steps and meaning of the data
- presentation: how to make the insights we generate easy to use by sales professionals
Data
The UK government lists upcoming contracts through its Digital Marketplace.
Within this marketplace, we will look at the contracts being offered for Digital Outcomes and Specialists.
There are three sets of data:
Data set | URL | Remarks |
---|---|---|
Live opportunities | Live data | This contains live data: contracts being tendered. This is the data we want to make predictions on. |
Historical opportunities | Training data | This contains past contract awards (“Closed opportunity data”). Use this to train your models on. |
Opportunity details for live and historical opportunities | Opportunity ID 14793 details | We can access a further page of details for that opportunity using this link. Essential skills and experience may be useful. Look at ‘Specialist’ skills to see what skills required to win. |
As of July 2021, the historical (ie algorithm training) data set contains 4273 records, including:
- The opportunity and its status
- The public sector buyer
- When the opportunity was advertised
- How many questions were raised
- How many organisations applied
- The winning supplier
- The awarded contract value
Analytics Toolchain
A suggestion to get you started:
Phase | Techniques |
---|---|
Extract | Spreadsheet (data set 2), Beautiful Soup (data sets 1 and 3) |
Transform | Process mining (data sets 2 and 3) |
Load | Tools of your choice – Spreadsheet / PowerBI / Pandas / SQLite |
Challenge Outputs
There are two groups of output, both requiring visualisations as well as numerical data:
- Descriptive analytics dashboard
- Predictive/prescriptive analytics dashboard
Descriptive Analytics
This is what the historical data tells us – a useful guide to the future, but requires further interpretation and analysis by the user to work out what they should do.
Standard suite:
A report or interactive dashboard showing summary statistics of:
- live opportunities
- closed opportunities
- links to the live data via https://
- biases/clusterings in win prices (eg at the £10M mark)
- amount of economic activity, by buyer organisation
- who’s won the most work?
- for given types of work, who tends to bid, and who tends to win?
Advanced suite:
A report or dashboard showing the outcome of statistical and/or machine learning techniques, that indicate:
- for a given type of opportunity, what the ‘win probability’ is
- how the win probability alters depending on various criteria (I call this: “the win surface”)
Predictive and Prescriptive Analytics
For specific live opportunities, what are our chances (prediction), and can we do to maximise our chances of winning (prescription)?
Using the live data set, imagine your team represents a particular type of organisation – either select one, or ‘invent’ one (whichever is easiest).
From this vantage point, generate reports or an interactive dashboard showing, for each live opportunity:
- The most likely winners based on the training data set
- If your organisation were to bid this, who might the competitors be, and what are their win probabilities?
- What skills / price range etc might you need to beat them?
- What your organisation could do to improve its chance of winning work, based on the past data
This part is open to you: devise the questions and answers that demonstrate how we can optimise sales using project data analytics techniques.
Get in touch
If you’d like to discuss this challenge in more detail (including some detailed criteria, and how the analytics could be developed), please get in touch.