How can Machine Learning benefit the Bid Development & Approval processes. We present 4 use cases..
We all have been reading about Machine Learning a lot in the recent months. Machine Learning, along with its other siblings like NLP, Predictive Analytics, Deep Learning and Big Data have been heralded as some of the most disruptive technologies of our times. And while it is impacting many industries and different process, does it have any promise for the bid development, pricing, and approval process. Let’s discover.
If one follows the process from the time an opportunity is qualified or an RFP/tender is released to the time it is submitted to the client, the opportunities are enormous. Let’s look at 4 in particular:
RFP parsing: This is probably a mundane one. Not if you are a bid manager. Organisations spend 100s of man hours going thru the 100-1000 pager RFPs/tenders and the associated annexures - reading them, re-reading them, tagging them, scanning them, etc. Instead if you had an ML/NLP-enabled application that would parse the RFP, identify the key sections and populate it in the respective categories. Users could then just search the application, read the summary or the full document as and when needed. Lots of labour lost.
Develop a project Risk score: How risky is this project? Have we covered ourselves enough for the major risks? These are probably the key questions every CXO asks when s/he is reviewing a new proposal. Why not develop a risk score that classifies the project into – High risk, Medium risk, Low risk. Look for parameters such as – Is it in the technical sweet spot? Is location an issue? Have we done this before or is this the first time? Are the SLAs too stringent? What is the customer history?
This requires base lining and deciding upon the parameters, as the elements could vary based on the industry and solution offering. But after these hard yards have been put in, the solution should be able to derive, calculate and classify the project into high, medium, or low risk.
Capturing review calls: The pricing decision for most turn-key projects happen on a bid review call. Post which, the minutes are recorded by the account manager or the bid manager. In many of the cases they are not recorded at all. And even if they are, some points go missing, or all views are not recorded correctly, the sales person may give a (slight) twist to the discussion especially when it comes to the approvals related T&Cs. Better then if the solution can use voice-to-text technologies and convert the conversation to text. Today voice-to-text technologies are available for multiple languages, can identify speakers, translate proper or common nouns, and also give you time stamps. Such a solution can dramatically reduce the time required for documenting the call and more importantly improve the accuracy of minutes.
Is the solution/estimation in-line: It is a common scenario that sales teams work to ensure that the solution or the estimation is just about right or better still, slightly lower. Of course, their objective is to win the deal. But we all know how the story ends. On the contrary if the estimation is high, there’s a risk of losing the deal itself. Depending on what role you are playing and what stage your organization is in, one may prefer either of the outcomes. Imagine a solution then, that can analyse the estimations of similar projects which have been successful and highlight if the current estimation is in-line or not.
This capability can also be directly used by those who are developing the estimations to make/validate them even before submitting for review. Similarly,the reviewer can know if the solution components included in the current solution are accurate or not. Or if the commercial approver wants to know, before approving, all the previous proposals or bids that were similar in terms of value or discount or margin, analytics can quickly throw up similar deals and the associated details for him/her to take an appropriate decision.
Some of the above use cases will require organisations to hold enough historical data (may be 3+ years). Standardize and maintain the data parameters in a consistent format. Define and store proper legends/tags in terms of - nature of the project, the workload, non-standard characteristics, etc.
As is true with any data-related project, majority of the effort has to be devoted to providing correct data for analysis. But with more advancements in AI/ML/NLP/Analytics technologies, one can expect new, innovative use cases coming up. Stay tuned.
Comments and observations welcome at firstname.lastname@example.org.