Related Posts
Can I join in Nokia R&D unit for java, spring boot backend developer role considering current situation of layoffs in product based companies ?
Exp - 4 years
Tech stack - Java, Spring Boot, Microservices
EPAM Systems Cisco Nokia Dell Deloitte Deloitte USI Deloitte India Infosys Cognizant KPMG EY PwC Verizon Verizon Media Ericsson Huawei Technologies
Mentor
any telltale signs of layoffs in tiger?
More Posts
Hi , need likes for DM Feature , Thank you.
Community Builder
Chief
Additional Posts in Data & Analytics Consultants
Mentor
New to Fishbowl?
Download the Fishbowl app to
unlock all discussions on Fishbowl.
unlock all discussions on Fishbowl.
Excel makes simplicity really simple and complexity really complex. I know that can be frustrating when the task at hand is complex.
I actually like excel, but I grew up on it before I got to R. Generally to avoid mistakes in excel, I recommend three things:
(1) Build the building one brick at a time. Don’t write really complicated formulas in single cells. Each formulas should be really simple to understand.
—this is going to help you keep track of each step, even the trivial ones, of your model
—troubleshooting/debugging or updating the model will be much simpler and faster
—the model will generally be more “flexible”, meaning that because each step is written out, you’ll be able to experiment more with the inputs to your model and observe their impact to your overall answer
(2) Spend a considerable of your time transforming your data. Leave your original data set unchanged, but append columns that serve as “flags”. Usually for me flags are 0/1s that reflect some criteria.
—when I write my formulas for the actual model, I’m usually interacting with the flags. I might count them or use them to filter data more easily to perform operations.
—this also allows you to perform Sanity checks more easily. If you expect the sumifs formula you wrote to equal X, then you can go into the data set, filter the Columba by the flags you added and check that the column does indeed equal X.
(3) perform sanity check regularly. It’s instinctual for me. I just constantly check that what I calculated equals what I expect. Obviously take a ton more time up front, but I promise you it’s better than being wrong. Once you’re wrong publicly, you lose all credibility. And it’s really hard to climb out of that hole.
Automate and embed check formulas
Bowl Leader
Is Excel level transparency an explicit requirement on your projects?
Likely feeling that way due to unstructured/disorganised approach, which makes you unsure of if there are gaps/errors.
- Create a structure for the workbook (data, assumptions, calculations, outputs)
- Build checks for each. Checks should be (i) independent within each part of your structure e.g. do things add up AND (ii) consistency across e.g. given your assumptions, part x of your output should be within a particular range or greater than part y. You can set up all checks to be formula driven with conditional formatting etc. I recommend also a checks tab that shows at minimum total number of checks and how many have failed (more detailed versions recommended)
- Iterate early i.e. take the proposed structure and checks you plan to build in to someone more experienced, show them the plan and do they agree/disagree. They’ll naturally chip in with things to look out for. Similarly take a first cut of output for review - don’t let the client presentation (or last thing the night before) be the first time people see it. No one will care if you put in a good, logical first effort and you iterate together. They will care if surprise them with a fuck up at a key moment, so iterate as you go.