Which matters more, your markup or your prep time?
Forty bids scored on four things you know before you press send. Every point of markup cuts the win odds by about a third and every extra week of preparation nearly quadruples them, and the prior relationship every estimator swears by does not clear the bar.
Construction Intermediate Logistic Regression free
After you install, this is the model to open.
What Makes Us Win a Bid?
- In your spreadsheet, click the Sortia icon in the strip of icons down the right-hand edge. No strip? Click the arrow at the bottom-right to open it. You can also use Extensions, then Sortia, then Open Sortia.
- Click Start from a template and put that name in the search box.
- Pick the card with that name and click Load this template. It arrives on a new tab with real numbers already in it.
The answer
- One more week of prep
- 3.82x the odds of winning, p = 0.035
- One more point of markup
- 0.63x the odds; a point of markup cuts them by about a third, at p = 0.066
- Accuracy
- 70% on the 10 tenders kept back, exactly what calling every tender a loss scores; 93% on the 30 it fitted
- Fit
- 0.526 McFadden R squared on the 30 tenders it fits, with 10 kept back
Forty tenders, four things you know before you press send, and whether each one won. Twelve of the forty were won, so guessing "we lose" every time already scores 70%. Click Run: the model scores 93.3% at a 0.5 cutoff on the 30 rows it trains on, calling 8 of the 9 wins and 20 of the 21 losses, with a McFadden R2 of 0.526, and 70% on the 10 rows it kept back.
Read the odds-ratio column rather than the coefficients, because it is the one that speaks the language of a tender review. Weeks to prepare comes back at 3.82 with a p-value of 0.035, the only one of the four inside the bar, so every extra week of preparation nearly quadruples the odds of winning, and lead time is bought by qualifying earlier rather than by giving up margin.
Markup sits at 0.63 with a p-value of 0.066 and competitors invited at 0.45 with a p-value of 0.077, both pointing the way you would expect and both just outside the bar on forty bids. Having worked for the client before comes back at 0.57 with a p-value of 0.70, which is the one result on this sheet that says nothing at all. A coefficient that misses significance is a statement about your sample size at least as much as about the world, and the fix is another year of tenders, not a policy.
Notice that the bid value column sits outside the training range on purpose. Logistic regression reads the outcome from the LAST column of the range you give it, so anything to the right of the won-or-lost column is carried for reference and never modeled. To price a live tender, type its four numbers into an empty block on the sheet and put that block in Predict rows: the model returns a win probability, and multiplying it by the margin at that markup gives you an expected value you can compare across markups.
What the model cannot see is who else was bidding and how hungry they were. To use your own tender log, keep the won-or-lost flag as the last column inside the range and widen the range to your rows.
The model
It arrives on a tab called Template: What Makes Us Win a Bid, carrying these columns:
- Markup (%)
- Weeks to prepare
- Worked for this client before (1 = yes)
- Competitors invited (count)
- Won (1 = yes)
- Bid value ($K)
Once it is in your sheet
- The model arrives with real numbers in it and runs as it stands, so you can press the button first and understand it second.
- Change the numbers to yours. The sheet marks which cells are inputs and which hold formulas, and most labels carry a note explaining the row.
- Press the run button at the bottom of the panel. It is labeled for the tool you are in, and the result lands on its own tab, with a written reading of it beside the figures.
Never used Google Sheets? Start here goes the whole way, in seven steps, and assumes nothing.
Next question
- What is one more person on the gang worth?What Explains How Much a Crew Gets Done?
- Are you paying for machine days you never use?Which Plant Do We Hire and Which Do We Own?
- Which tenders should you not price at all?Score the Sites Before You Bid
- Do rain days really cost you money?What Moves With Cost on This Job?
- Is one estimator quietly pricing higher than the other?Do Our Two Estimators Price the Same Job the Same Way?
- What will a tonne of steel cost you next month?Where Is This Material Price Going?
Every model like this one, and the method behind them: Machine learning in Google Sheets.