Let me tell you how I prepared a spreadsheet for my own state of Connecticut, and it may be helpful to frame the best answer to your question.
First, I had to determine whether I should count cases or deaths to determine growth rate. The number of new cases is affected by at least three things: the actual infectiousness of the virus; the efficacy of our countermeasures (social isolation and sanitation/disinfection); and, crucially, the amount of testing that is done. Deaths, on the other hand, I would not expect to be so affected by our testing regimen. At this point, the number of deaths is still low enough that day to day fluctuations can distort the identification of any underlying trend. To combat this, I use a 4 day rolling average.
Then, I had to choose a measure that would detect progress or lack thereof. As you have done, I chose to compare the percentage increase both in new cases compared to cumulative cases and in new deaths compared to cumulative deaths, using a 4 day rolling average for each. I also calculated the doubling time based on the percentage increase.
What the data show, for Connecticut, is that over the past four days the daily increment has decreased from 29% to 21% and doubling time has increased from approximately 2.5 days to 3.4 days. I could simply extrapolate from today's numbers, but I think that would lead to inaccurate results, as I expect that the underlying trend will continue. That is, the percentage of new deaths every day will decrease and the doubling time will increase. Plotted on a logarithmic scale, the curve would be starting to flatten.
ETA: one important caveat is that if the medical system gets overwhelmed and they run out of ventilators, the death rate probably will be distorted upward. Combined with improved testing, eventually the case rate might be a better measure.