survey Domain Optimal Pricing Research - NP Survey

[Ostrados]

Hi,

I am currently doing a research on optimal domains pricing, the goal of the research is to:

1. Determine sell through rate (STR) for different pricing ranges
2. Derive a formula for the relationship between STR and Domain Price
3. Using formula from #2 we can make simulation for different pricing scenarios to find optimal pricing strategy
4. Ultimately deriving a domain pricing formula (in a separate future study).

This study requires a lot of data, I collected some data using Dofo, Sedo, Namebio.

Here is a sneak peak of what I got so far:

However the results I got are inconclusive, and I need more accurate data.

To get more accurate results I am asking Namepros community to help me collect more data, and for that I have created the following Survey:

https://freeonlinesurveys.com/s/OqGsfT2s

* The survey is totally anonymous there is no way to tell who sent it.

* The survey is for all extensions and not specific to .com

* The results of the study will be published at Namepros, I believe the results will be insightful for all of us.

* Contribution will be greatly appreciated especially from big portfolio sellers and from marketplaces that have enough data.

@Sedo @GoDaddy @DAN.COM @LaszloSchenk @GrantP @James Iles @DaaZ @aoxborrow
@AbdulBasit.com
@bmugford
@Recons.Com
@JudgeMind
@xynames
@twiki
@Acroplex
@Bob Hawkes
@Name Trader
@MadAboutDomains
@tonyk2000
@ResoluteDomains
@Leo Angelo
@Nikul Sanghvi
@TERADOMAIN
@Yusupbabay

...and all others please contribute.

Thanks

 

[Bob Hawkes]

This is a superb initiative seeking to answer a question of great importance to domain investors - how to optimize their pricing for return.

I encourage all who can to fill out the survey to expand the data. It is pretty fast to complete, and as noted anonymous. It asks how many sales in different price ranges in 2021 and 2022, and your portfolio numbers broken down by BIN.

I look forward to seeing the results.

Thank you for this initiative, @Ostrados.

-Bob

 

[Recons.Com]

Hi,

I am currently doing a research on optimal domains pricing, the goal of the research is to:

1. Determine sell through rate (STR) for different pricing ranges
2. Derive a formula for the relationship between STR and Domain Price
3. Using formula from #2 we can make simulation for different pricing scenarios to find optimal pricing strategy
4. Ultimately deriving a domain pricing formula (in a separate future study).

This study requires a lot of data, I collected some data using Dofo, Sedo, Namebio.

Here is a sneak peak of what I got so far:
Show attachment 230253Show attachment 230254

However the results I got are inconclusive, and I need more accurate data.

To get more accurate results I am asking Namepros community to help me collect more data, and for that I have created the following Survey:

https://freeonlinesurveys.com/s/OqGsfT2s

* The survey is totally anonymous there is no way to tell who sent it.

* The survey is for all extensions and not specific to .com

* The results of the study will be published at Namepros, I believe the results will be insightful for all of us.

* Contribution will be greatly appreciated especially from big portfolio sellers and from marketplaces that have enough data.

@Sedo @GoDaddy @DAN.COM @LaszloSchenk @GrantP @James Iles @DaaZ @aoxborrow
@AbdulBasit.com
@bmugford
@Recons.Com
@JudgeMind
@xynames
@twiki
@Acroplex
@Bob Hawkes
@Name Trader
@MadAboutDomains
@tonyk2000
@ResoluteDomains
@Leo Angelo
@Nikul Sanghvi
@TERADOMAIN
@Yusupbabay

...and all others please contribute.

Thanks

The data from actual sales from open sources won't help.

You are looking for correlation of sales % based on price points. But all you have is ... price.

You don't know how many of the names in the same class and price did the seller hold, hence you cannot deduce %.

That is why the experiment and research really be conducted by a large portfolio holder. And this is one of the biggest motivations for me to collect big portfolio. I could invest the same amount into high cost domains and earn similar returns, but opted for the model with more names as that provides better data set.

So, yes, I will conduct a/b/c test this year with about 2000-3000 domains in each where I will do +20%/0/-20% and see the results. I will be testing for the names currently priced at $2500. The graph should already give some food for thought and might direct me to test lower pricing points to check if there any inflection points there as well.

PS another issue with the charts above is that it is lumping all types of names together. E.g. if you have 1000 LLL.coms and price them $5000 each, your STR would be 100%. But if you take some random 5L.coms and price them $1000, your STR might be 0.1%.

For pure reliable experiment, it has to be fully randomized price change for the baskets of names priced the same prior to experiment. And the data is to be collected at the same long enough time frame for big enough data set.

 

[Ostrados]

I forgot to mention that the Survey will expire in 14 days

 

[Bob Hawkes]

You don't know how many of the names in the same class and price did the seller hold, hence you cannot deduce %.

Thanks for your comments, but I think that is exactly why he is asking people to complete the survey. In it you report how many names you have for sale in different price categories, and how many sales in those same bins for 2021 and 2022. Sure, it is self-reported, with all the caveats of that, but it is directly applicable to what he is trying to measure/
-Bob

 

[Ostrados]

The data from actual sales from open sources won't help.

You are looking for correlation of sales % based on price points. But all you have is ... price.

You don't know how many of the names in the same class and price did the seller hold, hence you cannot deduce %.

That is why the experiment and research really be conducted by a large portfolio holder. And this is one of the biggest motivations for me to collect big portfolio. I could invest the same amount into high cost domains and earn similar returns, but opted for the model with more names as that provides better data set.

So, yes, I will conduct a/b/c test this year with about 2000-3000 domains in each where I will do +20%/0/-20% and see the results. I will be testing for the names currently priced at $2500. The graph should already give some food for thought and might direct me to test lower pricing points to check if there any inflection points there as well.

That's why I need data from this survey.

I compared results from my own portfolio with the 2 graphs in my first post and it gave similar curve and close numbers.

But my portfolio is relatively small (size between 500 to 700 domains in last 3 years ) so I need more data for sure.

 

[Recons.Com]

Thanks for your comments, but I think that is exactly why he is asking people to complete the survey. In it you report how many names you have for sale in different price categories, and how many sales in those same bins for 2021 and 2022. Sure, it is self-reported, with all the caveats of that, but it is directly applicable to what he is trying to measure/
-Bob

I agree that it seems fine at first glance.

But... If an investor reports data for names priced at $900 and $2500, he is basically reporting data for different quality of names. So it is not apples to apples and STR reflects not only change in price but in quality as well.

It is like trying to deduce some information based on pricing of BMW at $60k vs Toyota at $30k. Of course, if BMW priced its cars at $30k, its sales would go through the roof, while if Toyota raised its price by just 1/3, its sales would collapse.

Again, the true accurate result can be obtained only by randomly changing prices for the names currently priced at $X price point.

As to the charts above, how did Ostrados even make the chart for Dofo? Did he take every name listed at $2000, e.g., and somehow arrive at STR 1.5%? How? We all know that only small fraction of names get reported. Let's say 10% of them. So is he saying the true STR of names at $2000 is 15%? Wow. That would be fantastic and we all could be making 1000% return annually. These results are not only accurate, they are flat out misleading.

 

[Recons.Com]

That's why I need data from this survey.

I compared results from my own portfolio with the 2 graphs in my first post and it gave similar curve and close numbers.

But my portfolio is relatively small (size between 500 to 700 domains in last 3 years ) so I need more data for sure.

Again, that data won't help. You check price elasticity by changing price and checking sales for the SAME goods, not lumped up basket of various ones without controlled price change.

So, again, you need DYNAMIC data, not STATIC.

 

[Bob Hawkes]

But... If an investor reports data for names priced at $900 and $2500, he is basically reporting data for different quality of names. So it is not apples to apples and STR reflects not only change in price but in quality as well.

Again, the true accurate result can be obtained only by randomly changing prices for the names currently priced at $X price point.

I completely agree with both of these points. Any study has caveats, and the important one here is that the quality of the names in the different bins is not the same.

Still, I think it will be interesting to see of those who list names in different price points, what the STR is. So I support the research. While in principle it would be better to take same names and do some A/B/C testing, one needs a huge number of names, and willingness to price what might not be considered optimum, so probably not many test that.

I often wonder what tests the big players, like BD and HD and DM do along these lines.

-Bob

 

[Dane]

Awesome.

 

[Recons.Com]

I completely agree with both of these points. Any study has caveats, and the important one here is that the quality of the names in the different bins is not the same.

Still, I think it will be interesting to see of those who list names in different price points, what the STR is. So I support the research. While in principle it would be better to take same names and do some A/B/C testing, one needs a huge number of names, and willingness to price what might not be considered optimum, so probably not many test that.

I often wonder what tests the big players, like BD and HD and DM do along these lines.

-Bob

It might be interesting to see, but completely useless ))

Someone like @AbdulBasit.com can achieve 1.5% STR at $5000+ average, while another investor will have 0.8% STR at $1500 average.

So, again, without actually looking at the names and grouping them into separate data sets based on quality, the data will be misleading and will be shaped only by the bias towards one of the subsets in the input. <selfpromotion>Incidentally, just got 0bias.com into my portfolio ))</selfpromotion>

In contrast, compare with this type of data:

6000 domains were priced all at $2500 prior to the experiment. 2000 of them were reduced to $1950 and 2000 of them were raised to $2950 (not quite +/- 20%, but observing the psychological limits as well for my own purposes), and 20% were kept at $2500. After N-months here are STRs for each group: x%, y%, z%.

Information like that could be actionable. I'd pay $1000 right now for info like that, as I am losing way more every month without that either by underpricing or overpricing and hurting STR.

Now I don't know how that randomly lumped data can help anyone with any conclusions about pricing besides "raising prices reduces sales quantity".

 

[AbdulBasit.com]

I've now a little over 8,100 domains portfolio.

I've some 42% of domains priced at $9,888 and above.

Remaining some 58% are priced between $1,988‐$6,888.

My STR in 2022 was around 1.5%

My average sale price last year was over $6,000.

Had the worst year (2022) considering the numbers and quality of domains.

 

[Recons.Com]

I've now a little over 8,100 domains portfolio.

I've some 42% of domains priced at $9,888 and above.

Remaining some 58% are priced between $1,988‐$6,888.

My STR in 2022 was around 1.5%

My average sale price last year was over $6,000.

Had the worst year (2022) considering the numbers and quality of domains.

Thanks for sharing, Abdul.

This confirms my point that STR/price correlation is heavily dependent on the quality and no conclusions can be drawn just by bulking all the numbers together.

Btw, what portfolio size did you use for STR calculation: beginning of year size, mid year size, end of year size, weighted average?

 

[AbdulBasit.com]

Thanks for sharing, Abdul.

This confirms my point that STR/price correlation is heavily dependent on the quality and no conclusions can be drawn just by bulking all the numbers together.

Btw, what portfolio size did you use for STR calculation: beginning of year size, mid year size, end of year size, weighted average?

Mid year size.
Thanks!

 

[Ostrados]

@Recons.Com

First I would like to issue this statement: This study is for educational purposes only, and this subject will continue to be an ongoing research after publishing results.

As I mentioned in my first post there are 4 objectives in this study, the first objective is to find the relationship between price and STR. We do that by dividing domains in subsets of prices ranges and we find the STR for each subset.

You raised valid concerns:

1. Not enough data from open sources like Dofo & Namebio
2. Data noise from underpriced & overpriced domains
3. No point in determining STR for price range
4. You need testing to find better price
5. Each domains portfolio has unique qualities

There are many details but I will try to explain as much as I can.

1. Not enough data:

While I admit that the results I have so far are far from being conclusive, but what I wanted to see when I started researching this topic is a non-linear relationship between STR & Price (I would be surprised if it was linear). And I got same non-linear curve shape regardless of source of data! which indicates that I might be in the right direction.

* Regarding the data I got from Dofo, I compared listed BIN domains at Dofo with sales at Namebio, first results showed same curve relation as in my first post, but the results were off by big margin, for example $1000-$2000 price range had 10% STR. I fixed this by:

• I found an STR points from my own sales data, I picked STR=3% point and corresponding average selling price as a reference
• I did a linear transformation by simple shifting the curve until the 3% STR touched at same price.
• All the STR values shifted and were very reasonable after the shift.

* Similarly I repeated the same for only Sedo and the curve shape also matched Dofo but with different mathematical equation.

* I repeated with my own portfolio data and the results was very close to Sedo!

And again the above results are not reliable and I need more data, that's why this survey will greatly help getting more conclusive answers

2. Data noise from underpriced & overpriced domains:

There is ofcourse the problem of overpriced & underpriced domains, but if we apply a Gaussian distribution then we can hypothesize that there are equal number of underpriced domains and overpriced domains, and so they will cancel out in the output result if we take enough big data.

Example: Let's say for the range [1000-2000] there is 20% overpriced domains that reduced average STR, then there is also 20% underpriced domains that increased the STR, so they will cancel each other in final STR.

3. No point in determining STR for price range

Domain prices are elastic, unlike cars where you can simply find exact comparable prices for example 2020 Toyota civic, in domains each domain is unique and there is no exact comparables. In the case of cars you can sell +10% max of the market value, while in domaining you can sell at -/+ 100% or more of the market value.

When we appraise domains we usually refer to ranges, so we say this domain is a low x,xxx, what does that mean? Some domainers will list that domain at $1000 while other may list it $3000. Because of that I believe that it is very useful to find STR for each price range.

In general their is an optimum price range for any domain (regardless of it is value) where ROI is maximized. If we go far above that range then domain STR will drop significantly, if we go too low then we lose revenue.

Please note that we need also to define STR cutoff limit (under study in my research), which is the limit after which STR drops very low, for example if we take a domain that has a value of $1000-$3000 and increase it's price to $10,000 then we reduce selling chance to near 0.

4. You need testing to find better price

I can't agree any more, and I repeat please do your own test, don't rely on my data or anybody else data.

The important thing from the results that I hope to get from this study, is not static STR values but the relative ones. For example you said you want to do +/- 20% test for domains priced at $2500, I don't know your STR but according to my "draft" formula your STR will:

Increase by 9% at $2000 (ex: for 2% STR => 2*1.09 = 2.18%)
Reduce by 7% at $2500 (ex: for 2% STR => 2*0.93 = 1.86%)

Ofcourse that can be totally wrong, so it is better to test yourself, and I highly encourage you to do that.

5. Each domains portfolio has unique qualities

True. Each domain portfolio is different, but on average most domain portfolios are of relatively similar quality. Some unique portfolios have different STR profile, but such portfolios are rare. So the idea of this Study is to find General STR formula, not specific formula for a specific portfolio.

In the final study results I can post an Excel template that you can use to derive STR formula for your own portfolio data.

Finally, this survey is to derive STR data only, we are still still far from talking about pricing strategy, we will use this formula to make simulation of different pricing strategies, which should be only educational and informational, you should do your own test, it is your portfolio and you know it better than anyone else.

 

[Ostrados]

Mid year size.
Thanks!

Thanks for sharing,
Actually there is a better more accurate method that I use, but it is complicated and hard to implement, so I might make a script to do it (I will share if one day I make it).

 

[Recons.Com]

@Recons.Com

First I would like to issue this statement: This study is for educational purposes only, and this subject will continue to be an ongoing research after publishing results.

As I mentioned in my first post there are 4 objectives in this study, the first objective is to find the relationship between price and STR. We do that by dividing domains in subsets of prices ranges and we find the STR for each subset.

You raised valid concerns:

1. Not enough data from open sources like Dofo & Namebio
2. Data noise from underpriced & overpriced domains
3. No point in determining STR for price range
4. You need testing to find better price
5. Each domains portfolio has unique qualities

There are many details but I will try to explain as much as I can.

1. Not enough data:

While I admit that the results I have so far are far from being conclusive, but what I wanted to see when I started researching this topic is a non-linear relationship between STR & Price (I would be surprised if it was linear). And I got same non-linear curve shape regardless of source of data! which indicates that I might be in the right direction.

* Regarding the data I got from Dofo, I compared listed BIN domains at Dofo with sales at Namebio, first results showed same curve relation as in my first post, but the results were off by big margin, for example $1000-$2000 price range had 10% STR. I fixed this by:

• I found an STR points from my own sales data, I picked STR=3% point and corresponding average selling price as a reference
• I did a linear transformation by simple shifting the curve until the 3% STR touched at same price.
• All the STR values shifted and were very reasonable after the shift.

* Similarly I repeated the same for only Sedo and the curve shape also matched Dofo but with different mathematical equation.

* I repeated with my own portfolio data and the results was very close to Sedo!

And again the above results are not reliable and I need more data, that's why this survey will greatly help getting more conclusive answers

2. Data noise from underpriced & overpriced domains:

There is ofcourse the problem of overpriced & underpriced domains, but if we apply a Gaussian distribution then we can hypothesize that there are equal number of underpriced domains and overpriced domains, and so they will cancel out in the output result if we take enough big data.

Example: Let's say for the range [1000-2000] there is 20% overpriced domains that reduced average STR, then there is also 20% underpriced domains that increased the STR, so they will cancel each other in final STR.

3. No point in determining STR for price range

Domain prices are elastic, unlike cars where you can simply find exact comparable prices for example 2020 Toyota civic, in domains each domain is unique and there is no exact comparables. In the case of cars you can sell +10% max of the market value, while in domaining you can sell at -/+ 100% or more of the market value.

When we appraise domains we usually refer to ranges, so we say this domain is a low x,xxx, what does that mean? Some domainers will list that domain at $1000 while other may list it $3000. Because of that I believe that it is very useful to find STR for each price range.

In general their is an optimum price range for any domain (regardless of it is value) where ROI is maximized. If we go far above that range then domain STR will drop significantly, if we go too low then we lose revenue.

Please note that we need also to define STR cutoff limit (under study in my research), which is the limit after which STR drops very low, for example if we take a domain that has a value of $1000-$3000 and increase it's price to $10,000 then we reduce selling chance to near 0.

4. You need testing to find better price

I can't agree any more, and I repeat please do your own test, don't rely on my data or anybody else data.

The important thing from the results that I hope to get from this study, is not static STR values but the relative ones. For example you said you want to do +/- 20% test for domains priced at $2500, I don't know your STR but according to my "draft" formula your STR will:

Increase by 9% at $2000 (ex: for 2% STR => 2*1.09 = 2.18%)
Reduce by 7% at $2500 (ex: for 2% STR => 2*0.93 = 1.86%)

Ofcourse that can be totally wrong, so it is better to test yourself, and I highly encourage you to do that.

5. Each domains portfolio has unique qualities

True. Each domain portfolio is different, but on average most domain portfolios are of similar in quality. Some unique portfolios have different STR profile, but such portfolios are rare. So the idea of this Study is to find General STR formula, not specific formula for a specific portfolio.

In the final study results I can post an Excel template that you can use to derive STR formula for your own portfolio data.

Finally, this survey is to derive STR data only, we are still still far from talking about pricing strategy, we will use this formula to make simulation of different pricing strategies, which should be only educational and informational, you should do your own test, it is your portfolio and you know it better than anyone else.

I appreciate you doing the analysis. And I appreciate your analytical skills.

However, the statistical analysis is way deeper and the conclusions obtained in this way can do more harm than good.

E.g. you adjusted the dofo data to 3% "because that is the number you have in your portfolio". Why not to 1.5% of AbdulBasit's portfolio or to 1% that most domain investors get? Because the output started fitting nicely? But that is not how you do analytics.

Normally you get all input data and then analyze to get the conclusions. You are doing iterations between input and output and adjusting input until it fits the output you expect. That is not how it works though. You could cook up any result you want if that was a normal method.

Again, there is no alternative golden method for this. To measure price elasticity of a product, you need to test SIMULTANEOUSLY sales of the product at different price points. Given that domains, as product, are each unique, you can do some kind of approximation where you say, e.g., all hand reg .coms (in past 1-2 years) from x to y length are Product A, all closeout buys are Product B, all auction wins 12 to 99 dollars are product C, from $100 to $500 are product D etc.

Or use my method which is: take a current portfolio which has thousands of domains priced at $xxxx price point. Randomly distributed into a/b/c and reprice accordingly. Test over the same time period. Consider the results only in the context of the price they had pre-test (as a proxy for quality).

 

[Ostrados]

E.g. you adjusted the dofo data to 3% "because that is the number you have in your portfolio". Why not to 1.5% of AbdulBasit's portfolio or to 1% that most domain investors get? Because the output started fitting nicely? But that is not how you do analytics.

That is why I launched this survey!
I didn't publish the results from my data
And I repeated many times that what I have so far is far from conclusive

Or use my method which is: take a current portfolio which has thousands of domains priced at $xxxx price point. Randomly distributed into a/b/c and reprice accordingly. Test over the same time period. Consider the results only in the context of the price they had pre-test (as a proxy for quality).

Most of us don't have big portfolio to run split tests. I highly encourage you to do a test because you have large enough portfolio, and I will be grateful if you share the test results with us.

 

[Ostrados]

Why not to 1.5% of AbdulBasit's portfolio or to 1% that most domain investors get?

Actually I picked 3% for a totally different Av. Price.

@AbdulBasit.com mentioned that:

* His STR in 2022 was around 1.5%, with average sale price of over $6,000.

* If I apply Av. Price = $6000 in my "draft" formula I get:
STR = 1.460%

Wow I was surprised!
It seems that my "draft" formula is not that bad!

 

[mdrn]

Hi,

I am currently doing a research on optimal domains pricing, the goal of the research is to:

1. Determine sell through rate (STR) for different pricing ranges
2. Derive a formula for the relationship between STR and Domain Price
3. Using formula from #2 we can make simulation for different pricing scenarios to find optimal pricing strategy
4. Ultimately deriving a domain pricing formula (in a separate future study).

This study requires a lot of data, I collected some data using Dofo, Sedo, Namebio.

Here is a sneak peak of what I got so far:
Show attachment 230253Show attachment 230254

However the results I got are inconclusive, and I need more accurate data.

To get more accurate results I am asking Namepros community to help me collect more data, and for that I have created the following Survey:

https://freeonlinesurveys.com/s/OqGsfT2s

* The survey is totally anonymous there is no way to tell who sent it.

* The survey is for all extensions and not specific to .com

* The results of the study will be published at Namepros, I believe the results will be insightful for all of us.

* Contribution will be greatly appreciated especially from big portfolio sellers and from marketplaces that have enough data.

@Sedo @GoDaddy @DAN.COM @LaszloSchenk @GrantP @James Iles @DaaZ @aoxborrow
@AbdulBasit.com
@bmugford
@Recons.Com
@JudgeMind
@xynames
@twiki
@Acroplex
@Bob Hawkes
@Name Trader
@MadAboutDomains
@tonyk2000
@ResoluteDomains
@Leo Angelo
@Nikul Sanghvi
@TERADOMAIN
@Yusupbabay

...and all others please contribute.

Thanks

Thank you very much for doing this. Even though the sample size will be small for the survey, it will be interesting to see the results.

For the graphs you have compiled, how far back do those sales go? Like are those sales from past 5 years or all data available (like all years)?

 

[Recons.Com]

Actually I picked 3% for a totally different Av. Price.

@AbdulBasit.com mentioned that:

* His STR in 2022 was around 1.5%, with average sale price of over $6,000.

* If I apply Av. Price = $6000 in my "draft" formula I get:
STR = 1.460%

Wow I was surprised!
It seems that my "draft" formula is not that bad!

AbdulBasit's numbers are like that because his average domain acquisition cost is mid $xxx for the whole portfolio.

The fact that you believe those numbers somehow validated your output is strange. If he were to lower his price to $2500 for the whole portfolio, e.g., he could get 5%+ STR, while if an average investor would try $6000 average pricing, his STR would collapse to 0.1% or lower.

 

[Recons.Com]

That is why I launched this survey!
I didn't publish the results from my data
And I repeated many times that what I have so far is far from conclusive

Survey is not going to help. In fact, if you get enough response and diligently run numbers for each, you should get Price/STR pictures all of the place depending on the mix each portfolio owner has. And, conversely, similar portfolios priced similarly could get different STR based on a) lander platform selected b) options on the lander selected (bin, bin+makeover, bin+ LTO) c) appetite for taking offers d) communication skills and timely responses.

 

[Recons.Com]

I don't want to go in circles, but if you are serious about the analysis,, I would recommend the books in the attachment. Or there newer version. These are my keepers from my MBA.

I would also recommend taking college or grad level classes on the subject.

 

[Ostrados]

AbdulBasit's numbers are like that because his average domain acquisition cost is mid $xxx for the whole portfolio.

The fact that you believe those numbers somehow validated your output is strange. If he were to lower his price to $2500 for the whole portfolio, e.g., he could get 5%+ STR, while if an average investor would try $6000 average pricing, his STR would collapse to 0.1% or lower.

You got it wrong
The study assumes that you price your domains correctly
Then it finds the STR for each priced domains
Not bulk blind pricing!

For example a domain that is worth $10,000 Then you price it at $10,000 => We want to find the STR for Price = $10,000.

That's it

 

[Ostrados]

I don't want to go in circles, but if you are serious about the analysis,, I would recommend the books in the attachment. Or there newer version. These are my keepers from my MBA.

I would also recommend taking college or grad level classes on the subject.

Show attachment 230299

I have Master's degree in engineering