Akamai Experiencing Growth in Public Sector Services [View article]
I'm one of the founding members of Trefis and I recently caught up with this thread. Let me try to address some of the concerns that have been highlighted by jgonion which encapsulate some of the other comments recently posted.
jgonion:
Your original concern regarding Akamai was that carrying out a certain set of actions led to unexpected results. Typically this is related to one division getting to a point where it would be negative in value and we stop this out (limit value to zero). We do provide a warning on the site related to this. As you continue to make the driver more negative, the division's value has been limited to zero, but the driver changes you're making are have secondary effects on company-wide indirect cash flow items (corporate costs, taxes, capex). This is then lead to unintuitive results such as an increase in the calculated price estimate.
For example, you point out "Drag the "Media Business Gross Margin down to -200% (i.e. The division is losing money). This *doubles* the stock price."
Once you've dragged the gross margin down to the point the divisional value is stopped at zero and you continue to drag down the margin, you're having an impact on company-wide gross margins which are being lowered as a result of the negative gross margin in this division. This is leading to lower corporate costs, capex and taxes (which are calculated off of gross profit or EBIT in the case of taxes). This reduced negative indirect cash flow is flowing back into the other divisions causing part of the unintuitive price increase. Furthermore, the allocation percentage of indirect cash flow items to the other divisions is becoming negative in such cases and we're stopping out the negative allocation. Had we allowed you to create a negative division for Akamai's media business based on such a negative gross margin, the calculated Akamai price estimate would have been about negative $100 per share.
That explains the mechanics of what is happening. Now let's talk about the concepts behind it.
The main conceptual point to your comments seems to be this: "It is incorrect to assign zero-value to a financial loss anyway. A financial loss has negative value. This is a bug."
Firstly, let's distinguish between different types of losses. There are losses you incur in any given year. We show this as negative free cash flow in our models. The sum total of all the discounted free cash flows for a division is the value of that division. What we're assigning a zero value to is the value of the whole division if it would otherwise be negative. We are not zero-ing out the negative free cash flow that may occur in any given year. Your argument is that this is incorrect, a division can contribute negative value.
Although this is true mathematically, in practice it would be pretty silly for someone to continue running a business that is contributing a negative value. In reality, the company will or should just shut that business down. There are two primary reasons why a company may continue to operate a business division that is losing money: (1) it subsidizes or helps the profitability of another (profitable) business division (think mobile carriers giving away mobile phones for cheap) (2) the company believes that eventually the business will turn profitable and the future profits will justify today's losses (net positive present value).
In the case of (1), we believe that such a division should be bundled with the division that it is subsidizing and therefore the combined business should not have a negative net present value. In the case of (2), if you disagree with management and think it will contribute negative value and your view eventually becomes apparent to management then they will presumably shut it down. I don't think it's reasonable to believe someone will run a true loss making business into perpetuity. When you drag the margins to negative for all forecast years, you're calculating a negative terminal value implying that this division will be negative into perpetuity.
If you believe that a public company is going to be in a loss making business forever within no benefit to its other business lines then you may have found a great opportunity in the land of private equity. Raise capital, buy the company, shut down the loss making division and return the company to public markets at a higher valuation.
There is also another problem with allowing you to create negative divisions. Suppose you decide to make all the divisions negative and let's assume the company has zero net cash or debt. Now the stock price is negative. Would you now say that this model is correct or incorrect? It's correct mathematically, right? But does it make sense? Let's say you own the stock, does a negative stock price mean you now owe money? Who do you owe it to?
I think the points I've made above apply to many of your more recent comments. I've also tried to address a few specifically below.
Regarding Your Comment on "Nvidia’s 3 Stock Drivers to Watch" - bit.ly/eIVXV4
Primarily the same issue as above. Dragging Nvidia Pro Graphics Cards Margins as negative as you propose triggers a negative valuation for the division and we stop out the value of that division at zero (this message is provided on the site).
Your point about then positioning the EBITDA margin at around 10% profitability is due to a second order affect. 10% is around the point at which the divisional valuation swings to negative and we stop it out at zero. Then as you drag it down slightly further you are causing second order effects that benefit the other divisions (reduced taxes for example) and this causes the unintuitive increase in the price estimate. This has an effect up to a point after which you're causing so many things to swing negative that we're stopping them out and the price is no longer responding to any further negative changes in the forecast.
Regarding Your Comment on "YouTube Use Gives Google a Boost" - bit.ly/eAX1Va
"If you make the 2011 profit margin -4933% (huge losses) Trefis *increases* the stock price from $632 to $1026."
I get a result close to what you describe by modifying YouTube EBITDA Profit Margin (smoothing on) down to around -5000%. The issue doesn't come up by only modifying the 2011 margin.
Similar to the Akamai point above, there are two things happening here: (1) we're stopping out the valuation of the YouTube division at zero before it crosses into negative land (2) the extremely high negative profits of YouTube are having secondary effects on the other divisions by flowing through indirect costs.
Let's say we had allowed the YouTube division to be negative based on a -5000% margin. You would be forecasting that YouTube loses $50 billion for every $1 billion in revenues which is absurd. Furthermore, the divisional valuation would have been negative $4.2 trillion and the Google stock price would have been negative $12K.
By stopping out to the negative valuation of the YouTube division, we've prevented the extreme downside result. However, we have not prevented you from creating secondary effects. YouTube is so negative in this case that Google is a negative EBITDA company in your example. Since company-wide indirect cash flow items like corporate costs, taxes and capex are calculated off of EBITDA (EBIT in case of taxes), you've positive cash flow items from these that bump up the price.
Regarding Your Comment on "Verizon FiOS Penetration Rising" - bit.ly/fd3HQo
Comment 1:
"In the case of this FiOS article: Try going to the Trefis website for VZ, and dragging FiOS profit margins down to -400% (i.e. VZ losing money on FiOS).
The Trefis website calculates the stock price will *increase 13-fold* from 31.69 to 424.67 due to FiOS losing money."
Same issue as YouTube and Akamai above
Comment 2:
"As you can see from your own model, you project that MSFT stock price will be higher at 4% profit margin than if they have a 27% profit margin. These are both realistic numbers. It would be advisable to fix the bugs in your model rather than make excuses for them." (This is related to another commment posted on a recent Microsoft article here - bit.ly/fzhpSL)
You're referring to modifying Windows OS Operating Margin within the Microsoft analysis. By dragging down margins to 27% in 2011 (with smoothing on), you're coming up with an overall Windows OS Operating Margin of around 15-20% over the forecast period as shown on the chart. Again, this tips the divisional free cash flows across most of the outer years to negative and calculates a negative terminal value implying that the overall divisional value should be negative. The warning pop-up appears on the site telling you that the division profits have swung negative and that the division has no value now. The Trefis price calculated now is about $21.35 as you point out.
As you drag down the margins even lower, you get a similar behavior as described above for Google where secondary effects related to company-wide indirect cash flow items (corporate costs, taxes, capex) are triggering a slight increase in the valuation to about $23.
Of the points you've highlighted, this seems the most reasonable given the range of numbers involved. As pointed out above, I don't think it makes sense to allow a negative division valuation. In this example, you would have had a valuation for Windows of about negative $13 billion.
Once we've stopped out the division value at $0 and you continue to drive down margins, the price is going up because you're continuing to drive negative EBIT for that division resulting in lower company-wide EBIT off which a lower tax bill is calculated. This lower tax bill is being allocated back to the remaining non-negative business lines leading to a rise in their valuation (this is how companies share the benefits of interest tax shields related to negative profits from some divisions).
Now, assuming for a moment that you accept the argument for stopping out a negative divisional value at zero, this is mathematically a correct outcome, but I understand how it can seem illogical without understanding the mechanics of the model. Providing better messaging on site in such a case is something we can incorporate to provide greater transparency in such cases.
"Another bug is that a division that loses money accounts for 0% of the stock price. You can't seriously believe this. It is well-known that a division that loses money weights negatively on the stock price."
Responded to above. Keep in mind the distinction between a division losing money for a finite number of years vs. into perpetuity.
jgonion, I've also privately messaged you with my office phone number. If you have any doubts about this or anything in the future, you can call me directly. I haven't had a chance to go through your entire history of predominantly Trefis related comments so if you'd like me to look at other specific issues, please call or message me.
Overall, we do want to improve by fixing things that are truly bugs and I have noted down a few things we can do to improve the user experience including better warning messaging on our site and within our chart widgets. Having said that, let me also say that I don't really consider a result achieved by dragging margins down to -200% a bug. This is a bit like driving your car off a cliff and complaining about the in flight aerodynamics. Let's all understand that the base Trefis model is a tool and like all tools can be used (abused) in unintended ways. Use your judgment. The fact that outlier cases can "break" the Trefis model isn't "proof" that the base Trefis analysis is bogus. I would appreciate it if comments expressing concern can focus on bugs related to reasonable estimates.
Thanks for your comments Richard. They are very insightful. While our intent with the article was to talk about the U.S. marketplace, we agree with you that Latin American market is becoming increasingly important for DirecTV. Although capital expenditures remain high for Latin American business since this is the initial expansion phase (thus cash flows are low currently). We intend to revise our forecasts during next earnings.
In the event of extreme scenarios where the division is never intended to be profitable (negative margins, negative revenues, etc.), a warning appears to say that an assumption is causing negative values for a division.
Akamai Experiencing Growth in Public Sector Services [View article]
Dear jgonion and Kanman,
We value constructive feedback as it helps us improve our platform.
Many of your comments involve dragging margins or revenues to zero or extreme negative values where the value of a division is negative.
Like most models, ours are designed to operate within reasonable limits for a business as our aim is to provide tools to gauge the impact of drivers on a company's stock price.
This is not a perfect science. As you have highlighted in other posts, evaluating stocks is a complex process. For this reason, we provide our rationale on how we derive our estimate and then allow you to make the adjustments as you see fit.
Unfortunately our models are not designed to handle the extreme values you have suggested. In such scenarios, a yellow box should pop up to say that you have made adjustments resulting in negative values for a division. In this scenario, you are right, you should not trust the output.
We are working on incorporating these warnings in widgets or tools that you might see in articles or in other locations (not on our site). Additionally we hope to add limits so that such extreme assumptions do not impact our forecast in the ways you pointed out.
If you have further feedback or would like to discuss specific scenarios, we would be happy to help in any way we can.
iPad Distribution Expands, Impact Small for Apple’s Stock [View article]
Thanks for your comments.
We updated our estimates a few weeks back to around 12.5 million iPads for 2010 and 20 million for 2011. This is our analyst's forecast that will get updated as more data become available.
GSlusher, You're probably right that these are conservative estimates. Feel free to modify the charts above with your own forecast to gauge the price sensitivity. Thanks again.
Netflix Faces Growing Competition From ISPs [View article]
Hi Dan,
We never intended to say that ISPs have a better content compared to Netflix when it comes to their services like Xfinity. Our statement, "This directly competes with Netflix’s growing online streaming offering and with better content, cable operators can significantly weaken Netflix’s value proposition.", is merely built upon the 1st argument that ISPs can potentially improve their content. When we say "with a better content", it reflects potential and not the current state.
We had requested the correction from Seeking Alpha on August 25th. We have limited control over Seeking Alpha's editorial policy and have requested that the change be further clarified.
Furthermore, the change reflects an editorial mistake at the time and does not represent a change in the Trefis forecast since publication of that article.
You can verify this by comparing to the iPad forecast shown (10 million in 2010) in an article we published earlier (August 6, 2010) on our own site: "Modest Upside to Apple from Higher iPad Sales" (link below)
We do apologize for the editorial mistake in the original August 25th article. It was not our intent to mislead.
In comparison, in February 2010, we did have expectations for 4 million iPad unit sales in 2010. This forecast was made in advance of the iPad release (April 2010) and was later revised based on indications of strong iPad sales. You can verify this also based on the link below.
By clicking through the different eBay divisions shown in our analysis, you can see our forecasts for each of the relevant drivers of eBay. Using these drivers, we're able to forecast divisional free cash flows which drive the DCF-based valuation.
You can access additional detail within our analysis (margins, discount rate) using the "All Forecasts" button on the right-hand side of page after selecting an eBay division. By toggling to the "All" tab within the "All Forecasts" window that appears, you'll have access to all of the forecasts that factor into our analysis.
You can also find more information on our methodology and why we do DCF-based analysis rather than multiples analysis within our FAQ: www.trefis.com/faq#25
Verizon FiOS Penetration Rising [View article]
bit.ly/fIKVL0
Akamai Experiencing Growth in Public Sector Services [View article]
jgonion:
Your original concern regarding Akamai was that carrying out a certain set of actions led to unexpected results. Typically this is related to one division getting to a point where it would be negative in value and we stop this out (limit value to zero). We do provide a warning on the site related to this. As you continue to make the driver more negative, the division's value has been limited to zero, but the driver changes you're making are have secondary effects on company-wide indirect cash flow items (corporate costs, taxes, capex). This is then lead to unintuitive results such as an increase in the calculated price estimate.
For example, you point out "Drag the "Media Business Gross Margin down to -200% (i.e. The division is losing money). This *doubles* the stock price."
Once you've dragged the gross margin down to the point the divisional value is stopped at zero and you continue to drag down the margin, you're having an impact on company-wide gross margins which are being lowered as a result of the negative gross margin in this division. This is leading to lower corporate costs, capex and taxes (which are calculated off of gross profit or EBIT in the case of taxes). This reduced negative indirect cash flow is flowing back into the other divisions causing part of the unintuitive price increase. Furthermore, the allocation percentage of indirect cash flow items to the other divisions is becoming negative in such cases and we're stopping out the negative allocation. Had we allowed you to create a negative division for Akamai's media business based on such a negative gross margin, the calculated Akamai price estimate would have been about negative $100 per share.
That explains the mechanics of what is happening. Now let's talk about the concepts behind it.
The main conceptual point to your comments seems to be this: "It is incorrect to assign zero-value to a financial loss anyway. A financial loss has negative value. This is a bug."
Firstly, let's distinguish between different types of losses. There are losses you incur in any given year. We show this as negative free cash flow in our models. The sum total of all the discounted free cash flows for a division is the value of that division. What we're assigning a zero value to is the value of the whole division if it would otherwise be negative. We are not zero-ing out the negative free cash flow that may occur in any given year. Your argument is that this is incorrect, a division can contribute negative value.
Although this is true mathematically, in practice it would be pretty silly for someone to continue running a business that is contributing a negative value. In reality, the company will or should just shut that business down. There are two primary reasons why a company may continue to operate a business division that is losing money: (1) it subsidizes or helps the profitability of another (profitable) business division (think mobile carriers giving away mobile phones for cheap) (2) the company believes that eventually the business will turn profitable and the future profits will justify today's losses (net positive present value).
In the case of (1), we believe that such a division should be bundled with the division that it is subsidizing and therefore the combined business should not have a negative net present value. In the case of (2), if you disagree with management and think it will contribute negative value and your view eventually becomes apparent to management then they will presumably shut it down. I don't think it's reasonable to believe someone will run a true loss making business into perpetuity. When you drag the margins to negative for all forecast years, you're calculating a negative terminal value implying that this division will be negative into perpetuity.
If you believe that a public company is going to be in a loss making business forever within no benefit to its other business lines then you may have found a great opportunity in the land of private equity. Raise capital, buy the company, shut down the loss making division and return the company to public markets at a higher valuation.
There is also another problem with allowing you to create negative divisions. Suppose you decide to make all the divisions negative and let's assume the company has zero net cash or debt. Now the stock price is negative. Would you now say that this model is correct or incorrect? It's correct mathematically, right? But does it make sense? Let's say you own the stock, does a negative stock price mean you now owe money? Who do you owe it to?
I think the points I've made above apply to many of your more recent comments. I've also tried to address a few specifically below.
Regarding Your Comment on "Nvidia’s 3 Stock Drivers to Watch" - bit.ly/eIVXV4
Primarily the same issue as above. Dragging Nvidia Pro Graphics Cards Margins as negative as you propose triggers a negative valuation for the division and we stop out the value of that division at zero (this message is provided on the site).
Your point about then positioning the EBITDA margin at around 10% profitability is due to a second order affect. 10% is around the point at which the divisional valuation swings to negative and we stop it out at zero. Then as you drag it down slightly further you are causing second order effects that benefit the other divisions (reduced taxes for example) and this causes the unintuitive increase in the price estimate. This has an effect up to a point after which you're causing so many things to swing negative that we're stopping them out and the price is no longer responding to any further negative changes in the forecast.
Regarding Your Comment on "YouTube Use Gives Google a Boost" - bit.ly/eAX1Va
"If you make the 2011 profit margin -4933% (huge losses) Trefis *increases* the stock price from $632 to $1026."
I get a result close to what you describe by modifying YouTube EBITDA Profit Margin (smoothing on) down to around -5000%. The issue doesn't come up by only modifying the 2011 margin.
Similar to the Akamai point above, there are two things happening here: (1) we're stopping out the valuation of the YouTube division at zero before it crosses into negative land (2) the extremely high negative profits of YouTube are having secondary effects on the other divisions by flowing through indirect costs.
Let's say we had allowed the YouTube division to be negative based on a -5000% margin. You would be forecasting that YouTube loses $50 billion for every $1 billion in revenues which is absurd. Furthermore, the divisional valuation would have been negative $4.2 trillion and the Google stock price would have been negative $12K.
By stopping out to the negative valuation of the YouTube division, we've prevented the extreme downside result. However, we have not prevented you from creating secondary effects. YouTube is so negative in this case that Google is a negative EBITDA company in your example. Since company-wide indirect cash flow items like corporate costs, taxes and capex are calculated off of EBITDA (EBIT in case of taxes), you've positive cash flow items from these that bump up the price.
Regarding Your Comment on "Verizon FiOS Penetration Rising" - bit.ly/fd3HQo
Comment 1:
"In the case of this FiOS article: Try going to the Trefis website for VZ, and dragging FiOS profit margins down to -400% (i.e. VZ losing money on FiOS).
The Trefis website calculates the stock price will *increase 13-fold* from 31.69 to 424.67 due to FiOS losing money."
Same issue as YouTube and Akamai above
Comment 2:
"As you can see from your own model, you project that MSFT stock price will be higher at 4% profit margin than if they have a 27% profit margin. These are both realistic numbers. It would be advisable to fix the bugs in your model rather than make excuses for them." (This is related to another commment posted on a recent Microsoft article here - bit.ly/fzhpSL)
You're referring to modifying Windows OS Operating Margin within the Microsoft analysis. By dragging down margins to 27% in 2011 (with smoothing on), you're coming up with an overall Windows OS Operating Margin of around 15-20% over the forecast period as shown on the chart. Again, this tips the divisional free cash flows across most of the outer years to negative and calculates a negative terminal value implying that the overall divisional value should be negative. The warning pop-up appears on the site telling you that the division profits have swung negative and that the division has no value now. The Trefis price calculated now is about $21.35 as you point out.
As you drag down the margins even lower, you get a similar behavior as described above for Google where secondary effects related to company-wide indirect cash flow items (corporate costs, taxes, capex) are triggering a slight increase in the valuation to about $23.
Of the points you've highlighted, this seems the most reasonable given the range of numbers involved. As pointed out above, I don't think it makes sense to allow a negative division valuation. In this example, you would have had a valuation for Windows of about negative $13 billion.
Once we've stopped out the division value at $0 and you continue to drive down margins, the price is going up because you're continuing to drive negative EBIT for that division resulting in lower company-wide EBIT off which a lower tax bill is calculated. This lower tax bill is being allocated back to the remaining non-negative business lines leading to a rise in their valuation (this is how companies share the benefits of interest tax shields related to negative profits from some divisions).
Now, assuming for a moment that you accept the argument for stopping out a negative divisional value at zero, this is mathematically a correct outcome, but I understand how it can seem illogical without understanding the mechanics of the model. Providing better messaging on site in such a case is something we can incorporate to provide greater transparency in such cases.
"Another bug is that a division that loses money accounts for 0% of the stock price. You can't seriously believe this. It is well-known that a division that loses money weights negatively on the stock price."
Responded to above. Keep in mind the distinction between a division losing money for a finite number of years vs. into perpetuity.
jgonion, I've also privately messaged you with my office phone number. If you have any doubts about this or anything in the future, you can call me directly. I haven't had a chance to go through your entire history of predominantly Trefis related comments so if you'd like me to look at other specific issues, please call or message me.
Overall, we do want to improve by fixing things that are truly bugs and I have noted down a few things we can do to improve the user experience including better warning messaging on our site and within our chart widgets. Having said that, let me also say that I don't really consider a result achieved by dragging margins down to -200% a bug. This is a bit like driving your car off a cliff and complaining about the in flight aerodynamics. Let's all understand that the base Trefis model is a tool and like all tools can be used (abused) in unintended ways. Use your judgment. The fact that outlier cases can "break" the Trefis model isn't "proof" that the base Trefis analysis is bogus. I would appreciate it if comments expressing concern can focus on bugs related to reasonable estimates.
Best,
Cem
Akamai Experiencing Growth in Public Sector Services [View article]
Your feedback is appreciated and we are working on incorporating this in some way to avoid confusion going forward.
DirecTV’s Subscriber Gains Continue [View article]
Akamai Experiencing Growth in Public Sector Services [View article]
Verizon FiOS Penetration Rising [View article]
Verizon FiOS Penetration Rising [View article]
Akamai Experiencing Growth in Public Sector Services [View article]
We value constructive feedback as it helps us improve our platform.
Many of your comments involve dragging margins or revenues to zero or extreme negative values where the value of a division is negative.
Like most models, ours are designed to operate within reasonable limits for a business as our aim is to provide tools to gauge the impact of drivers on a company's stock price.
This is not a perfect science. As you have highlighted in other posts, evaluating stocks is a complex process. For this reason, we provide our rationale on how we derive our estimate and then allow you to make the adjustments as you see fit.
Unfortunately our models are not designed to handle the extreme values you have suggested. In such scenarios, a yellow box should pop up to say that you have made adjustments resulting in negative values for a division. In this scenario, you are right, you should not trust the output.
We are working on incorporating these warnings in widgets or tools that you might see in articles or in other locations (not on our site). Additionally we hope to add limits so that such extreme assumptions do not impact our forecast in the ways you pointed out.
If you have further feedback or would like to discuss specific scenarios, we would be happy to help in any way we can.
feedback@trefis.com
Thanks, Trefis Team
iPad Distribution Expands, Impact Small for Apple’s Stock [View article]
We updated our estimates a few weeks back to around 12.5 million iPads for 2010 and 20 million for 2011. This is our analyst's forecast that will get updated as more data become available.
GSlusher,
You're probably right that these are conservative estimates. Feel free to modify the charts above with your own forecast to gauge the price sensitivity. Thanks again.
Netflix Faces Growing Competition From ISPs [View article]
We never intended to say that ISPs have a better content compared to Netflix when it comes to their services like Xfinity. Our statement, "This directly competes with Netflix’s growing online streaming offering and with better content, cable operators can significantly weaken Netflix’s value proposition.", is merely built upon the 1st argument that ISPs can potentially improve their content. When we say "with a better content", it reflects potential and not the current state.
Thanks
Apple iPad Not Hurting Mac Sales [View article]
Furthermore, the change reflects an editorial mistake at the time and does not represent a change in the Trefis forecast since publication of that article.
You can verify this by comparing to the iPad forecast shown (10 million in 2010) in an article we published earlier (August 6, 2010) on our own site: "Modest Upside to Apple from Higher iPad Sales" (link below)
www.trefis.com/company...#
We do apologize for the editorial mistake in the original August 25th article. It was not our intent to mislead.
In comparison, in February 2010, we did have expectations for 4 million iPad unit sales in 2010. This forecast was made in advance of the iPad release (April 2010) and was later revised based on indications of strong iPad sales. You can verify this also based on the link below.
www.trefis.com/article...
Still Large Upside on eBay [View article]
We're not valuing eBay off of multiples but rather doing a DCF-based analysis. You can access our full analysis using the link below.
bit.ly/b6mrJl
By clicking through the different eBay divisions shown in our analysis, you can see our forecasts for each of the relevant drivers of eBay. Using these drivers, we're able to forecast divisional free cash flows which drive the DCF-based valuation.
You can access additional detail within our analysis (margins, discount rate) using the "All Forecasts" button on the right-hand side of page after selecting an eBay division. By toggling to the "All" tab within the "All Forecasts" window that appears, you'll have access to all of the forecasts that factor into our analysis.
You can also find more information on our methodology and why we do DCF-based analysis rather than multiples analysis within our FAQ: www.trefis.com/faq#25
Hope that helps.