discuss Artificial Intelligence and how it applies to outbound sales

[ThatNameGuy]

No doubt domainers way smarter than me are working toward a better way to market/sell domain names. While artificial intelligence is being used by retailers like Amazon to target prospects, realtors are also using AI to target home buyers.

I asked the question, is SEO artificial intelligence? No, but if you google "Is SEO artificial intelligence", you'll learn how AI empowers SEO.

Our brains are actually computers that if coordinated could develop one hell of a program that would target domain prospects rather than wait for these prospects to find us.

I believe I know enough about sales and marketing in general to provide credible input to make more domain sales to the real estate industry, the healthcare industry, and the food/restaurant industries.

Final questions...what, if anything do you know about AI and how it relates to selling domains? Do you know any one individual or company that's actively pursuing AI solutions for outbound domain sales? Inquiring minds would luv to know. Thanks

 

[Ategy]

Domainers who want to outbound already have access to one of the strongest AIs in history .. it's called Google!


That being said .. for domainer customised solutions, AI needs massive amounts of data before spitting out results with any kind of accuracy. Amounts of data that domainers (and even most registrars/marketplaces) simply don't have anything nearly close to enough of to get any meaningful results from even if they did have access to the computing/processing power.

Not only that .. but it's total overkill when there are plenty of lead-generation tools already available at your fingertips for anyone who wants to do outbound on their domain. The most obvious being the afore mentioned Google .. LinkedIn, business directories, or any large user-based website with a searchbox.


The bigger problem with AI is that when it comes to brandables, it wouldn't be very effective as every brandable is at least somewhat unique .. as are most end-user buyers .. not only that, but many also have artistic components that would be almost impossible to program into an algorithm with any semblance of accuracy (it's why automated appraisals are just about useless when it comes to most brandables) ... so while an AI solution most certainly could likely find some patterns to generate some leads, it's pretty overkill and in the end probably wouldn't be much better than a simple google/LinkedIn search.

 

[Bob Hawkes]

Does someone know if the big operations, like Huge Domains, use AI in their acquisition, pricing and renewal decisions. I presume they use some kind of algorithm, probably overseen by human intervention, but do they actually use true AI as in software that gradually teaches itself to be better. They clearly have a big enough operation that it would seem they could do it if they decided to.

Also interested to what degree that GoDaddy implement true AI. They have access to a huge resource to learn from.

I have always wondered what Google's long term plan in the domain space is. Clearly they have an unfair advantage through dominance in search. Now they have become a top tier registrar - e.g. in top 10 in world for .com registrations. They have delegated some new gTLDs and have a bunch more sitting there. As @Ategy notes they are the masters of AI. So they could, if desired, leapfrog to a really dominant tool for valuation, information sources, etc. for any domain. Those who have invested a lot in developing domain intelligence services must be constantly looking over their shoulder at Google. Would be interesting to know their aspirations with respect to domain names.

Thanks for thread. Interesting topic. Probably far more questions than answers.

Bob

 

[Bob Hawkes]

The bigger problem with AI is that when it comes to brandables, it wouldn't be very effective as every brandable is at least somewhat unique .. as are most end-user buyers

Does anyone know if the brandable marketplaces use any kind of algorithm, whether AI or not, as a first pass or second opinion for their acceptance decisions? It must be a rather huge job, and while people can undoubtedly do it much better currently, I am not sure that will always be the case. AI already is successful at tasks that in my mind are more complex than a good brandable, or at least arguably as complex.

Bob

 

[Ategy]

The biggest problem with discussions on AI is that definitions of what AI is ranges so wide from person to person that in the end many people aren't even talking about the same thing.

To some AI means sentience .. while at the other end of the spectrum my own personal domaining process could be considered AI because I occasionally add data and tweak my excel equations.

When it comes to automated valuations, the reason they all fail is because it would be excessively complex to program algorithms to score things like phonetics and word flow. Certainly not impossible, but there are so many sub-conscious "artistic" components to what makes up brandables that it would truly be a monstrous task to set up the groundwork and foundation for an AI system to start scoring. Not only that, but I'm not even sure all the data that would be needed even exists as a large component of the data needed to grow such an AI would be "human response/reaction" to those specific attributes. (Remember that the main raison-d'etre of a domain brand is for your website's IP address to be memorable to a fellow HUMAN)

It would certainly make for a very interesting project .. but extremely complex and expensive .. and in the end there would still be a huge margin of error because the actual value we want from an appraisal is what another HUMAN would pay for the domain .. and in the end much of the value a buyer attributes to a potential domain they want to purchase is based on the business concept they have planned for the domain. That's something an AI could sometimes figure out to a small degree .. but most certainly not always .. particularly with brandables.

 

[Bob Hawkes]

My feeling on this is that while AI is far from domain perfect, it can already probably do much better than over reliance on a simple set of rules such as:

1. only .com
2. no hyphens
3. only buy in expired auctions
4. never mix letters and numbers
5. etc.

Just those first two, for example, would rule out all of top four on today's 10 highest value sales on NameBio, and 5 of the top 6. And while that many ruled out by just two simple widely adopted rules is not common, a number ruled out by just the most popular few rules is not at all unusual. For example the preceding day the top two sales would be ruled out by using only the first two rules again. Most days some of top 10 sales would be ruled out with only the most repeated common rules, and now and then a majority of top 10 eliminated by just a very few simple rules I see repeated on Request thread all off the time.

What I am stressing is that AI could be far from perfect to be better than following simplistic rules.

Bob

EDIT Just to not be misunderstood, I agree that the simple rules work a fair amount of the time. Not saying they are bad advice. But depending on them as absolutes, is reliance on actually ultra-simplistic "AI" and is ruling out many opportunities. I think something considerably better could be implemented. One advantage of AI is it is more immune from human biases and can better follow new trends.

 

[Ategy]

Does anyone know if the brandable marketplaces use any kind of algorithm, whether AI or not, as a first pass or second opinion for their acceptance decisions? It must be a rather huge job, and while people can undoubtedly do it much better currently, I am not sure that will always be the case. AI already is successful at tasks that in my mind are more complex than a good brandable, or at least arguably as complex.

Don't take my comments above as me saying algorithms and automation is bad. I think in the end a smart domainer will use automation when appropriate, but also know when a human eye is better.

Every day there are about 50,000 expired domains at auction at GoDaddy. I obviously don't go through all of them .. it would be stupid and a waste of time for me to go through 23 character .info's .. about 80% of all expiring domains are 99.9+% garbage .. but on the other side, if I filter out too many via automation, that's when I start to miss some good ones (particularly the ones that other people miss that I can get for cheap).


But all this is acquisitions .. when it comes to outbound, getting results noticeably better than google is going to be a huge problem in some settings because the data for many (particularly brandable) domains simply does not exist. By definition itself founders looking to brand their start-ups don't have much if any data about their start-ups available to be found by AI systems. So while a perfect buyer might exist, the information/data needed by an AI system to find that buyer simply isn't there.

Most definitely when it comes to more straight forward domains like real estate geo domains, there is plenty of information and data .. but there really is no need to come up with AI for those as the leads are already at your fingertips with a couple of keystrokes into Google search.

 

[Bob Hawkes]

Do you know any one individual or company that's actively pursuing AI solutions for outbound domain sales?

I wish I knew the answer to this. I don't. It would seem surprising if none were.
Bob

PS I think this is one place where the relative absence of a scholarly side of domain analysis (I know there is a tiny bit of activity) is unfortunate. If scholarly papers were being written on topics like how to use AI to predict business owner choices for brandable domain names not only would we know what is being done, but we would benefit from what they found.

 

[Joe N]

I wish I knew the answer to this. I don't. It would seem surprising if none were.

I'm not sure what something like that would look like. @Ategy nailed it above when he said the necessary data to target people who would like your name simply doesn't exist (beyond the tools he already mentioned: Google, LinkedIn, etc).

We already have the ability to easily identify potential buyers in targeted industries. The only way AI could improve upon that would be to crawl into people's heads and determine how likely they would be to favour certain names.

I suppose there's always the possibility of using social media and online behaviour data to better target potential buyers (much the way political parties have been targeting undecided voters for years), but I'm not sure the techniques can be ported to selling domain names. In elections, people only have a few choices. When buying domain names, the choices are endless.

 

[Ategy]

One advantage of AI is it is more immune from human biases and can better follow new trends.

That's kinda the problem. With the exception of one-word/term, 3L, numerics and exact match domains; "Brandables" are specifically all about human bias based interpretation.

Look at what we discussed here yesterday. The level of complexity and sophistication an AI would need to be to recognise the value in a brand like "AlphaMail" is extremely difficult. Same holds true to distiguishing the 0.1% of good 5L brands from the 99.9% that are worthless garbage? There definitely are metrics like phonetics that you can get with AI .. but you aren't going to get results much better than a human. In fact .. in the end you'll end up facing the choice of speed (Algorithms or AI) vs accuracy (Human) .. and the best result will be a hybrid solution.

 

[twiki]

I am pretty sure big companies like Huge Domains use AI. It would be stupid not to. And they have the ability and funds to do it.

AI and automation and machine learning aren't the same.

Using AI for domain filtering, even for brandables is definitely doable. It won't be perfect though, so there will be constant work to refine the results by feeding new training to the AI based on human verification of the output. There are even startups that claim using AI to generate brandables (kw gen). (Edit: seen their output and I have my doubts on that)

It is to be said though that you pretty much need an entire team to do it, and especially a very good linguist and AI coders. Otherwise it would be hard to impossible to tame the difficulties of language and to reach the proper understanding of the domain/contained words etc.

Otherwise, I think AI-based automation is overkill for regular domainers. Good standard filtering based on negative words, + rules based on consonants, vowels, number of words, etc. is enough. This is still coding/some form of artificial intelligence that does classification and filtering, but it is not machine learning, what people commonly refer to as AI. It uses fixed algorithms instead of neural networks. (Edit: there's also deep learning but it's not the case to go in too deep here)

But regarding outbound, it's hard to say. Until the bots will be able to call people directly, and taken seriously and being negotiate with, the feedback loop is missing, therefore no machine learning can be used since the chain is definitely broken.

In the end, in domaining, it's always a mix of AI and human. An eye needs to be kept continuously on the AI, retraining is constant work, otherwise it will start going in the wrong direction. Think trends, think social changes, think language evolution, think business decisions, etc etc - a human hand needs to be present there. But computers can do the grunt work for you.

 

[Joe N]

There definitely are metrics like phonetics that you can get with AI .. but you aren't going to get results much better than a human. In fact .. in the end you'll end up facing the choice of speed (Algorithms or AI) vs accuracy (Human) .. and the best result will be a hybrid solution.

This is it exactly. I'm convinced that most of the decent domains I find are due to my willingness to put in the time evaluating a filtered list of names. Domainers love automated tools, and as a result I think many of us filter too much, or rely on other investors bidding on names to find the good ones.

The more that domainers rely exclusively on automated tools, the more good names will be available for those of us using our brains.

 

[Bob Hawkes]

Thank you for your comments, @Joe Nichols - common sense clearly expressed!

I guess I would see one component as exactly this....

and determine how likely they would be to favour certain names.

Let's imagine that Google does it but in concert with a large source of updated information on the most recent domain names sold, including how and if they are used. Also they have their coverage of content on social media, their access to the TM and copyright databases, access to the various registered company name registrations, Google trends and search data of course, etc.

Let's assume that they partner with the major brandable marketplaces to provide access to the AI in return for brandable sales data. It seems to me that it would not be a huge step to actually get a product that was better than humans in saying this domain is more likely to be selected than this one. Or this one is likely to be accepted for a pharmaceutical product and they typically would pay xxxx for domain names, so help with valuations could come as well.

Let's say if one brandable marketplace could suddenly make better choices, so their sell through rate goes from 5% to 20% (just using made-up numbers here). Suddenly domainers would say 'hey, that is the place to list'. Then they get more quality domains, and their sell-through rate and prices and inventory quality further go up, and then the AI gets smarter with this new volume and data, etc.

Or, what about advertising. When did you last see an online advertisement for a specific domain name? I mean we see registry and registrar adds, but I mean for a specific name. Smart enough AI might be able to target a specific name that is valued enough so it gets seen in the social media feeds of exactly the kind of person who might buy it.

Just early morning wondering....

Bob

Edit addition: I think the other way AI could enable the domain industry would be tools to allow a potential client to compare the benefit of a domain name. That might work against other interests of Google, so let's imagine this is developed by someone not in the advertising provision services. If an automated report could compare for a given company, based on some questions about existing company, location, niche etc., see how two different domain choices would impact future in terms of leads, exposure, prestige, etc. Not easy to quantify but remember we are not talking yesterday's AI but tomorrows. If people came to trust such a product, people would pay for the reports I feel sure.

 

[twiki]

This is it exactly. I'm convinced that most of the decent domains I find are due to my willingness to put in the time evaluating a filtered list of names. Domainers love automated tools, and as a result I think many of us filter too much, or rely on other investors bidding on names to find the good ones.

The more that domainers rely exclusively on automated tools, the more good names will be available for those of us using our brains.

Don't worry - there will always be good names out there for human-based selection.

The reason is simple. Each AI implementation is unique; and the uniqueness comes not necessarily from the software (can be the same) but how it is trained and used. Each business will configure and train it based on their particular needs and model; data being fed and looped back will be different; as a result each implementation is like a different person, and 2 different people don't necessarily go for the same things.

 

[twiki]

Thank you for your comments, @Joe Nichols - common sense clearly expressed!

I guess I would see one component as exactly this....

Let's imagine that Google does it but in concert with a large source of updated information on the most recent domain names sold, including how and if they are used. Also they have their coverage of content on social media, their access to the TM and copyright databases, access to the various registered company name registrations, Google trends and search data of course, etc.

Let's assume that they partner with the major brandable marketplaces to provide access to the AI in return for brandable sales data. It seems to me that it would not be a huge step to actually get a product that was better than humans in saying this domain is more likely to be selected than this one. Or this one is likely to be accepted for a pharmaceutical product and they typically would pay xxxx for domain names, so help with valuations could come as well.

Let's say if one brandable marketplace could suddenly make better choices, so their sell through rate goes from 5% to 20% (just using made up numbers here). Suddenly domainers would way hey, that is the place to list. Then they get more quality domains, and their sell-through rate and prices and inventory quality further go up, and then the AI gets smarter with this new volume and data, etc.

Or, what about advertising. When did you last see an online advertisement for a specific domain name? I mean we see registry and registrar adds, but I mean for a specific name. Smart enough AI might be able to target a specific name that is valued enough so it gets seen in the social media feeds of exactly the kind of person who might buy it.

Just early morning wondering....

Bob

Definitely possible, even probable, but we aren't there yet.

 

[Joe N]

Thank you for your comments, @Joe Nichols - common sense clearly expressed!

I guess I would see one component as exactly this....

Let's imagine that Google does it but in concert with a large source of updated information on the most recent domain names sold, including how and if they are used. Also they have their coverage of content on social media, their access to the TM and copyright databases, access to the various registered company name registrations, Google trends and search data of course, etc.

Let's assume that they partner with the major brandable marketplaces to provide access to the AI in return for brandable sales data. It seems to me that it would not be a huge step to actually get a product that was better than humans in saying this domain is more likely to be selected than this one. Or this one is likely to be accepted for a pharmaceutical product and they typically would pay xxxx for domain names, so help with valuations could come as well.

Let's say if one brandable marketplace could suddenly make better choices, so their sell through rate goes from 5% to 20% (just using made up numbers here). Suddenly domainers would way hey, that is the place to list. Then they get more quality domains, and their sell-through rate and prices and inventory quality further go up, and then the AI gets smarter with this new volume and data, etc.

Or, what about advertising. When did you last see an online advertisement for a specific domain name? I mean we see registry and registrar adds, but I mean for a specific name. Smart enough AI might be able to target a specific name that is valued enough so it gets seen in the social media feeds of exactly the kind of person who might buy it.

Just early morning wondering....

Bob

Yeah, I think the methods you're proposing are likely already in use by the bigger players in the industry (at least to some degree). And using these tactics would certainly improve your sell-through rate and would inform better buying decisions.

But how would these methods help you better target outbound leads? Even amassing vast amounts of personal data would have a limited benefit, since there are often multiple people involved in high priced domain buying decisions.

And then of course there's the cost/benefit issue...

 

[Internet.Domains]

The biggest problem with discussions on AI is that definitions of what AI is ranges so wide from person to person that in the end many people aren't even talking about the same thing.

In recent times, AI has been softened to mean algorithms completing tasks. AI is now very simple. Further down the spectrum is Deep Learning which is more complex. We are in times when these terms are absolutely misconstrued due to definition tampering.

 

[Bob Hawkes]

But how would these methods help you better target outbound leads? Even amassing vast amounts of personal data would have a limited benefit, since there are often multiple people involved in high priced domain buying decisions.

I don't know. But I think if someone could build a much better place to sell domain names, we would benefit. At least our good names would benefit.

I think there is a place for some innovative company with AI skills to build some sort of scaleable solution like the world's best place to find domain names. As I argued in another thread, I think the search in all of the big general purpose marketplaces is far from optimum. They present big lists of results most of which the client would never consider so they lose trust in the marketplace. Imagine an AI-based system, that learned from each sale how to get better, bot that worked something like this.....

"Hi, I am Denise and want to help you find just the right name today. Can you first just in a few words tell me what you are looking for?"​

[response]

"So just to be clear, you are mainly looking for a name for your startup company that is going to use AI to automate help to legal clients for routine issues. Do I understand that right?"​

[response]

"Just to get a feel for the sort of name you might prefer, I am going to show you three possible names. Even if you don't like any of them, can you please order them in preference?​

[would be shown a made up brandable type name, and two word name, and an unusual single word name or a name from another language, etc.]

"Thanks. That helps me search more effectively for you....​

and so on ultimately leading to a curated list of say 15 attractively presented names, and then from that the person chooses 5 to better define what they are looking for and so on.

If someone can make the best way to sell domain names, they will win a lot of market share (and others would rush to copy them of course).

This would need to be coupled with a startup advertising budget showing happy people playfully interacting with Denise, and then months later their successful company name on a building.

This is partly non-AI conversational bot, but it is more than that, using machine learning to get better with each sale.

Bob

PS Not much of a step from this to an acceptance similar dialog that would help domainers know why a marketplace accepted name A but not B

Addition: Of course ultimately it ends with an actual purchase and transfer all in one session or when impossible it saves your session so you later resume it when you are ready to move to sale. Selecting and buying domains made easy.

 

[Joe N]

I don't know. But I think if someone could build a much better place to sell domain names, we would benefit. At least our good names would benefit.

But I think that means there is a place for some innovative company with AI skills to build some sort of scaleable solution like the world's best place to find domain names. As I argued in another thread, I think the search in all of the big general purpose marketplaces is far from optimum. They present big lists of results most of which the client would never consider so they lose trust in the marketplace. Imagine an AI based, that learned from each sale how to get better, bot that worked something like this.....

"Hi, I am Denise and want to help you find just the right name today. Can you first just in a few words tell me what you are looking for?"

[response]

"So just to be clear, you are mainly looking for a name for your startup company that is going to use AI to automate help to legal clients for routine issues. Do I understand that right?"

[response]

"Just to get a feel for the sort of name you might prefer, I am going to show you three possible names. Even if you don't like any of them, can you please order them in preference?

[would be shown a made up brandable type name, and two word name, and an unusual single word name or a name from another language, etc.]

"Thanks. That helps me search more effectively for you....

and so on ultimately leading to a curated list of say 15 attractively presented names, and then from that the person chooses 5 to better define what they are looking for and so on.

If someone can make the best way to sell domain names, they will win a lot of market share (and others would rush to copy them of course).

This would need to be coupled with a startup advertising budget showing happy people playfully interacting with Denise, and then months later their successful company name on a building.

This is partly non-AI conversational bot, but it is more than that, using machine learning to get better with each sale.

Bob

PS Not much of a step from this to an acceptance similar dialog that would help domainers know why a marketplace accepted name A but not B

That would actually be fantastic. The brandable marketplace search functions could use a major upgrade, since I agree with you that too many names are presented to potential buyers.

This still targets inbound leads (as opposed to outbound), but there would be a lot of value in that approach. Couple it with a massive portfolio of names (to cover all potential preferences), and you could really have something.

 

[Bob Hawkes]

This still targets inbound leads (as opposed to outbound), but there would be a lot of value in that approach. Couple it with a massive portfolio of names (to cover all potential preferences), and you could really have something.

I think that traditional domain outbound is going to die. I may be wrong. I think though a different kind of loose outbound will succeed. It will use social media promotion, industry interactions, influencers, targeted ads, and a few well respected places to go when you want a domain name with the right advertising that tugs at emotions and success stories.

Some acceptable way to highlight exactly the kinds of names that might be of interest to a client. Not some person you have never heard of phoning or emailing with some domain name that in most cases is not a great fit or even if it is you have zero interest in any new name. More akin to real estate. Look up the domain agents either online or maybe even local presence. I want to live in these areas. I have only this amount of money. I would like a yard for the dog. What do you have? or for domain names My company is going to sell Canadian travel experiences globally with an emphasis on Asian markets. We have this much to spend, or could do a payment plan for x per month maximum. We want to feel elegant but also a little cute. Terms that help define what we are all about are enviro-conscious, unique, local-friendly, safe. What do you have?

Bob

PS the massive portfolio is not that the person dealing with the client will own all the domains. Rather like the real estate market there would be a standard agreement that domainers would share inventory with idea that the domainer who owns the domain gets say 45%, the domainer who actually makes the sale 45% and some centralized coordinating service the other 10%. I realize easier to say than to set up.

 

[Joe N]

PS the massive portfolio is not that the person dealing with the client will own all the domains. Rather like the real estate market there would be a standard agreement that domainers would share inventory with idea that the domainer who owns the domain gets say 45%, the domainer who actually makes the sale 45% and some centralized coordinating service the other 10%. I realize easier to say than to set up.

This would be an interesting place to get to, but there's no way a domainer would be willing to regularly sacrifice that much profit after putting up the initial investment. So I think it would in fact be brokers splitting the commissions.

This model works well in real estate because it's a highly structured and regulated industry, and virtually all real estate holds some value. But compare this to domain names, where the majority hold little to no value, and the industry is global and highly unregulated... The model may very well work, but most domain investors would be unable to access it due to the low quality of their names.

 

[Bob Hawkes]

This would be an interesting place to get to, but there's no way a domainer would be willing to regularly sacrifice that much profit after putting up the initial investment.

Obviously thought needs to be given to percentages, but I think a significant portion needs to go to the person who actually manages to sell the name. Most artistic product commissions are about 60% 40%. Person who creates the art or craft item get 60% and the firm marketing get 40%.

If we go to a model where the sellers invest in seeing clients, helping them choose, follow-up, convince of worth to client, etc. they should be compensated well enough to make it worth while.

 

[frank-germany]

In recent times, AI has been softened to mean algorithms completing tasks. AI is now very simple. Further down the spectrum is Deep Learning which is more complex. We are in times when these terms are absolutely misconstrued due to definition tampering.

as far as I know
AI is defined as

"software that writes software
and updates itself"

 

[Joe N]

Obviously thought needs to be given to percentages, but realistically a big part needs to go to the person who actually manages to sell the name. Most artistic product commissions are about 60% 40%. Person who creates the art or craft item get 60% and the firm marketing get 40%.

I guess a lot of it would depend on industry norms, available options and competition, and how much the domain investor actually needs an agent to act on their behalf.

 

[Internet.Domains]

as far as I know
AI is defined as

"software that writes software
and updates itself"

That's the point that I agreed with @Ategy.

The biggest problem with discussions on AI is that definitions of what AI is ranges so wide from person to person that in the end many people aren't even talking about the same thing.