Method For Creating A Relational Description Of A Formatted Transaction: Fill & Download for Free

This will be a long one, so let’s get started. Note that Codd’s 12 Rules (really 13, since the count starts with “Rule 0”) are about the implementation of the db engine itself, but sometimes you can define a schema that complies with a rule by skipping engine features that don’t comply…These are Codd's 12 rules:Rule 0: The Foundation rule:For any system that is advertised as, or claimed to be, a relational data base management system, that system must be able to manage data bases entirely through its relational capabilities.GK: This rule is interpreted as meaning that the way a “relational database system” can define and access data is through its data definition language, including constraints, and its query language. You shouldn’t have to have extra code to impose constraints in languages other than the main one used to access the db.SQL DDL + DML meet this requirement.Rule 1: The information rule:All information in a relational data base is represented explicitly at the logical level and in exactly one way — by values in tables.GK: I’d argue that database triggers violate this rule, as triggers may hide information or “alias” information, and thus become “information” themselves. And triggers live outside the “table world”.Rule 2: The guaranteed access rule:Each and every data (atomic value) in a relational data base is guaranteed to be logically accessible by resorting to a combination of table name, primary key value and column name.GK: SQL violates this rule by allowing tables to exist that don’t have primary keys. If a table has a PRIMARY KEY, this rule is met.Rule 3: Systematic treatment of null values:Null values (distinct from the empty character string or a string of blank characters and distinct from zero or any other number) are supported in fully relational DBMS for representing missing information and inapplicable information in a systematic way, independent of data type.GK: SQL NULL is not quite “Codd’s NULL”. NULL behavior is hotly debated among relational types, and SQL NULL semantics is “sort of, but not quite” what Codd wanted, so Codd himself would probably argue that this rule is not met.Among other things, Codd wanted two flavors of NULL.Rule 4: Dynamic online catalog based on the relational model:The data base description is represented at the logical level in the same way as ordinary data, so that authorized users can apply the same relational language to its interrogation as they apply to the regular data.GK: Most modern SQL databases have logical tables that allow you to query metadata such as table names, column types, etc. They may or may not be stored as physical tables, but you can query them using “standard” SQL queries. Such databases would meet this rule.Rule 5: The comprehensive data sublanguage rule:A relational system may support several languages and various modes of terminal use (for example, the fill-in-the-blanks mode). However, there must be at least one language whose statements are expressible, per some well-defined syntax, as character strings and that is comprehensive in supporting all of the following items:Data definition.View definition.Data manipulation (interactive and by program).Integrity constraints.Authorization.Transaction boundaries (begin, commit and rollback).GK: This is mostly about whether you have a query language or use some type of direct API to access the database. SQL is a query language that has all of the above, so as long as you’re using a storage engine that supports transactions, this rule is met.Rule 6: The view updating rule:All views that are theoretically updatable are also updatable by the system.GK: Most VIEW implementations are either done as logical views that are essentially live queries to other tables and thus are always “up-to-date”, and/or as “materialized” views that are tables that are auto-updated whenever the underlying tables are manipulated. Either implementation satisfies the rule.Rule 7: High-level insert, update, and delete:The capability of handling a base relation or a derived relation as a single operand applies not only to the retrieval of data but also to the insertion, update and deletion of data.GK: This basically says if you have an INSERT, UPDATE, or DELETE statement which can completely update a record in a well-defined way, you’re good. SQL does cover this.Rule 8: Physical data independence:Application programs and terminal activities remain logically unimpaired whenever any changes are made in either storage representations or access methods.GK: This rule means that you can do stuff like rehost a table from one disk to another, change its storage engine, repartition it, or whatever, and applications won’t change. This is “mostly” true.It’s unclear to me whether this rule covers stuff like creating and dropping indexes and whether performance changes so implied would cause this rule to fail.If “logical impairment” is just about “correct query answers”, this rule is met. If it implies some type of performance criteria, it isn’t.Rule 9: Logical data independence:Application programs and terminal activities remain logically unimpaired when information-preserving changes of any kind that theoretically permit unimpairment are made to the base tables.GK: This rule says that if you change lower-level table definitions, apps should still work without changes. To do this, your apps have to access all your tables using views, which few do.SQL as normally used would not meet this rule, although since it does support CREATE VIEW, you could put together schema designs using SQL that did meet it. (But why - performance would suck, possibly very badly.)Rule 10: Integrity independence:Integrity constraints specific to a particular relational data base must be definable in the relational data sublanguage and storable in the catalog, not in the application programs.GK: SQL has all the syntax needed for integrity constraints, so this is met.Rule 11: Distribution independence:The end-user must not be able to see that the data is distributed over various locations. Users should always get the impression that the data is located at one site only.GK: This rule is a “root rule” for widely distributed databases. Most SQL databases would not meet this rule as apps do need to connect to a specific DB instance and data stored in different instances would have to be queried separately. A few DB engines support “remote table links” and such things, so you may be able to cobble together something that met this rule if you used these + views, but it would be a pain.Most SQL implementations do not meet this rule (but a few do, such as MemSQL or MySQL Cluster).Rule 12: The nonsubversion rule:If a relational system has a low-level (single-record-at-a-time) language, that low level cannot be used to subvert or bypass the integrity rules and constraints expressed in the higher level relational language (multiple-records-at-a-time).GK: Some DB’s do have server-side programming language APIs that don’t use SQL, but for the most part, even accessing data by programming language APIs doesn’t defeat integrity and other table-level constraints - as this could be a security hole and could corrupt the database. So, this is met in any sanely-implemented low-level API.

Summary: Don't try to create a Frankenstein website that leverages all of the recommendations below. You should consider this a toolbox of techniques to test on your site. Use it to figure out the practical and impractical usage of these techniques as it relates to your product and customers, and how these techniques fit into your overall product and design vision.The subject of conversion rate optimization can be discussed in encyclopedic length. So, I'll do it justice and speak about it at length (especially after Tim Morgan's thoughtful response to "What do all the controls in an airplane's cockpit do?"). This is a long read though, so I warned you. I'll also continue to add to it over time as I learn more and more and find other great conversion rate case studies.I think there are two generic ways in which you can frame your conversion rate optimization efforts:First, I'll speak about on-page optimization of design/copy/layout/etc. for increasing conversion rates, since that is the substrate upon which all referring sources are built and that's what I think this question is initially interested in exploring.Secondly, I'll speak about conversion rate optimization from referring sources since there are methods for improving the performance of referring channels (e.g., SEO vs. email vs. social, and so forth). I'll talk about these points in less detail than the on-page optimization elements, but optimizing the funnel of referring sources to the page you ultimately need the user to convert on is equally valuable. I don't think you can discuss on-page conversion optimization without also addressing improvements further up the funnel. Otherwise you miss the point that conversion rate optimization starts well before the user lands on the page intended on converting them.On-page conversion rate optimizationWhat is it: Changing the elements of a page's design and function to improve the rate at which users take a key action on your site.Conversion event: The most common conversion events are sign-ups, logins, purchases, subscriptions, and shares (brought on via the growth in social media). There are several other conversion types that are more granular, but I'll focus on those core types since the methods for optimizing for those core actions are interchangeable with pretty much any other core action you choose to optimize for.1. Buttons vs. text linksButtons tend to have higher click rates than links simply because they are more obvious. Where you currently have text links for major calls to action, try using buttons instead. There are countless tests where people have shown in increase in CTR anywhere from 20–200% over the control where text links were used.An interesting aside: You shouldn't expect a button's A/B test performance to remain constant 'N' weeks after the test, specifically with email marketing. The button essentially experiences click fatigue since you're likely sending the email (such as a newsletter) to a relatively static group of users. Users get used to the design and may realize that the email's text links and buttons link to the same data—so clicking on any of the linked objects, whether a text link or a button, delivers the value they were looking for. You can read blog posts like this one for statistics that demonstrate button click fatigue: http://www.aweber.com/blog/case-studies/buttons-vs-text-links.htmNonetheless, buttons are still optimal next to text links in most cases because they increase CTR even if the increase in CTR begins to fatigue after 'N' number of weeks. The absolute value may not be a sustained incremental CTR of 30% for perpetuity, but at least you can expect that sustained CTR with some rate of decay for several weeks or months after.That's not to say that you can't have a sustained increase in CTR by using a button on a webpage though. Button behavior on websites vs. emails can be quite different. You tend to not have the same click fatigue behavior on website buttons. Typically, if you realize a gain via an A/B test on the web, you can expect that gain to hold.2. Location of the buttonAbove-the-fold is better than below-the-fold. For example, Chitika ran a study that compared the CTR of ads above the fold vs. below the fold, and found that ads above the fold had a 36% higher CTR (http://www.searchenginejournal.com/placement-matters-placing-online-ads-above-the-fold-increases-ctr-by-36/40930/). The same effect generally applies to any clickable action you place above the fold (vs. below the fold).Some marketers have run interesting studies that demonstrate the effect of showing less information above the fold to encourage users to scroll down the page to discover more content and click on your call to action. The evidence is clear that you can encourage users to scroll and click below the fold (http://www.cxpartners.co.uk/cxblog/the_myth_of_the_page_fold_evidence_from_user_testing/).Take a look at this heat map from the Bristol Airport website and an eye tracking study that was conducted on it. This shows that by showing less content above the fold you can get more users to track down the page and click on those actions.But from a user's perspective, requiring less work on their part generally seems like the right thing to do. So, if you're capable of delivering enough information and value to users that encourages them to click your call to action button above the fold, it seems reasonable to do that instead of coaxing them below the fold by purposefully omitting information at the top of the page. At least, think deeply about minimizing the number of available actions above the fold to simplify their decision-making.3. Size, color, and contrast of action buttonsYou want to make the call to action button obvious but not intrusive. I think Square does a great job at this. Check out their home page at https://squareup.com/. The button color contrasts with the full-bleed image and other page elements. It is large, but not freakishly big, and has a bunch of calls to action surrounding the button and on the button itself.If you're looking for inspiration for more call to action types and styles then check out this list here: http://www.webanddesigners.com/35-creative-call-to-action-buttons-for-inspiration/4. Improve page speedAn often overlooked but hugely valuable component of conversion rate optimization is the speed of the page you are trying to deliver to users. Several tests have shown the benefits of site speed. For example, back in 2007 Amazon released a study that showed that they lost 1% of sales for every 100 ms of additional site latency. You can find more test studies on site speed on the ConversionMedic blog: http://www.conversionmedic.com/website-performance-optimization/ Think about ways in which you can reduce the payload of images/CSS/JavaScript/etc. (especially on critical webpages) to reduce load times. Or figure out ways to chunk the loading of different components of the page to make sure critical components load before non-critical components.Through various tests I've been a part of at both Facebook (product) and Twitter (product), I've seen firsthand how speed improvements can dramatically increase conversion rates. I actually think poor site speed is one of the factors that caused the demise of MySpace (product). Think about it... a slower site means less friend requests, less status updates, fewer picture uploads, fewer ads being clicked on, fewer invites sent, etc., etc., etc. All of those actions contribute to what people would generally refer to as the network effects of a social networking product. Site speed reduces the momentum in the flywheel of network effects. With less momentum, you have less growth, and with less growth, you have an increased chance of losing. Facebook came in and was faster, simpler, and had better network effects. By focusing on critical elements like site speed, Facebook was able to succeed.Google has provided a great Page Speed Testing Tool for anyone to use here: https://developers.google.com/pagespeed/. It generates detailed reports giving specific site speed recommendations. YSlow is another great tool for looking into site speed issues.If you're a WordPress site owner, site speed can be a real issue (especially if you're a heavy plugin user). Try using the Plugin Performance Profiler to look at individual site speed implications for each plugin in use: http://wordpress.org/extend/plugins/p3-profiler/It generates reports like this that help you diagnose latency contributions per plugin:Thanks to unbounce.com for pointing out these site speed analysis tools (http://unbounce.com/conversion-rate-optimization/how-to-improve-conversions-by-increasing-page-speed-tips-tools/).5. Headlines, subheadlines, and body copyUsing a prominent headline that delivers the core value proposition to users is really important. I would imagine that conversion rate to signup would decrease if Square (company) removed the "Start accepting credit cards today." language from above their sign-up form. Test different language in your headline to see what resonates. If you don't have enough impressions on your site to quickly run an A/B test, then try using Google Ads or Facebook Ads as a platform for testing the language that resonates best with your users (since you can get a lot of impressions relatively quickly, especially with Facebook ads). For example, the headline in your Facebook Ad that gets the highest CTR may be an ideal candidate for the headline on your website's home page. When it comes to body copy, less is usually more. You don't want to overwhelm the reader with too much text to read.I did a search in Google for "small business websites" and clicked on a few of the ads that I was shown to see what the landing pages looked like. One landing page looked like the image below. Clear calls to action, not too busy, a little bit of text to read, and some testimonials for validation. They even have a video introduction that explains the product. This is a pretty solid page:Source: http://www.volusion.com/But another ad landed me on the page below. My knee-jerk reaction was to leave immediately since it seemed like too much to figure out! I do the same thing with any blog post or email that is overwhelmingly wordy and doesn't lead my eye to the valuable bit of information on the page such as data, graphs, bullet point lists, etc. This site also does a terrible job with the use of contrast. The "visit site" action links aren't as obvious as the orange calls to action for "14 day free trial" in the site above:Source: http://www.webhostingfreereviews.com/Sure, there are several calls to action. And since they are bidding on a very expensive keyword they must be managing this to an effective ROI (I hope!). But with all of that text, I'm sure they are losing conversions. In many cases, less text is more when it comes to consumers. Remember, you want your product to be more like In-N-Out than Denny's (company). Offer a short list of excellent products with clear value, not a long list of crap you can barely digest.6. Think about how color impacts mood, emotion and decisionsAs Colm Tuite has shown in his answers to "Is there a science to picking the colors that work well together in a design, or is it just subjective?" your site's use of color can also set the tone for driving conversions (or not). For example, here is a breakdown of color connotation amongst users in the Western world:Take a look at his answer and think about how color may impact conversions and then try a few tests to see if this applies.7. Use of imagesImages can increase and detract from conversion rates. I've run several A/B tests that have shown that clickable images on home pages often lead to fewer net signups overall (even including revisits in a 30-day window after dropping previous sessions). I think Pinterest understands this concept and doesn't let users "escape" from the primary action such as on their "Request an Invite" page: http://pinterest.com/landing/.Notice the carousel of rotating images at the bottom. They are not clickable links so the users isn't going to be driven away from the page intended to initiate signups. But the images are there nonetheless to help convey the value of using Pinterest. From my previous experience, I would anticipate a 5% drop in the rate of email address submissions on that page if the carousel of images was clickable. I've tested clickable vs. non-clickable images on logged out home pages in the past, and in every case, the non-clickable images version increased signups by 5–7% and the quality of the users relative to the control group remained constant.Images can also be used to explicitly improve conversion rate. One of the sayings in Internet marketing is "chicks get clicks." (That's another way of saying "sex sells.") The consulting company ion interactive ran a test for a gaming company and found that the below design outperformed the same design but with less emphasis on breasts by a whopping 35%.Here is a side-by-side comparison of two of the landing pages that were tested with the version on the right getting 35% more registrations:Source: http://www.neurosciencemarketing.com/blog/articles/sex-sells-sometimes.htmThis isn't to say that every website should start putting half-naked women (real or fake!) up front on their landing pages. Clearly this is an effective model for some video game developers. Zynga has also used this approach effectively with ads for games like Mafia Wars.But in general, there are a few key points when using images for driving conversion rates, especially for sites that sell products:use high resolution, not cheap clip arttry image carousels or 360° image rotatorsenable a large product image by default with clickable thumbnails for additional product imagesuse image thumbnails in site search features like BrickHouse Security has done (http://www.internetretailer.com/2010/04/29/product-images-site-search-window-boosts-conversions)More info here: http://conversionxl.com/how-images-can-boost-your-conversion-rate/8. Show validation/testimonialsSometimes testimonials can drive validation, especially for B2C products. Mint.com uses testimonials on their home page. I don't have data on its efficacy, but you should consider using testimonial validation on your product like Mint.com (product) does. Testimonials don't always make sense. For example, Facebook.com would never need testimonials, but a financial services product might.Source: https://www.mint.com/9. Leverage video to demonstrate product and drive conversionsThis subject was brought up by William Wai Wong, who has a great deal of knowledge when it comes to testing video. Video can be a great asset when trying to explain the product and value add of your product. I think that video is especially valuable when your product has anything to do with personal finance, wealth, or the handling of information that the user may consider sensitive. Dropbox (product), for example, has a video on their homepage directly above the "Download Dropbox" button:Dropbox's success with moving to a video-based landing page has been widely documented. Path.com also offers a good example of a first-visit autoplay video walkthrough of their iPhone app. I think that video is very helpful in Path's case since there is a general sense of "not another social network" fatigue and users are becoming increasingly sensitive to sharing personal information. But this video aligns nicely with Path's vision because it wants to focus on closer, more tightly controlled networks of friends. They actually offer a couple of videos on their homepage:10. Pairing intent with landing page contentWhen William Wai Wong first joined Bloomspot, a startup in the daily deals space, one of his first observations was that every competitor in that industry copied the generic landing page Groupon used in all of their marketing campaigns.Despite targeting a wide range of keywords (salon, spa, dining, etc.) on paid search, broad demographics on Facebook, and a range of placements on display networks, Groupon and other daily deal sites use the same landing page. While this saves engineering time, it didn't take very long to build a generic template that allowed them to test text, background image, and other elements tailored to specific campaigns or verticals in their space.The result was over a 20% spike in our conversion rate.Example: Groupon, LivingSocial and other daily deal sites use a single multi-purpose landing page for the following keywords (spa deals, restaurant deals, fitness deals):However, moving to a vertical and intent-based approach pairing the background image and supporting text to match intent, Bloomspot saw significant lifts in conversion rate:As William Wai Wong points out, what works for your mainstream competitor doesn't necessarily work for you. Or, you may in fact, be more sophisticated at it than the mainstream competitor.International implications and costs associated with using a videoThere are a few things to take into consideration before hopping into the deep end and making a video. First, these videos aren't cheap. I've spoken with several vendors and one- to three-minute videos can range in cost from $5,000–20,000. For a startup, that could be a healthy chunk of change. You may want to consider testing as many other "free" conversion rate optimization components as you can before deciding on making a video. You may find that a video isn't needed if you can get enough CRO optimization out from the other methods.Secondly, think about the international implications. If you are starting to expand globally into several languages and countries you'll need to (1) pay for the production cost of translating your video into each locale/language, and (2) consider the latency implications of a video in a country with low broadband penetration. Sure, it's trivial for someone in San Francisco to stream a video. However you shouldn't assume the same even in major markets like Brazil and India where broadband penetration is still nascent compared to markets like the US and the UK. Remember, latency kills conversion rates. A slow video may mean high abandon rates and low signups since the two tend to be inversely correlated. Consider these costs as you work through the CRO tools.Simplify signup formsWhen it comes to getting users to sign up, there are several best practices you should follow. Some of the largest consumer acquisition optimizations I've ever had came from making improvements to signup forms.11. Don't ask for unnecessary fieldsI think that Yahoo! (company) is one of the worst perpetrators of producing scary signup forms and asking for far too much information. Take a look at this signup form for registering an email account:Source: http://www.unmatchedstyle.comThey do not have to ask for gender or birthday. That data is entirely self-serving for Yahoo! since they can use that data to better target you with ads. They could also get rid of the CAPTCHA requirement (which typically has a 10% failure rate) on your first attempt to signup from an IP address. After the second attempt to signup within a few minutes from the same IP address they could then hit the user with the CAPTCHA requirement because that could tip Yahoo! off that you're a bot trying to create multiple accounts. But on the first attempt from an IP, there is high confidence that it is a user and not a bot—they shouldn't be required to provide CAPTCHA. Not to mention that the form is so long that the primary call to action is below the fold.12. Automate fields when possible and use inline validationTwitter does a great job at using validation and automation in their signup form. On the homepage you start signup by entering your name, email address and password. When you hit the button to submit the form they land you on the second page of signup with some fields pre-populated with recommendations (such as the field for choosing a username) and validation that the other fields you entered are correct:They then use a simple algorithm to make username suggestions that you can click and select. Imagine having a common name like Melissa Anderson and being the 53rd Melissa Anderson to try and sign up for Twitter. It's unlikely you can choose anderson.melissa, melissa.anderson, melissa1, etc. since common permutations are already in use. Twitter recommends a name for you based on the available name space. Surely this improves signup rate.13. Give social signup options and use social pressureRegistration using Facebook Connect or Twitter authentication makes signup much simpler for users. You can use the data culled from the Facebook and Twitter APIs to pre-populate several fields such as their profile information, email address, etc. and then just ask the user to accept permissions and maybe create a password.Digg does a great job using social context (the Facepile plugin) in tandem with a bold Facebook Connect social signup option when prompting their social reader integration with Facebook's Open Graph. Mixcloud claims to have increased their conversion rate to signup by 200–300% using the Facepile plugin along with Facebook Connect (http://www.netmagazine.com/news/report-f8-london-part-2-111504). Although most sites won't see the same level of improvement, I've chatted with several other developers and designers that have quoted a 20–50% increase in signups when the signup button is accompanied by the Facepile plugin. Here's what their implementation looks like for Digg:Hipmunk also does a great job at offering social sign in options in their modal signup dialogue:You'll also want to look at the downstream value of the users produced depending on what social signup option they choose. It's common to find variance in the value per user produced whether they came from Facebook Connect or another option. Depending on the value of the user produced you should look into promoting one button over the other.Furthermore you should focus on one social signup button depending on the referrer of the users. For example, if I visited Hipmunk by way of a link on Facebook that I clicked, I'd test showing Facebook Connect as the primary method of showing up. They might be able to optimize signup rates by doing that because the signup process is then within the context of the site they were referred from.Email MarketingWhat is it: Sending transactional or bulk email to users who have opted into receiving email from you or your business/product.Conversion event: From an email marketing standpoint, the conversion event is typically defined by a click in the email that leads the user to a destination page where a final conversion action is emphasized. In this case we'll talk about things you can do to optimize your email funnel for getting the most clicks possible. In other words, conversion rate optimization for email is about increasing the % of email recipients that click on your email and land on your site or mobile app.14. Subject lines and sender addressEmail Optimization 101 asks that you test your subject lines as well as the sender address. In your subject lines, you can test the length (long vs. short) as well as use-descriptors like "Exclusive." In this case study, the client was able to increase email open rates by 8.2% by using a longer subject line that was more descriptive of the event and leveraged enticing descriptors (http://www.marketingexperiments.com/blog/research-topics/email-marketing/subject-line-testing-relevance.html).Other best practices include using social signals. For example, Summify is well known for their excellent email digests. They provide a service that aggregates the top stories across all of your social networks (e.g., LinkedIn + Twitter + Facebook) and give you a daily email with those top stories. In the subject line, they list the names of your social contacts to drive high open rates. A subject line could read, "Here are your top stories from John Smith, Sally Jones, and Rob Johnson" (all of which would be friends/connections of mine on Facebook/Twitter/LinkedIn). Anecdotally, users appreciate that experience as well because the personalization drives relevancy:You should also avoid sending your email from address like "[email protected]." Users have come to associate unpersonalized sender lines like that to spam accounts. You should test your way towards higher open rates with personalized sender lines just like you can personalize the subject line.15. Test various lengths of email copyLess is usually more. Several studies have shown that too much text leads to lower click rates and higher opt-out rates, especially for transactional emails. Test the length of copy you include in your emails to see what works.16. Plain text vs HTMLMany designers and developers get caught up in trying to build the perfectly designed email. But the same rules of design for the web don't apply to email. In many cases, plain text email performs better at scale than HTML email because HTML-rendering across email clients varies wildly (your design could be completely broken from one mail client to the next) and because transactional emails are usually meant to deliver a click (e.g., an email from Facebook notifying you that someone wrote on your wall). Test whether plain text does better than HTML.17. Sending email at the right timeIt's amazing how important it is to send email at the right time of day and the right day of week, especially when sending digest emails once a day/week/month. Below is a graph of the typical open and click decay curve that happens after an email is sent out:Source: http://www.ryansolutions.com/blog/2012/three-core-principles-behind-successful-email-optimization/Generally speaking you can expect 90% of clicks/opens to happen on a particular email campaign within 48 hours of being sent. Usually, 80% of opens/clicks happen within 24 hours.Also consider the time of day that you're sending an email. MailChimp (product) has an amazing study on the optimal time of day for email open/click activity:As the graph demonstrates, the optimal time for opens/clicks is 2–5 p.m. local time. If you plan on sending a weekly or monthly digest email to your users, make sure you don't deliver them to inboxes at midnight. You'll dramatically impact email performance. Check out the rest of MailChimp's recommendations in this awesome blog post about email optimization: http://kb.mailchimp.com/article/when-is-the-best-time-to-send-emails/Facebook AdvertisingWhat is it: Advertisements on Facebook for promoting Facebook Pages/Applications and external websites.Conversion event: In most cases advertisers are trying to get Likes for their Page or installs of their Application (e.g., a Zynga game). Sometimes advertisers are driving users to a custom landing tab on their Facebook Page for lead generation purposes for something like an email database. Below are the top factors to consider when thinking about optimizing Facebook Ads as a conversion funnel.18. Country targetingThis is by far the biggest factor to consider when setting up Facebook ads. The cost to drive a conversion in the United States could be 5–10x higher than in other countries, including other major English-speaking markets like the United Kingdom and Canada. This is primarily influenced by the supply side where there are fewer advertisers bidding for impressions in countries outside the United States. However, as advertiser acquisition scales in other countries, they will continue to place a constraint on international Facebook ads inventory and the price will float over time. Be sure to setup separate advertising campaigns/targeting based on country so that you can independently optimize each locale.19. Ad creative (title, image, and description) and image rotationSome images will get higher CTRs but lower conversion rate while others will get lower CTRs but a better conversion rate. Here is one example provided by AdParlor:Make sure your ad text is relevant to the product and the underlying landing page. You also need to cycle images often because Facebook ads typically experience a rapid CTR decay since you're often showing the ad to the same set of users over and over again, unlike Google ads where you tend to have higher variance in the cluster of users (unless someone goes back to the search results and searches for the exact same keyword day after day). With Facebook ads you are being targeted based on your interests, which is a relatively steady set of data associated with you as a user. That's why you are likely to see the same ad multiple times. But rotating images regularly you can optimize CTRs for your ads, which usually means your CPC gets priced lower and hence the cost you pay per conversion decreases.20. Target demographicGenerally speaking, women click on ads more than men, and younger ages click more than older ages. Think about the target demographic your product appeals to and look at performance data to see which demographics cost more or less to convert on.21. Pipeline from click to conversionYou also want to require as few clicks as possible. I recently got an ad on Facebook from AppSumo. The ad looked like this:Not only did the ad use social context by showing me that one of my friends likes it (which boosts CTR) but the underlying landing page that drives a conversion event is really simple as well:The conversion event is driven by a single click after I enter my email address and click "show me the deal" or if I just login with Facebook Connect. It is not uncommon for Facebook ads to have a conversion rate of 40–60% after the click if the ads are well targeted and the conversion event is simple.22. Market saturation of your product/applicationAs you start your Facebook campaign you can define "exclusion targeting" to make sure you don't serve ads to users who have already interacted with your Page/Application. But imagine if the total potential audience for your business or product is realistically 10,000 users. After a few months let's assume you have reached 5,000 of them. As you saturate the market your CTR drops and as your CTR drops your CPC may go up since you're now starting to target semi-interested users. According to John Marsland, Acquisition Manager at Zynga, "Once you've acquired the majority of the early adopters, CPIs in a particular market can increase by 3x-5x from your day cost-per-install. Fortunately, this change doesn't happen overnight. Refreshing creative, creating a compelling message, and refining your target audience are the best ways to combat application saturation."23. Market conditionsAd supply continues to increase as more advertisers enter Facebook's system. There is also a question of demand fatigue in that users could be getting tired of applications and deals. Sudden changes in market conditions from one month to the next can cause your cost-per-install to change dramatically over time. You have to stay vigilant when running Facebook ads to also adjust for these market changes.You can find more best practices on Facebook Ads optimization from Facebook here: http://ads.ak.facebook.com/ads/FacebookAds/OptimizationGuide_060611.pdfSearch Engine Advertising (AdWords)What is it: Bidding on a cost-per-click basis on Google ads based on user search keywords.Conversion event: Google is known for being a demand fulfillment product in that the conversion rate downstream from clicks on ads is very high since the searcher already has a strong idea for what they are looking for. The ads are intended to deliver on that user demand. Below are some of the best practices for optimizing your Google ads.24. Optimizing for Quality Score and ad performanceAdWords Quality Score is a score as defined by Google is "a dynamic metric assigned to each of your keywords. It's calculated using a varity of factors and measure how relevant your keyword is to your ad group and to a user's search query. The higher the keyword's Quality Score, the lower its cost-per-click and the better its ad position." The breakdown of Quality Score factors includes these elements: CTR of the ad, relevance of the keyword to its adgroup, landing page quality, relevance of keywords to the creative (ad copy), overall account performance history, and other elements like page load time.Here is a list of techniques from PPChero on improving ad CTR (http://www.ppchero.com/how-to-increase-your-ppc-ctr-%E2%80%93-it%E2%80%99s-staring-you-right-in-the-face/). It boils down to:test using trademark (™) or registered trademark (®) symbolsuse calls to action or incentive-based action words (e.g., "buy now, limited offer!")enable sitelinksuse Dynamic Keyword Insertion (DKI)You can do things like setup negative keywords to ensure you don't bid on keywords that don't convert for you, thereby cutting wasteful spend. Also try lots of landing page rotation and testing to find the best combination of keyword + match type + ad text + landing page.Social Sharing (Like/Twitter buttons)What is it: Optimizing for the placement and location of social features on your site (e.g., the Like and Tweet button) as well as contributions to social media site to drive qualified, targeted value to your site.Conversion event: In this case, the conversion event is either a click of the Like/Tweet/etc. buttons or an interaction with a status update or tweet that drives referral traffic to your site.25. Testing placement of the Like/Tweet buttonsYou can use products like Kissmetrics to run quick tests to figure out which social button placement on your site generates the most clicks. The same general rules apply to the lessons at the top of this answer regarding button placement since Like/Tweet buttons are also conversion events. Generally speaking for blog owners, the Like/Tweet buttons are best when placed immediately next to or below the title of the blog post and immediately above the comment box on the page. Typically you'll get fewer Likes/Tweets if those buttons are placed in the sidebar of the site. Here's an example of a typical placement of social buttons on AllThingsD (although they should probably remove the "share" and "print" buttons—I'm pretty sure people don't use those).26. Intelligent status updates and tweets using analyticsYou can use products like Tweriod or Timely to figure out when your followers/fans are online. Then you can schedule status updates and tweets via products like Crowdbooster so that your social media production coincides with when fans/followers are online. Rand Fishkin at Moz found that he had slightly higher tweet interaction rates with his tweets that were shorter in length than longer in length.Source: http://www.seomoz.org/blog/calculating-and-improving-your-twitter-clickthroughrateThere's lots of other amazing data points on the appropriate use of Twitter here (http://www.whitefireseo.com/infographics/twitter-psychology-for-marketers/) thanks to the team at White Fire SEO.SEO (Search Engine Optimization)What is it: The art and science of improving on-site and off-site signals that help a web page rank higher in the search results for a particular keyword or set of keywords.Conversion event: Search engines are known for being demand fulfillment businesses (i.e., the results tend to satisfy the demand of the searcher more so than they tend to generate the demand). Not to say there is no branding effect in search engines since discovery of a new brand can be serendipitous, but generally speaking, the user knows what they want and the search results satisfy that demand. So from a conversion rate optimization standpoint, you optimize conversion rates from SEO referrers by optimizing for the click in the search results. Here are some best practices:27. Page meta dataUse targeted keywords in the meta title and meta description of the page since most search engines tend to highlight matching keywords in titles/descriptions from the user's search query. So if I search for "adidas running shoes" any organic or paid search result will show bolded keyword matches like in the screenshot below. Those keywords attract the eye of the searcher, increasing the chances that the user will click on your search results, hence increasing the conversion rate of a user clicking on your site vs competing sites. Using those keywords in the meta data also improves the chances of your site ranking highly, which is a conversion optimization in itself.28. Calls to action in the page metadataTest using calls to action in the meta title and description such as, "sign up for free," "free shipping on all orders!" etc. to see if clicks on search results with calls to action correlate with a higher conversion rate on the underlying page. Twitter does a great job with their meta descriptions by using the "sign up for Twitter to follow" call to action:

The simple answer is, write an API accessible on the Internet so Siri can access your data for service delegation. This will be the primary method I believe Siri will use to form developer relationships. I will call this a Back end API for clarity.It is very clear that Apple has not even suggested that they will have an API available. However I think one can surmise from the past and the need for Siri to be more functional API access will become available if not encouraged. Apple has already created primary API relationships with Wolfram|Alpha, Yelp and Wikipedia.Siri Is About Completing TasksIt is important to understand the basis of Siri. Siri is not a search engine. Siri is a Task Engine or perhaps a Do Engine. This means the end result is the completion of the task.Siri's power comes from accessing APIs that are called upon once the task(s) have been extracted. A developer can be very certain that there is absolutely no need to be coding for the Context, the Conversational Interface, Speech Recognition, Speech To Text, Text To Speech, Dialog Flow. One can also be reasonable sure that Siri will produce well presented Speech to text results with full context for the developer's API that may include Location Awareness, Time Awareness, some degree of Task Awareness with appropriate levels of personal data based on the type of service and Domain (classification of service).To complete a task there can be dozens of APIs accessed. Over time there could be hundreds. Siri will learn over time and optimize, however new API development will create a constantly dynamic environment.APIs EverywhereSiri uses APIs as a service once the Task And Domain Models have been determined by the Intent engine. There are any number of intermediary service(s) and APIs that may be accessed to complete a Siri task.This is in contrast to APIs using Siri as an overlay to an app on the iOS device is something else entirely. In many ways this forces Siri into more of a Speech Recognition system. I will call this a Front end API for clarity. At first blush it may seem that the benefit to the developer is with a Front end API because the connection would be directly to the app. However this does not seem to be the modle that Siri will use, although it is possible that this sort of API would be made available. I believe in the short to medium term Siri will be more focused on the Backend API.How Can A Developer Get Ready?Get to understand the theory behind Siri, its history and how it really works in the real world. This is a starting point: Why is Siri important?.Lets assume that you will be creating a Backend API for Siri to access your data sets. This would be in the sprit of the typical API you will find at ProgrammableWeb.com, for example. However, there would be a number of things a Developer can do to optimize for Siri's in bound and out bound API. One can gain much knowledge on what Siri may like in an API by studying the API of one of it's major organic relationships, Wolfram|Alpha. I am told that Wolfram|Alpha did not change much for Siri other than optimize the primary results to Speech rather than Display, although some tasks will produce a visual result with Speech.There is a treasure of information that can be found on the Wolfram|Alpha API. I would start this journey here:http://products.wolframalpha.com/api/documentation.htmlIt is very important to gain as much experience with the concepts of the Semantic Web (http://en.wikipedia.org/wiki/Semantic_Web), Artificial Intelligence, dialog and natural language understanding, machine learning, evidential and probabilistic reasoning, planning, reasoning and service delegation. Much of the work will be done by Siri, but it is important to understand these concepts so as to be able to predict how Siri's APIs will be developed.The most important understanding would be to study Ontology (http://www-ksl.stanford.edu/kst/what-is-an-ontology.html) and knowledge representation. I answered a "Self Posed, Quora Question" here: Is Quora important to Siri? on this subject as I feel Quora will become very important to Siri, but Quora would need to develop schemas that would be optimized to Siri.This paper is the definitive work on Ontology:http://www-ksl.stanford.edu/kst/what-is-an-ontology.htmlArranging data in an Ontology will be of the greatest benefit. However, Siri may not demand this, but I feel rather strongly that it would give APIs higher priority that have done so. Of course, most data on the Internet is represented in relational database schema. Taxonomies or relational database schema is in stark contrast to how Ontologies work to organize information. No matter what the domain or scope, an Ontology is a description of a world view of the data. This view might be limited and minuscule, or it could be global and expansive. However, unlike alternative hierarchical views of concepts like taxonomies, Ontologies will have a linked or networked “graphical” structure. Multiple things can be related to other things, all in a potentially multi-way series of relationships.In the simplest form, Siri will no doubt parse a simple text string to the API and look for a simple text string in return. However I believe that data formated in Ontologies would allow far more potential for Siri for questions that are inbound as well as data Ontologies on the outbound as the answer.To see a road map of how Apple may view the future, there is a patent for that:http://c2499022.cdn.cloudfiles.rackspacecloud.com/wp-content/uploads/2011/10/iPhone-Siri.pdf"Local Services (including location- and time-specific services such as restaurants, movies, automated teller machines (ATMs), events, and places to meet); personal and Social Memory Services (including action items, notes, calendar events, shared links, and the like); E-commerce (including online purchases of items such as books, DVDs, music, and the like); travel Services (including flights, hotels, attractions, and the like), navigation (maps and directions); database lookup (such as finding businesses or people by name or other properties); getting weather conditions and forecasts, checking the price of market items or status of financial transactions; monitoring traffic or the status of flights; accessing and updating calendars and schedules; managing reminders, alerts, tasks and projects; communicating over email or other messaging platforms; and operating devices locally or remotely (e.g., dialing telephones, controlling light and temperature, controlling home security devices, playing music or video, etc.)"Business ModelsDo not assume that just because, for example, Yelp already is part of the organic API, that another service that may rate a dining experiences would not be needed. There is no doubt that Siri will use a form of Consensus Building to arrive at the best results. Thus many, if not 100s of competing services could be accessed. In fact, if an API is for example, authoritative on just Pizza, Siri would over time be biased to this API's results if it turns out to produce the best answers.I can see literally thousands of APIs that will be needed as Siri grows. I am predicting the same sort of land rush we have seen in the app store developing around Siri. How the compensation models will work under Apple is a closely guarded secret at this point. There is a revenue model that will be created around Siri. But clearly Apple is not going to do anything to make the experience become "commercialize" in a sort of Pay For Performance or Pay For Click model. I can say that the financial ecosystem that is built around Siri may become larger than one could imagine. My advice at this point would be to build the best APIs and focus on the highest Quality and not Quantity and assume that at some future date it will pay off. This may be obvious, but I feel it can not be overstated.Siri's Quasi APIsThere is a very interesting and perhaps useful Quasi APIs available today, although limited and a bit quirky.Sending SMS messages, Siri does this very well and can read back the results. One could, have access a Q&A system, setting the state of some web app, checking in, status updates to social networks.The CalDAV (http://en.wikipedia.org/wiki/CalDAV) protocol. Siri allows for interaction in real time to any CalDAV calendar. One could use this feature as a communication system parsing data in and out of a system that could implemt this protocol. More info here: How does CalDav integration work with Siri and what commands will work through this implementation?This Is Very Early DaysPeople are just now getting the first iPhone 4S models (10/13/2011). I have advised my clients, and just about anyone who would listen, to be Siri ready over the last Two years. Frankly I am certain some thought I was goofy about this stuff. However I have studied this stuff since I first saw LISP running on a PDP 11. Today we have the perfect convergences of Moore's Law and the amazing work from SRI International/Stanford University to thank for Siri. It is estimated that over $750 Million Dollars was invested in the Meta technology behind Siri.When the Siri API opens up, the explosion of opportunities will be boundless and Siri's usefulness will soar.