Effective advertising on Google

Last week (May 14 and 16) we asked teams how they were buying ads on Google. The answers were not very coherent. “We just say how much we want to spend, and what words.” But why those words? How much did you end up paying for every potential customer you brought in? How do you decide if these words are working? If they are cost-effective?

Learning how to bring people to your web site is, obviously, crucial. Right now you have a relatively short term goal: Maximum learning about what works, subject to a limited budget.

There are several ways to bring in people, including:

  • Referrals from sites that mention you (discussed in class on Wednesday)
  • Paid search – showing ads to people who are searching for specific terms that suggest they might be interested in you. This is our topic. You will see it referred to as S.E.M., Search Engine Marketing.
  • Organic search – Getting ranked near the top on a general search by potential customers.

Your task this week is to learn about, devise, and implement a good strategy for paid search. This is a big and important field in e-commerce, for obvious reasons.  Document what you did, and measure how well it worked. Iterate fast to get better results. Show that you know how to run fast experiments.  (see below)

Some Resources

Initial explanation: https://searchengineland.com/guide/what-is-paid-search

A short “cartoon” explanation of many factors affecting  success. For example, how to target different customer characteristics. https://www.wordstream.com/blog/ws/2015/05/21/how-much-does-adwords-cost

Google’s own pages for AdWord beginners are good. Start with them, and the associated videos. https://adwords.google.com/intl/en/home/how-it-works/search-ads/
Google’s adwords help center, still quite elementary: https://support.google.com/adwords/answer/6146252

I expect that there are also many good videos on these topics. If you find some, please post comments, or send emails.

A page listing multiple AdWords tutorials. https://www.disruptiveadvertising.com/adwords/adwords-guide/
One specific tutorial:     https://neilpatel.com/what-is-google-adwords/

Two books, downloadable on Springerlink.com:

  • The Definitive Guide to Google AdWords: Create Versatile and Powerful Marketing and Advertising **Campaigns**


Use a planning tool such as this one to analyze your choices and make a  plan. You don’t have to implement it (e.g. you can be unrealistic about your budget). But we want to see if you understand this technology.  Here is an example of a planner: https://support.google.com/partners/answer/7337243.
There are many others around.

Why is this complicated?

There are multiple levels for designing your advertising. Google started with a sophisticated “double auction” system, invented by a well-known economist, and it set them on the path to riches. But it is based on some sophisticated choices that small companies are not set up for. Recently, Google seems to have shifted to a system that hides what is  going on, and in particular does not show you the cost for different words. Despite this simple interface, behind the scenes Google still uses bidding to set a constantly changing price of ads on different search terms. You need to figure out the basics of pricing, although you certainly do not need to understand the underlying algorithms. Fundamentally, you want search terms that people who are interested in your site/product/service will use often, but that are special to a narrow group so that the ad price will be low. You can add negative words and use other methods to reduce who sees your ad.

Fortunately, Google makes it easy to find relevant, high-demand keywords. Simply enter your keyword ideas into the AdWords Keyword Suggestion Tool, and Google returns lists of similar keyword terms along with their estimated monthly search volumes and various other metrics. Estimated costs per click are shown, but these figures are often incorrect. Definitely pay attention to the level of competition for each keyword term; keywords with higher levels of competition are being bid on by more AdWords users,

A refinement is that you only pay for people who click on your ad. Therefore as much as possible:

  • Write your ad to attract people who might want your product, but nobody else.
  • Use different ads to appeal to different groups, associated with different search terms.
  • Experiment with different ads just as you experiment with search terms. Be systematic (keep good records of what ads you tried and what they produced)
  • The key ratio is not the click-through ratio, but the ratios of conversions per click. A conversion is a positive event by a potential customer. You can define it how you want, such as filling out a request for information, making a purchase, etc.


Keeping and generating revenue from customers

May 14, 2018

Parand’s view graphs. UCSD Startups_ Metrics 5-14

Acquisition costs, burn rate, runway, and other key metrics for startups.


Web site with live  dashboards of startup firms:  http://baremetrics.com/open-startups.

Assignment: Look for some outliers on one of these metrics, eg.

  • Recurring revenue,
  • Churn
  • ARPU

Assignment: Interview target users

For next week April 30 – May 4,  there are two written assignments and two readings.


Pages  52  to 57   of a very short book called TALKING TO HUMANS: Success starts with understanding your customers, by Giff Constable. The book’s home page is  talkingtohumans.com. The book is available in several formats including downloadable  PDF version (free).   You will want to read the whole book – it is even has cartoons!

A brief excerpt from a book on product development, finding customer needs A more academic style, but still words of wisdom.

cartoon1. Interview reports

Due Monday in class: Interview reports for at least two people in your target audience. If the interviewees turn out to be inappropriate, do others. Email reports to all three faculty. Format them so we can look at them on screen.

For each “set” of interviews,  report on your selection strategy and attach your pre-written interview guide.  Individual reports include information about who, what, when, where, direct quotes (not the whole verbatim interview), and interpreted statements. (See readings – how do you interpret what they said?) Finally, write a summary of the whole package. What answers did you receive to your hypotheses? What other significant insights did you hear (even if you cannot yet be confident that they generalize)?

2. Weekly team update

As discussed in class. Due Wednesday by email by 2pm. Other dates can be negotiated for each team. Include the URL for your team web site.

Put serious thought into what your hypotheses really are. For example, major hypotheses should answer several of the following questions. If they don’t, then they are not major. Keep digging until you figure out the hidden assumptions  you are making, and restate them as hypotheses.

  • If this is not correct, will we want a major change in our target market?
  • If this is not correct, should we reconsider the concept?
  • Will an answer resolve an uncertainty which has slowed down our progress?

Some of your investigations will not start out as hypotheses.  For example, “How many different news sites do people look at over the course of a day?” Or “How can we identify, and then attract, people who are most interested in multiple points of view?” Once you answer such a question it leads naturally to a hypothesis such as “We should go after XYZ customer demographics.”

How to conduct interviews:

There is a big difference between casual interviews, and systematic ones. See the readings for advice on conducting interviews. For starters

  • Two interviewers  with agreed on roles.
  • How will you record results? Direct recording intimidates people if you don’t already know them. Take at least one photograph.
  • Where? The best place to interview is relevant to your product concept. For example, while they are online so they can demonstrate something. For clothing, have them show you what they already own, or walk through a clothing store with them.
  • Use props. Photographs of similar ideas that will be familiar, for example. There are several purposes: to evoke a situation or mood to enhance recall and creativity in the interview, and also to see what they currently use/do.
  • The earlier in your development process, the less directive your interview should be (especially the first few minutes). Nondirective questions allow unexpected results that may give you new ideas  “What kinds of things do you like to look at on social media sites?” is better than “Do you look at memes when you use Facebook?” Later in the same interview, you can narrow in on something, but it should reflect what they told you earlier.



Creating a web page

Yesterday, Parand challenged everyone to create a web site. Here are some resources to make that easy. Primarily they are one-stop hosting sites where you can use a variety of templates to customize a landing page and some subsidiary pages. Some offer free plans; others offer first 30 days free.

Alternatives to LaunchRock 

A list on Quora of about 10 hosting software packages. Little information about each, but look in the comments. From 2015.

A list 0f 25, with descriptions written and paid for by the sites. It also has filters. It’s probably more up to date, because companies that vanish don’t pay. This links to the site, filtered to show only sites that include on-line host, A/B testing, and web based (they provide hosting).

If it’s not obvious by now, you need 4 things: One or more web pages (“content”), a server where you will host it (site), a URL, and software to easily build web pages (“package”). Usually you will also start with a template that has generic versions of everything you intend to put on your page. The template sets the overall “look and feel” of the page. I recommend finding a single web site to provide all of these things except your own content. At this stage there is no virtue in messing with hosting sites.

Things to consider

Domain name: Consider registering your own domain name at the very start. That way, if you change hosting sites you won’t have to change your URL. BUT registering a site is roughly $50 a year, and it may be prudent to start with a compound URL. (ucsdstartup.wordpress.com versus ucsdstartup.org, for example.)

Criteria: There are many criteria for choosing a hosting site. For now, ease-of-use ( mainly how easy it is to make changes) and low cost may be the lead contenders. Some of these sites are much more ambitious and designed to offer the ability, for example, to turn your site into a full store. Here is one list, in the form of a filter for sites.

  • A/B Testing
  • Analytics
  • Drag & Drop
  • Form Creation
  • Real Time Editing
  • Responsive
  • SEO Management
  • Templates

Reviews: Please write a few  reviews of sites/software that you look at. Rather than each team investigating all these sites, if you look carefully at a site please write a 1- 2 paragraph review and put it in the Comments on this page. We will gradually move them from comments to the main part of the page. I will open up ‘comments’ at least for a short while.  We will also request/require a review of whatever site you end up using, as part of your final deliverables.

Plan to throw one away: One way to make experiments faster and easier is to resolve, at the start, that you will start over midway through the project (rebuild your web page from scratch, with a new host). As a good friend used to say, “You will have to start over anyway; you might as well plan for it from the start.”

WordPress: WordPress.com offers its own ecosystem for building entire sites.  I use it. It’s overkill for what you need, but if you are already familiar with it, consider using it. I’m sure you can find several templates that fit your needs rather precisely.

Aesthetics and functionality: This site has thoughtful critiques of 15 landing pages froma variety of sites. His comments will help you think about how you set up your own page. For example, how to make best use of images.   Don’t look at this too early. First get something up on the web, no matter how crude.

Weekly status reports: hypotheses, pitches, tasks

We have prepared a template for each team to prepare weekly. It covers your current vision of your product/service,  your key hypotheses, and how you plan to test them. We provide some hypothetical examples to get you into the spirit applying these ideas, which were discussed last week, to your own situations.

The framework we are using is the “lean startup” approach, or more generally framing your knowledge as hypotheses, rather than facts or assumptions. You progress by identifying the most critical assumptions and testing them as early as possible. Some will fail the tests, setting off the process of rethinking what you are doing. Others will “pass” the test but still lead to recalibrating and adjusting.

Continue reading “Weekly status reports: hypotheses, pitches, tasks”

Random ideas for Meme T-shirts

This page is a forum for faculty to post ideas/suggestions in a form that everyone can see them. We may ultimately decide that another, more bidirectional, method is better. Such as email!

  • https://medium.com/@vacord/why-mechanical-turk-is-the-greatest-market-research-tool-you-never-thought-to-use-872120766ee. This guy owns a screen printing company. So his article could be doubly relevant.
  • Another article by the same guy.  https://medium.com/@vacord/validate-your-new-business-idea-for-just-fifty-bucks-605eb14bf64d
  • Picture sharing sites used by members of a target market could be a quick way to survey what people are interested in. Image search, with some coaching, can find images of people wearing T-shirts. It may be possible to use some beta software from Google or others to look for similarities among images, and in that way perform simple frequency counts. This will be a lot easier than scouring videos or physically walking around multiple cities, although each method is worth using as a check on the other methods.
  • How long does a meme last? Identify 10 to 50 memes of interest, then use online tools to count how often they are mentioned on specific web sites, week by week.