This page contains a description of some of the Marketing Analytics solutions we at Kmbara provide. Our Marketing Analytics solutions use advanced technology, machine learning and statistical models to help marketing professionals succeed.
Fundamentally, business-to-business marketing is about (1) finding potential clients, (2) convincing them to sign contracts, and (3) retaining them as clients. All of these are easier said than done. Finding potential clients is difficult not only because we may not even be aware of the existence of some companies, but also because it's hard to know whether they have the right characteristics to be suitable clients, and also the best way to contact their decision makers. Convincing prospects to become clients notoriously has a single-digit-percent success rate. Client retention is often a resource allocation issue, in which a few client managers have to work on retaining thousands of business clients, and need to know which clients are at greatest risk of terminating their contracts, so they can focus their efforts there.
In each of these cases, the professional expertise of marketing practitioners can be augmented, improved, and multiplied by software tools and rigorous quantitative analysis. We at Kmbara can provide the quantitative software tools that you need to succeed with your most difficult marketing challenges.
Our solutions use advanced software tools and advanced quantitative analysis to augment the efforts of marketing professionals. Here, we'll present some details about several of the most commonly needed solutions:
Clicking through Google searches and web links for hours to find suitable prospective clients can take hours and feel tedious. Today, any tedious task can be automated by suitably advanced software tools. We at Kmbara can create "web scrapers," sometimes called "crawlers" or "bots," that automatically visit websites and download and parse their contents. Scrapers consist of scripts written in programming languages like Python or PHP that are run on servers connected to the internet. They run automatically on regular, pre-set schedules, download content from websites, and can store the data and even generate reports about what they find.
Scrapers can be configured to visit any website, though of course many websites present unique challenges to scrapers, including security measures that prevent non-human visitors, terms of service the prohibit scraping, and content that's stored in formats that are difficult to parse.
One of the website's we've scraped before is Wikipedia. Wikipedia presents public data in a relatively friendly format. Our scraper can follow links between Wikipedia articles about corporations, so that starting with only one or a few prospects, we can expand to generate a much longer, global list. Most pages about prominent corporations have tables that summarize important information about them, including revenue, number of employees, and locations of headquarters. We can parse these tables and record whatever information is useful from them, to get full profiles of prospects. Wikipedia does not contain contact information (phone numbers, email addresses) of corporations, but other sites can sometimes be scraped to supplement profiles with that useful information. All of this can be done automatically with regularly scheduled scripts.
Scraping bots can be very useful, but we can go a step further by creating "screening bots." A screening bot will analyze the information that a scraper obtains, and check whether it matches the criteria you've set for prospects. This could be extremely simple - potentially a check against manually set criteria. It could also be more advanced, using machine learning tools to make an intelligent prediction about the likelihood of a prospect's suitability. A custom scraper/screener bot setup could make your whole process of finding and vetting prospects easy, reliable, and fast.
Like any business challenge, marketing requires wise allocation of limited resources. A business might put a huge investment into an email advertising campaign, and then be unsure about whether the investment was worth it, or whether there could have been some better use of the investment, or whether the performance was as high as it could have been.
The most elegant way to answer these questions is through an experimental approach commonly referred to as A/B testing. The idea of A/B testing is quite simple: we try two approaches simultaneously - for example, two different email campaigns - and compare the results (reads, clicks, conversions) from each approach. Whichever approach that we find has better performance becomes the one we use full-time in the future.
A/B testing requires some care to ensure statistical rigor. One extremely important element of the statistical design is random assignment: we create "A" and "B" groups that are the same size, randomly selected so that we expect that they don't differ in any crucial respects. Some shortcuts can undermine the statistical rigor that comes from random assignment. For example, we may send our first 1000emails to group A on Thursday, and our second 1000 emails to group B on Friday, and expect to be able to compare results from groups A and B to draw conclusions about our email designs. However, if we notice a difference between Group A and Group B response rates in this scenario, it may not be a measurement of the efficacy of different email designs, but rather a measurement of how people react to emails on different days of the week. True random assignment requires that each group receive emails on all of the days during which emails are being sent out - again, to ensure that Groups A and B differ only in the emails they're receiving and not in any other essential respect.
After random assignment and the collection of results, we can use statistical hypothesis testing to measure whether the differences between group reach the threshold of statistical significance. There are many tests we could use to determine this. Among the most common of these tests is a two-sample proportion test (a type of t-test), which can be implemented simply and easily, given the right data.
More advanced tests can be performed for some purposes, but a typical A/B testing scenario is quite simple from the mathematical perspective.
When a series of A/B tests are run in a row for the same business challenge, we have the makings of a "champion/challenger" scenario. One approach - for example, one email campaign - begins as the"champion," and an A/B test compares this champion to a "challenger." If the champion performs better than the challenger, it remains champion, but if the challenger does better, it becomes the new champion. In the next round of testing, the new champion is challenged by yet another challenger. As successive challengers continually dethrone strong champions, we arrive, via natural selection, at a champion that is extremely powerful and performs extremely well.
A/B tests can be run to optimize performance in a variety of scenarios: email campaigns, landing pages, advertising copy, cold calling scripts, website designs, postcard content, UI/UX choices, and more. They provide a simple way to make decisions between complex alternatives where final answers are never obvious. By finding the right approach to get the best results, their one of the most powerful optimization tools available to businesses today.
Marketing professionals have enough on their plates, and sometimes struggle to keep up with the many demands placed on them. We want to make their lives easier: don't do all that hard work yourself, but rather let our bots scrape the web for you, and let our scientists design rigorous A/B tests for you. Contact us today to talk about how we can help you with marketing analytics.
Kmbara © 2019 - 2020