写在前面（attribution modeling 相关知识科普）

归因模式（attribution modeling）是一套规则，这套规则决定了一次购买或者说转化过程中每个触点（touch point）的贡献分配。最基本的归因模式有：last interaction （也称last click鼠标最后点击）；first interaction（也称first click鼠标最初点击）；linear（线性分配）；time decay（时间衰减分配）；position based（基于位置的分配）。下面小编给各位读者举个例子具体说明以上五种归因模式。

例子：

顾客A点击了B商户的某一个互联网广告，找到了B的网站。一个星期后，A通过社交网络渠道又到了B的网站。同一天里，A通过一封B的推销邮件在B网站进行了一次购买。那么，根据不同的归因模式，这三次触点对最终购买的贡献分配如下：

1、遵照last interaction模式，最后一次触点——那封邮件——做了100%的贡献。

2、遵照first interaction模式，第一次触点——那个网络广告——做了100%的贡献。

3、 遵照linear模式，这三次触点的贡献率各占33.3%

4、遵照time decay模式，距离购买行为时间最近的触点贡献最多，在此例中，即email和social network的贡献率最大，因为这两次触点跟购买行为是同一天，而那个互联网广告则只占了较小的贡献。

5 遵照position based模式，那么第一次触点和最后一次触点贡献最大，均为40%，第二次触点为20%。（这种模式适合最看重第一次触点和最后一次触点的广告主）

可以看到，根据不同的归因模式，每一个触点的贡献率是不同的。那究竟该选择何种归因模式才能最终优化广告投放，提高ROI成了营销人员最头疼的问题。这次谷歌发布的归因模式工具一方面可以给营销人员提供各个基本归因模式的报告，也可以让营销人员自定义符合自己需要的归因模式。

以下为新闻稿内容：

Google recently launched an update to its DFA reporting suite with the addition of attribution modelling capabilities natively into the DFA interface. Beyond the basic models (namely: linear, first position, last position, time decay, etc.), DFA also offers the ability for users to build their own custom model to manipulate the rules of the model itself.

谷歌近期宣布升级其DFA（doubleclick for advertiser）报告套件，具体来说，就是在DFA界面集成归因模式功能。除了基本的模式（即：线性分配，最先鼠标点击，最后鼠标点击，时间衰减等归因模式）外，DFA还能让用户自定义归因模式，进而自由操纵自定义模式的规则。

By offering it as a value-add within an existing ad server, it may force existing attribution software clients (who currently deploy the likes of c3metrics, Clearsaleing) to review their overall investment in attribution modelling.

这次升级只是作为在原有广告服务器上的一次增值服务而非新的解决方案。通过这次升级，它可以让现有的归因软件客户（即像c3metrics和clearsaleing的客户）审查自己所有的归因分析投资情况。

Many agencies have been investing in developing their own proprietary solutions, often requiring the need of some form of ‘big data’ solution to execute this. Will Google’s recent roll-out be seen as a positive for the industry? Will it help to increase more scaled adoption of attribution modelling as a practice by marketers? Will it devalue the internal tools being developed by agencies? We asked some industry leaders what they thought of this. The general consensus seems to be that, whilst added analysis is agreed to be a good thing, in the hands of non-analysts, the benefits could range from slim to adverse:

许多代理商在开发自有解决方案时，需要用到一些“大数据”解决方案。那么谷歌的这次升级是否能给归因分析行业带来积极有效的作用？是否能大范围促使营销人员接受归因分析服务？它是否会造成代理商开发的内置归因软件贬值？带着这些问题，让我们一起来看看业内权威人士的评价：

Samuel Watts, Associate Director, Starcom MediaVest Group,“This has certainly led to plenty of discussion our end. Anything that helps to move the industry away from judging on the last click is a good thing, although it is important that we do not move from one arbitrary model to another without thinking about the role that each channel and site is playing in the media mix. Adding an ability to use a rules-based approach will obviously help the adoption of attribution in DFA, however in our experience, generating the attributed results is the easy bit. The more difficult, and unsurprisingly more valuable, bit is having the expertise to analyse the results to make solid recommendations to increase ROI. This is the difference between a reporting and an actionable insight. At Starcom MediaVest Group we see the need for an automated attribution model as one aspect of best practice digital evaluation. Hidden within the web logs are a treasure trove of insight, the challenge for our analysts is getting to gems to inform better planning.”

Samuel Watts, Associate Director, Starcom MediaVest Group：

“我们这边一直在讨论谷歌的这次升级。我认为，任何有助于改善‘最后点击last click’模式的东西都是好的。但是我们不能从一个极端转移到另一个极端，根本不考虑每个渠道、网站在整个媒介组合中的角色，这一点是很重要的。这种基于规则的升级很明显可以提高DFA归因分析市场接受度。但是，以我们的经验来看，生成归因结果报告小菜一碟。更难的同时也是更有价值的部分是运用专业知识能分析结果并提出有力的投放建议进而提高ROI。这就是简单的报告与实际洞察结果的区别。在starcom集团，我们需要一种自动的归因分析模式作为最佳的数据评估。网络日志背后的是一个巨大的意见宝库，我们分析师所面临的困难就是挖出这些有洞察力的意见去做更好的规划。”

Kate Tickner, Business Solutions Executive for Big Data Solutions at IBM,“There aren’t many companies that can match Google for investment in data analysis tools, so in my opinion many will welcome enhanced analytical capabilities within their existing DFA solution. Others are likely to be less enthusiastic as they will see this as yet another way to extend Google’s hold on the market place. Amongst the fans however, it will probably be a subset of people will be able to use analytics effectively or accurately. There is a balance to be struck between making analytics accessible and ‘dumbing down’ sophisticated analytical techniques too far. Users without the right skill set could recommend actions based on inaccurate data or flawed models. Organisations who are basing campaigns on the results of any complex analyses need to have checks in place to ensure they understand the quality of the base data and the transformations it has been through during the modelling process.”

Kate Tickner, Business Solutions Executive for Big Data Solutions at IBM：“事实上，没有几家公司在数据分析工具方面的投入能跟谷歌相提并论。所以在我看来，很多公司会乐意接受谷歌在原有的DFA解决方案上增加这项功能。而那些不愿意看到谷歌继续扩大市场占有率的竞争公司会表现失望。在所有追捧者之中，可能只有一小部分人能够准确有效地使用分析工具。在让分析技术受到普遍接受和“傻瓜化”复杂的分析技术之间要做到一种平衡。没有掌握正确使用方法的用户会根据错误的数据或有缺陷的模型作出推荐行为。那些根据复杂分析结果制定广告campaign投放策略的公司，需要建立不间断的审计功能以确保他们能了解建模过程中基本数据的质量和这些基本数据的转型。

Peter Wallace, Head of Performance, Total Media,“There is still an air of mystery amongst clients as to how these attribution models are actually created and therefore how they benefit them. Essentially, Google is providing too much autonomy to agencies by allowing them to develop custom models. To increase the pace of development for attribution modelling, you actually need a much greater level of education and an industry-accepted standard. This product will produce greater client uncertainty and allow parties without the appropriate levels of expertise to create models which could in fact be inaccurate. Will this product devalue the agencies’ proprietary technology? No, not at all. In fact, it will accelerate the development of tools across the industry. Agencies will strive to add value through their experience and expertise until attribution modelling is fully accepted by clients and represented across every plan.”

Peter Wallace, Head of Performance,Total Media：“客户们现在仍在疑惑归因模式究竟是怎么形成的，也不知道归因模式究竟能给自己带来怎样的利益。而谷歌这次升级，给代理商的自由度太高，他们可以自定义归因模式。要加快归因分析模式的发展，你必须要有很扎实的教育背景而且要有一个行业普遍接受的标准。这个产品会加重客户的疑惑，也会让一些没有相关专业层次的公司建立实际是错误的模型。不过，关于这个产品会不会使广告代理商自主技术贬值这个问题，我觉得不会。实际上，这个产品可以加速业内归因分析工具的发展。它会让代理商不断努力在自己的产品中加入他们的经验和专业直到自己的归因模式完全被客户接受并且在每一次广告投放计划中得到体现。”

Andy Mihalop, Head of Digital,Moneysupermarket,“The launch of Google’s attribution modeling tool is a positive step for marketers and builds on the existing path to conversion functionality in terms of custom modeling. This is undoubtedly aimed at SME marketers, who don’t have substantial media budgets and therefore the opportunity to leverage existing media agency or pure play attribution management solutions. I can’t see this gaining much traction with more sophisticated advertisers who require a more customised solution, such as integrating offline data. The other challenge is data analysis. Regardless of the solution, advertisers will still require skilled data analysts in order to leverage and action the insight. That may be tough for SME’s.”

Andy Mihalop, Head of Digital,Moneysupermarket：“谷歌发布这样一个归因分析工具在我看来是对营销人员定制化的归因转换路径的正面肯定。这个产品无疑是针对中小企业的营销人员设计的，因为他们没有充裕的媒体预算因此也没有机会利用现有的媒体代理或是纯归因管理解决方案。我不认为这个产品能带来更多的要求繁复的广告主，因为他们需要更加高级的自定义解决方案比如整合线下数据。另外一个问题是数据分析。撇开解决方案不谈，广告主为了能利用和实践有洞察力的建议还会要求专业数据分析人员做支持。这对中小企业营销人员来说是很难做到的。”

小编总结：虽然各方人士都认为谷歌推出的新工具对提高归因模式产业接受度有积极作用，但是大家都普遍认为，这个工具给营销人员的自由度太高，如果背后没有专业的建模分析人员支持，反而会给不明就里的广告主带去适得其反的效果。

注：科普内容译自google analytics；新闻稿内容译自exchangewire.com