Purpose
This article answers commonly asked questions when analyzing InnovidXP data.
Are custom reports available?
Yes, InnovidXP provides a selection of custom charts within the Analysis tab of the platform. If the chart you need is not there, contact your Innovid representative or submit a ticket here (or through your brand portal).
Why do sessions show as decimals?
Note: This question and answer relate to spike modeling.
Sometimes sessions display with decimal places instead of whole numbers in the Pivot tables, for example:
This can happen because these results are based on probabilities, since we do not know exactly who has landed on the website due to the TV ad. Our unique methodology attributes a probability score to each user who has landed on the site above the baseline.
For example, in the graph below, 80 people landed on the site after the spot aired, but 20 of them fell within the baseline. We don’t know exactly who has fallen into the baseline and who has been driven to the site by the ad, so we apply a probability percentage to every visit. In this simple case, we would apply a probability of 60/80 or 75% to all visits:
See Linear Spike Model Campaigns for more details on our methodology and a full visual picture.
Why is my response lower than 1 and my cost per response extremely high?
Note: This question and answer relate to spike modeling.
Each attributed response is assigned a probability, a value between 0 and 1; therefore, it is often 0.5. After each session is assigned a probability, we sum these probabilities to obtain an overall attributed response number, which is usually greater than 1.
In the case you have seen, the spots have received very little response, or, more accurately, the response after the spot has a very low probability of being attributed to the spot. This results in an overall response that is below 1; therefore, you will see a spot with a 0.01 response attributed to it.
If the cost per response is high, it is due to the basic mathematical rule of dividing by a decimal. Cost per response is calculated by dividing the cost by the number of responses (cost/response). This works well when the response is 1 or above, but in a rare example like this, where the response is a decimal, we get something like the following:
CPR = (cost = 100 / response = (1 / 100))
To perform this calculation, we multiply the cost by the denominator. Therefore, the formula actually looks like this:
CPR = (cost = 100 x inverted response = 100/1)
Why doesn't my spot count match my file?
This discrepancy can occur due to errors or when the system detects duplicate spots. Your confirmation email will specify the exact issue. Duplicate spots are defined as spots that occur simultaneously on the same channel. Deduplication is determined by date, time, channel, and market (if available). When detected, these are merged to combine cost and audience values while retaining data from the topmost row.
Why is my audience multiplied by 1000?
This can happen if you selected thousands as the audience metric when configuring your file settings and your spot file contained individual numbers.
To fix this, re-upload the spot file and select individuals as the audience metric.
Why is it important to group and compare like-for-like inventory?
If you use the InnovidXP spike methodology, you will see an immediate response only about 10-15 minutes after the ad airs. We recommend grouping similar channels together and optimizing within those groups.
Consider the following:
- Bigger brand channels & spots could provide more response in the long term.
- Advertising is more expensive on-brand channels or during peak periods, so expect a higher CPR.
- Content on peak times and brand channels tends to be more engaging.
- Peak time and brand channels provide much greater reach.
To ensure efficient inventory comparisons, we recommend creating like-for-like inventory groups and benchmarks based on relevant KPIs for each group.
For example, if you compare daytime vs. peak inventory, you can expect different results. Daytime tends to look efficient, but peak is better for long-term reach. For daytime, you could optimize for efficiency (cost per response), and for peak, for reach.
Note: Our recommendations are based on historical data in the InnovidXP platform.
Since we are unsure what the future holds, any changes in external factors are not known to the platform, for example, changes in (but not limited to) the following:
- Media landscape
- Industry
- Seasonality
- Inventory availability
- Inventory cost
How much data do I need before I can optimize?
The more, the better, but a week is a good start.
A week allows you to see daypart or hourly trends over that period. Of course, two or three weeks allow you to spot better trends across your spot dimensions, as you have two or three samples of each day of the week to gather insight.
Can I plug the data into a visualization platform like Tableau, Qlik, or MicroStrategy?
Yes, you can set up manual or automated exports that can be sent via email, SFTP, or to an Amazon S3 bucket. These exports will allow you to input the data into your desired visualization platform.
Contact your Innovid representative or submit a ticket here (or through your brand portal).
What is a Report in InnovidXP?
A Report is a customized chart from your Pivot table data that you can save and export. You can create and save customized charts of your Pivot table data to send in report format.
You can have multiple rows and columns, and column splits to show how the response details break down.
The Report view allows you to use a Pivot, create a more 'typical' table view, and reorder the metrics (in a Pivot, metrics appear only in the order added).
Where do I find advanced reports in InnovidXP?
The Analysis tab provides a complete range of predefined reports and lets you create your own outputs.
How to interpret your results
| Strategic Action / Goal | Recommendations & Guidelines |
| Build custom reports, e.g., compare channels |
➡️ What does the TV converter look like?<br> ➡️ Are you targeting the best audience? |
| Redefine the metrics to improve campaigns | ➡️ Look at using a different URL mix to see where people are exploring on their first visit |
| Compare drop-off rates and user drop-off rates |
➡️ Identify where the drops happen<br> ➡️ Split the user journey down further to gain better insights |
| Check advertisers with low response rates | ➡️ Look at the response volumes to optimize campaigns |
| Provide quick, regular, and consistent reporting | ➡️ Report at the same times regularly to build up an effective analysis |
| Use charts effectively to identify performance | ➡️ Which weekdays and times are performing well and which are not? |
| Are you using both brand and DRTV channels in the spot file? | ➡️ Include a breakdown by channel type to give you the option to filter by each group and see a like-for-like comparison |
| Compare creatives | ➡️ Are you running or planning to run any tandem spots or any A/B testing? |
| Check adverts that promote a unique action, e.g., donation or site visit | ➡️ Does your creative call to action push the audience to act as required? |
| Review device and medium use | ➡️ How are people responding? |
| Consider benchmarking similar clients or a call to action | ➡️ Gain comparative like-for-like insights |
| Use multiple dashboards for different teams | ➡️ Build to individual team specifications |
| Automate reports | ➡️ Plug the data into internal warehouses or data visualization tools |
| Ongoing training | ➡️ Offer InnovidXP training to new employees and refresher/Q&A sessions for current staff |
How to change the date range
All data displayed in InnovidXP is determined by the selected date range, which you can find just beneath the main menu:
The set date range defines the data volume available for analysis, so in the date picker, you will also see a top-line summary of your spot and response data for the selected range.
- To view the spot response and cost summary for your date range, select the earliest date on the left and the latest date on the right.
- To switch the start/end month, click the listed month, then either select a month from the revised calendar menu or scroll using the keyboard's back and forward arrow keys.
- To switch start/end year, select ‘month,’ then select ‘year’ to expand, or select ‘month,’ then use the keyboard back and forward arrows to select a year.
- There are two quick-jump options available: select the last active week or the last 4 active weeks to view data for a specific time period.
Troubleshooting web response volume issues
Sometimes web response and action volumes displayed in the platform do not match the numbers reported by a third-party analytics site (e.g., Google Analytics). We call this "response volume discrepancy," but it does not necessarily mean that InnovidXP is processing your data incorrectly.
Note: Our JavaScript-based web tag measures total website visits. Only one visit tag is needed per website, and it must be implemented on all pages to measure total visits.
The web and action tags fire on the page(s) where they are placed, while complying with the required opt-in/opt-out mechanisms. Any comparison of web volumes and a third-party tool is on a like-for-like basis, as the tag is intended to interact with a consistent consent criterion.
The web manager should place web tags on the client's website. We recommend installing tags two weeks before your campaign goes live to allow time for data QA.
Contact your Innovid representative or submit a ticket here (or through your brand portal).
Compare volumes
You can compare and validate web response volumes via the Raw Response Breakdown page; select Data>Responses>Raw Response Breakdown:
Daily: When you compare volumes by day, InnovidXP should be within 7-10% of the third-party analytics volumes. It is important that the lowest day in InnovidXP is the lowest day on your third-party analytics tool.
Hourly: It is also worth comparing volumes on an hourly basis for one or two days to check if the trends align, i.e., whether the data peaks and dips at the same time.
Volume discrepancy issues
There are several possible reasons why web visits and user action volumes are not exactly the same in other systems:
- Differences in how sessions are classified across systems: Innovid defines a "session" as a user interacting with your website that expires after 30 minutes of inactivity. When a user arrives on your site, we record the time and set the session to expire in 30 minutes. If 30 minutes pass without any user interaction, the session ends. However, each time the user interacts with the website, the session timer is reset to 30 minutes from that moment, rather than being added to the remaining time.
- Note: To change any default referrer classifications, add one or more rules using regular expressions (what is regex?). See Setting Web Traffic Classifications for more information.
- De-duplication of multiple instances of an action in the same session: Innovid measures counts containing an action, not the sum of total numbers.
- Differences in time zones between systems: The time zone for raw response is platform-specific. Check with your Innovid representative to confirm the time zone for volume comparison.
- The difference in geolocation is that Innovid uses IP addresses, while others may use domain names.
Low volume of actions
If the volume of actions appears significantly lower in Innovid, check the following:
- Whether there are multiple versions of the action page, for example, if the user can check out on different parts of the site, and if so, whether all appropriate pages have been tagged.
- Whether there is a separate mobile site, and if it has been tagged.
- Whether the other system measures the sum of conversions rather than the count of sessions containing a conversion.
Troubleshooting impression pixel volume issues
Sometimes, impression volumes displayed on the platform do not match the number of impressions served by you or your ad server. We call this "impression pixel discrepancy," but it does not necessarily mean that InnovidXP processes your data incorrectly.
This section outlines why this can happen and how to avoid it, so you can be confident that any discrepancies that may occur do not affect your overall data quality.
Use this checklist as a reference pre-flight so you know what to expect. If you have any questions, contact your Innovid representative.
Reasons for impression pixel discrepancies between Innovid and ad server
- Impression counting differences: Ad servers have different methods for counting impressions and filtering valid impressions. InnovidXP counts an impression regardless of whether it was filtered.
- Ad blockers: Ad-blocking software can prevent the line item from being delivered; however, the impression pixel has fired, so InnovidXP counts the impressions.
- Different geography definitions: For RoW clients, other ad servers map IP address location data differently, leading to significant discrepancies.
- Latency: The lag between an initial line item and the creative's appearance can affect counts.
- Network connection and server reliability: An ad server may fail briefly or encounter an issue that prevents it from logging an impression.
Note: The acceptable difference between counts is 10%.
During XP Campaign Setup
While you are setting up your Campaign:
- Ensure the Campaign Start Date matches the date the impression pixel was implemented.
- Ensure the pixel is appended to ALL creatives.
- Check whether the pixel has been cut off during implementation.
- Check whether your macros are hard-coded, i.e., the dynamic value is not being passed.
General OTT/CTV Limitations
Be aware of the following limitations for OTT/CTV data:
- Innovid can only capture video impressions (no display impressions will be reflected in the platform)
- For all OTT/CTV Partners, Innovid cannot do the following:
- Measure YouTube inventory.
- Measure Sky (UK) and ITV (UK) inventory.
- Measure set-top box/VOD inventory.
- Track Facebook videos.
- Track publisher-level data.
- Track impressions in "walled garden" sites, such as Amazon, YouTube, and Facebook.
- Track impressions for IPV6 devices.
Publisher-Specific Limitations
Be aware of the following limitations for each publisher listed:
| Publisher | Limitation details |
| Viacom / EyeQ | We cannot measure set-top box / VOD inventory. This partner's inventory tends to run on Charter Communications, so we cannot measure their performance. |
| Roku | Creatives and Placement Groups come through in Base64 and therefore need to be decoded to perform mappings. |
| Xandr | Cannot measure Roku inventory. |
| Disney | Cannot provide publisher-level data, so all publishers will display as "Disney". |
| Discovery / Warner | Include YouTube inventory in their OTT campaigns that InnovidXP cannot measure. |
| Extreme Reach | Some impressions are served in an environment that does not allow multiple trackers and only gives counts to Extreme Reach and InnovidXP, e.g., Google and YouTube. |