QBR settings: CSAT sources, metrics and AI

Three tabs of the QBR settings decide what the numbers in a review mean and how the AI behaves around them: CSAT sources, Metrics and AI. A single Save settings button on the settings page saves all six tabs together. How to open the page, save and discard is described on QBR settings: brand, workflow and HaloPSA items, which also covers the other three tabs.

Before you start: the QBR Reporting module must be on for your organisation, and you must be an Administrator or a Service Desk Manager. The numbers themselves are listed on the QBR metric library, where you can try any of them for a client after a change.

Choose where satisfaction scores come from

Open the CSAT sources tab. Every source you turn on is pooled into one CSAT score per client. Each source keeps its own scale and is normalised to 0 to 100 before pooling. If no source is on, the CSAT metrics show as unavailable.

1. Under Where satisfaction scores come from, tick QuantumOps surveys to use the ratings from the CSAT faces in your HaloPSA e-mails, matched to the client of each ticket. It is on by default.
2. Tick Include NPS campaigns to report NPS as well. NPS is reported on its own and is never mixed into the CSAT score. It is on by default. See NPS campaigns.
3. Under Halo satisfaction, tick Use it to include the score that HaloPSA stores on each ticket from its own feedback survey. It is off by default. Then:
   • Under Which end is best, choose whether 1 is the best score (HaloPSA's default runs from 1 Excellent to 4 Bad) or the highest number is the best.
   • Set Number of levels, the highest score on your HaloPSA scale, from 2 to 10. HaloPSA ships with 4.
   • Select Show score distribution to see the scores HaloPSA has recorded over the last year, with the best and worst ends marked. Use it to check that the end you chose is the one your happy customers pick. If no closed ticket in the last year carries a score, the page says so.
4. To include a rating tool that writes into HaloPSA custom fields, follow the next section.
5. Select Save settings.

If your reviews used the earlier Crewhu keyword setting, a notice at the top of the page says so and links to CSAT sources. A Crewhu source is waiting there, switched off, for you to finish it. The notice goes away the next time you save.

Map ratings from other tools in HaloPSA custom fields

Tools such as Crewhu and Simplesat write each rating onto the ticket as a HaloPSA custom field. Map the fields once and their ratings join the pool.

1. Under Ratings in Halo custom fields, select Load Halo fields. The page tells you how many ticket custom fields it found; start typing in a column box to pick one.
2. Select Add Crewhu, Add Simplesat or Add another tool. Crewhu starts as a word-or-face scale and Simplesat as a number from 1 to 5.
3. Check the Source name, then enter the Rating column. Optionally enter a Comment column and a Response date column. Columns are HaloPSA ticket custom fields written as CF followed by the field name; you can type the field name with or without the CF. Without a date column, the ticket's close date is used. The rating column must be a ticket custom field that HaloPSA lists. Saving checks only the shape of the name. If HaloPSA does not have the column, the source is skipped when a review is generated and the satisfaction figures use the sources that remain; if none can answer, they show as unavailable with the reason.
4. Next to The rating is, choose how the rating is stored:
   • A number: set Lowest and Highest, and tick Lower numbers are better if a low number is the good end.
   • A word or face: select Sample values (available once Rating column holds a valid column name) to read the distinct ratings from the last 180 days of tickets, then mark each value as positive, neutral or negative. A value you leave unmapped is ignored. To add a value by hand, type it and select Add value.
5. Keep Pool these ratings ticked so the source is used. Untick it to keep the mapping but leave the source out. To delete a mapping, use its trash icon.
6. Select Save settings.

Positive ratings count towards the CSAT percentage, and net CSAT is positive minus negative.

Set the minimum responses before a score shows

Under Scoring, Minimum responses is the number of responses a client needs before its score is shown, from 1 to 100. The default is 5. Below that, the score shows as "not enough responses" instead of a number that could swing on a single survey.

Set the reporting time zone and business hours

Open the Metrics tab. Under Time, periods start and end at local midnight, and the last day is always counted.

1. Pick the Time zone. The first choice uses your organisation's time zone. This time zone also decides the dates on QBR schedules.
2. Set Business hours (the default is 08:00 to 18:00). Tickets logged outside these hours, or on a day that is not a business day, count as after hours.
3. Switch the Business days on or off. The default is Monday to Friday.

Say which tickets count as closed or cancelled

Under Which tickets count as closed, HaloPSA's Resolved and Closed statuses always count as closed. Instances add statuses of their own, so name any others here.

1. Select Load from Halo to fetch your HaloPSA statuses and ticket types as buttons. If you have not loaded them, you can type status numbers separated by commas.
2. Under Other closed statuses, pick the statuses that mean a ticket is done.
3. Under Cancelled statuses, pick the statuses that mean a ticket was cancelled. Cancelled tickets count as opened but never as resolved, so they do not flatter resolution times.
4. Leave the option "Also treat any status named resolved, closed or completed as closed, and cancelled as cancelled" ticked if you want statuses recognised by name as well. It is on by default.

Choose what is counted

Under What is counted:

• Leave out these ticket types removes ticket types from every metric. Opportunities and projects are always left out of service metrics.
• Sources that mean "raised by us" is a comma-separated list of ticket source names that count as proactive work, such as monitoring and automation. The default list is Monitoring, Alert, Automation, RMM and Integration.

Set SLA and satisfaction targets

Under SLA and targets, gauges and scorecards colour against your targets.

1. Enter Response within SLA and Resolution within SLA as percentages. The defaults are 95 and 90.
2. Enter CSAT as a percentage. The default is 90.
3. Leave a target empty for no target.
4. Decide whether clients see service-level figures with Show SLA metrics to clients, on by default. When it is off, SLA blocks are skipped when a review is generated and are hidden from everything the client sees. Your team still sees those metrics in the metric library.

Turn AI commentary on or off and set its tone

Open the AI tab. AI writes the words around the numbers. It never produces a number, because every figure comes from the metric library.

1. Under AI commentary, use On to switch the commentary on or off. When it is off, AI blocks show "AI is turned off" and every data block still builds. The assistant in the template and review editors is also unavailable until you turn it back on.
2. Choose the Tone:
   • Candid and constructive celebrates wins and names problems plainly, each with a plan. It is the default.
   • Partnership advocate leads with the value delivered and frames issues as initiatives already under way. It suits renewals and relationships that need reassurance.
   • Neutral analyst puts facts first, with few adjectives and no framing. It suits clients who want the numbers and little else.
3. Optionally write a House style, up to 4,000 characters, that is added to every AI block's instructions. For example: "Write in British English. Refer to us as 'the service desk'."
4. Select Save settings.

Let AI blocks look things up

Under Evidence the AI may gather, prompt blocks whose option "Let the AI look things up with connected tools" is ticked can use read-only tools to find backup, security and licence data in your connected sources. They can search but never change anything. The option is set per block in the template or the review.

1. Tick Allow. It is on by default.
2. Under Tool sources, pick the sources the AI may use. These come from your MCP sources (see MCP sources). If you pick none, every source the run can reach is allowed; if you pick some, only those can be used. A scheduled draft that runs as the organisation reaches only organisation-wide sources, and one that runs as its owner also reaches the owner's own. See Run as on QBR schedules.
3. Set the limits for one block. Tool calls per block runs from 0 to 25 and defaults to 6. Seconds per block runs from 15 to 600 and defaults to 180; a block that runs out of time is marked for review, not left half written. Tokens per block is the most that one block's look-ups may use: enter 0 for no cap, otherwise a number from 5,000 to 1,000,000. The default is 60,000.
4. The option "Use what QuantumOps remembers about the client (the memory graph) as evidence" lets look-ups read what QuantumOps has learned about the client. The screen says "memory graph" for what these pages call support memory. The option is on by default and has an effect where support memory is enabled for your organisation. See Support memory.
5. Select Save settings.

Use Design with Claude

Design with Claude is an extra export. Claude builds a designed PowerPoint copy of the approved review and keeps every number as it was reviewed. It is off by default.

1. Under Design with Claude, switch On.
2. Read the notice on the card. When someone uses the export, the client name, the numbers, the commentary and the speaker notes of the visible slides are sent to Anthropic using your organisation's own API key. Images and your logo are not sent. The export uses Anthropic's code execution tool, which is not eligible for zero data retention. Leave it off if your agreements with clients require otherwise.
3. Look at the key message under the notice. The page says whether your organisation has an Anthropic API key. Without one, the export stays unavailable until an administrator adds a key to your organisation's AI settings (see AI providers and models).
4. Choose the Claude model by entering a Claude model name; the field starts with the default. Optionally write a Design direction, up to 2,000 characters, which is sent with every export. For example: "Minimal, plenty of white space, our brand colours, one idea per slide."
5. Select Save settings.

How the export is run from a review's Export menu, and what it sends, is in Presenting and exporting a review.

Related pages

• QBR settings: brand, workflow and HaloPSA items covers the Brand kit, Workflow and Halo items tabs and how saving works.
• QBR metric library shows every number these settings change.
• NPS campaigns explains where NPS scores come from.
• MCP sources explains the tool sources the AI can use.
• Presenting and exporting a review explains the export menu.